Accurate and Efficient Cleavage of the Human Insulin Proreceptor by the Human Proprotein-processing Protease Furin CHARACTERIZATION AND KINETIC PARAMETERS USING THE PURIFIED, SECRETED SOLUBLE PROTEASE EXPRESSED BY A RECOMBINANT BACULOVIRUS*

Maturation of the insulin proreceptor in a late Golgi compartment requires cleavage at an Arg-Lys-Arg-Arg processing site, suggesting involvement of furin, a trans- membrane serine protease of the Kex2 family of processing enzymes. A genetically engineered secreted, soluble form of human furin (ss-furin), expressed by infection of insect cells with a recombinant baculovirus, was purified to near homogeneity. ss-Furin exhibited rapid and efficient cleavage of both isoforms of the human insulin proreceptor in solubilized extracts of cul- tured mammalian cells expressing preproreceptor cDNA. Proreceptor cleavage occurred at the physiolog- ical processing site as judged by the effects of mutations in this site on cleavage by purified ss-furin. Moreover, purified ss-furin exhibited specificity for proreceptor cleavage identical to that of the endogenous insulin pro- receptor-processing enzyme. Furin thus displays the properties expected of an insulin proreceptor-process-ing enzyme in that it (i) cleaves the proreceptor efficiently and at the correct site; (ii) exhibits the same specificity in processing variant proreceptors ATG ATG pUN70 inserted into the Autographa californica nuclear polyhedrosis virus (baculovirus) transfer vector pVL using BclI adaptors (311, creating plasmid pVL-fur-595. The insect cell line Sf9 (an ovarian cell line of the fall armyworm Spodoptera frugiperda) cotransfected with pVL-fur-595 and linear wild type baculovirus genomic DNA (Invitrogen Corp.) for recombination in uiuo. Recombinant baculovirus (bac:fur-595) in the was isolated from wild type baculovirus by four rounds of plaque purification, and a high titer stock was obtained by infection of Sf9 cells. Hi5 cells (ovarian cell line from the cabbage looper moth, Tkzchoplusia ni, Invitrogen Corp.) were infected with either wild type (bac:wt) or bac:fur-595 viral stocks at a multiplicity of infection of two, and media samples (300 pl) were analyzed for ss-furin expression.

Two isoforms of the insulin receptor differ by the presence or absence of a 12-residue segment 3 residues amino-terminal to the proreceptor processing site, the result of alternative splicing of exon 11 in the hIPR pre-mRNA(h1R-11 lacks and hIR+11 has exon 11) (10). The two are expressed in a tissue-specific fashion. hIR-11 alone appears to be expressed in lymphocytes, whereas both hIR-11 and hIR+11 are expressed in liver, kidney, and muscle.
Physiological importance for proteolytic maturation of hIPR was established with the discovery of two sisters presenting extreme insulin-resistant diabetes, apparently due to substitution of Ser for Arg,,, at the P12 position of the proreceptor processing site (11,12). Additionally, in an animal model for diabetes, a defect in proreceptor processing was indicated from an increase in the insulin proreceptor to receptor ratio in ketotic rats (13). It is unclear precisely how processing affects receptor function, and the hIR-11 and hIR+11 isoforms may differ in this regard (14). Proreceptors acquire insulin binding capacity before proteolytic cleavage in the late Golgi, indicating that processing is not absolutely required for binding (15). Epstein-Barr virus-transformed lymphocytes derived from one patient with the Ser substitution at the P, position in the processing site exhibited only proreceptors, presumably the -11 isoform, at the cell surface. These cells exhibited dramatically reduced insulin binding and signal transduction. In contrast, when a series of substitutions was made in the hIPR+11 type receptor (see below), uncleaved proreceptors exhibited near normal insulin affinity, although somewhat higher levels of insulin were required for activation of receptor tyrosine kinase activity. Thus, the absence of processing may have a more drastic effect on the function of the -11 isoform than on that of the +11 isoform.
Residues amino-terminal to a cleaved bond are designated P,, P,, P,, and P4, etc.
Mutations created in the hIPR processing site revealed that Arg was essential for cleavage at the P, and P4 positions. Substitution of Ala for Lys at P3 or Arg at P, had little or no effect on cleavage of hIPR (16). This pattern of specificity (-Arg-Xaa-Xaa-Arg-) matches that of the recently discovered protease, furin (17)(18)(19). Human furin was the first member of a family of mammalian processing enzymes identified by homology to yeast Kex2 protease, a Ca2+-dependent serine protease that cleaves the a-mating pheromone precursor at -Lys-Arg-sites (20)(21)(22). Expression of the fur gene, which encodes furin, appears to be ubiquitous, although the level of expression is developmentally modulated (23). Colocalization with TGN38 and failure to redistribute to the endoplasmic reticulum in the presence of brefeldin A suggest localization of furin in the truns-Golgi network (241, although it should be cautioned that experiments have been performed only with high levels of furin expression. The nearly ubiquitous expression of both furin and hIPR, the correlation between furin localization and the cellular location of hIPR cleavage, and the nature of the hIPR processing site make furin a likely candidate for the physiological hIPR-processing enzyme. Furin has been implicated in processing the profusogens of several lipid enveloped viruses including fowl plaque hemagglutinin (25) and the 160-kDa glycoprotein from HIV-I (26) and so is a potential target for antiviral drugs. Knowledge of furin's physiological roles will become important in assessing the consequences of chronic or acute inhibition of the enzyme.
This paper presents biochemical evidence that furin processes the insulin proreceptor. Using recombinant baculovirus expression, we have expressed a secreted, soluble form of human furin (ss-furin) and purified the enzyme. We have found that the purified enzyme cleaves both isoforms of insulin proreceptor efficiently and in a Ca2+-dependent manner. Finally, site-directed mutagenesis of the hIPR processing site has demonstrated that purified ss-furin exhibits the same specificity as the physiological hIPR-processing enzyme.
Cloning Human fur cDNA-Oligonucleotide primers corresponding to the 5' end of exon S and the 3' end of exon T of the human fur gene (8) were used to amplify a 300-bp fragment from a human KB cell cDNA library in the plasmid vector pCD (27) (supplied by Takashi Yokota, DNAX, Palo Alto, CA) using the polymerase chain reaction. This fragment was used as hybridization probe to identify plasmid pCD-fur, from the same library, a -2,300-kilobase pair fur cDNA that lacked the 5'-untranslated region and 250 bp of coding sequence and ended with a poly(A) stretch beginning 145 bp 3' to the translational termination site. This apparently represents utilization of an alternative poly(A) addition site in that previously published fur cDNAs contain -1.5 kilobase pairs 3' beyond the stop codon (28). Publication of a full-length fur cDNA (22) permitted synthesis of additional primers that were used to obtain the remainder of the fur coding sequences. The polymerase chain reaction was used to amplify a fragment from total human KB cell RNA from 32 bp 5' to the ATG codon to a KpnI site 370 bp 3' to the ATG codon. The 5' primer incorporated a n EcoRI site, and the resulting EcoRI-KpnI fur cDNA fragment was inserted into pCD-fur to produce a 2,786-bp fur cDNA that was subsequently inserted into vector pUN70 (29) linearized by digestion with EcoRI plus SalI to generate PUN-fur.
Expression of ss-Furin-For expression of ss-furin, a SalI adaptor containing tandem termination codons in the three sense reading frames (30) was inserted at a StuI restriction site immediately after codon 595 of the fur cDNA in PUN-fur creating PUN-fur-595. The 1,816-bp EcoRI-Sal1 fur cDNAfragment was excised from PUN-fur-595

Proreceptor by Human
Furin 25831 and inserted into the Autographa californica nuclear polyhedrosis virus (baculovirus) transfer vector pVL using BclI adaptors (311, creating plasmid pVL-fur-595. The insect cell line Sf9 (an ovarian cell line of the fall armyworm Spodoptera frugiperda) was cotransfected with pVL-fur-595 and linear wild type baculovirus genomic DNA (Invitrogen Corp.) for recombination in uiuo. Recombinant baculovirus (bac:fur-595) in the medium was isolated from wild type baculovirus by four rounds of plaque purification, and a high titer stock was obtained by infection of Sf9 cells. Hi5 cells (ovarian cell line from the cabbage looper moth, Tkzchoplusia ni, Invitrogen Corp.) were infected with either wild type (bac:wt) or bac:fur-595 viral stocks a t a multiplicity of infection of two, and media samples (300 pl) were analyzed for ss-furin expression.
Purification ofss-Furin-2 x lo7 Hi5 cells, grown to -90% confluence in T-225 flasks (Corning) in 40 ml of Grace's medium plus serum (marketed as TMN-FH by Life Technologies, Inc.), were used to seed three fresh flasks each containing 40 ml of serum-free medium EXCELL 401 (Jackson Laboratories). When cells reached 70% confluence (60 x lo6 celldflask), medium was removed, and monolayers were infected with bac:fur-595 stock virus at a multiplicity of infection of two. After 1 h, the inoculum was replaced with fresh EXCELL 401 medium. After an additional 90 h, medium was harvested by centrifugation at 1,000 x g for 15 min and dialyzed against four changes of 6 liters of buffer A (10 mM Tris-Cl,, pH 8.0, 1 mM CaCl,, 0.1 IILM o-phenanthroline) in Spectrflor 7 membrane (50-kDa cutoff, Spectrum Medical Industries, Inc.). Dialysis was necessary to remove a low molecular weight contaminant in the serum-free medium that interfered with binding of ss-furin to Q-Sepharose. The dialysate was clarified by centrifugation a t 100,000 x g for 30 min, and the supernatant fraction (S100) was applied to a 44-ml (2 x 14 cm) fast flow Q-Sepharose column (Pharmacia) equilibrated with buffer A. The column was washed with 300 ml of buffer A and eluted with 50 ml of buffer B (buffer A with 10 mM CaCl, and 10% glycerol). Fractions and pools were frozen in liquid N, and stored at -80 "C without loss of activity for more than 10 months.
Assays ofss-Furin Actiuity-In 50 pl the standard assay contained 20 mM NaMES, pH 7.0, 1 mM CaCl,, 0.01% (v/v) Triton X-100, 0.5% (v/v) dimethyl sulfoxide, and either 100 p~ Boc-RVRR-MCA or 150 pg of detergent-solubilized cell extracts. Reactions with Boc-RVRR-MCA were initiated by adding enzyme, incubated at 37 "C for various times, and terminated with 0.85 ml of ice-cold 0.125 M ZnSO,. Released AMC was determined fluorometrically (32). Aunit was defined as release of 1 pmol of AMC/min from Boc-RVRR-MCA under standard assay conditions. Furin was stable under assay conditions for >>1 h. Reactions with detergent-solubilized cell extracts were terminated by adding ice-cold trichloroacetic acid to 6% (w/v), and precipitates recovered by centrifugation a t 12,000 x g for 15 min were washed twice with ice-cold acetone, air dried, and solubilized in SDS-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer (10 mM Tris-C1, pH 6.8, 5 mM P-mercaptoethanol, 1% SDS, and 10% glycerol) at 100 "C for 5 min before immunoblotting. To determine steady-state rate constants (kcav K,) for cleavage of Boc-RVRR-MCA, the initial rates of AMC release were made by continuous measurements of AMC release from 0 to 500 s in a fluorometric cuvette a t 37 "C in 2.5 ml of 20 mM NaMES, pH 7.0, 1 mM CaCl,, 0.01% Triton X-100 with various concentrations of Boc-RVRR-MCA. Data were analyzed using Windows Enzfitter program (Microsoft, Inc.) as described (32). Rapid kinetic experiments were performed using a n EnzTeck, Inc. quenched flow apparatus as described (32).
Immunological /Immunochemical Methods-Affinity-purified rabbit anti-hIR P-subunit polyclonal antibodies (anti-IRP) were as described (33). To obtain anti-furin antiserum, a lacZ-furin fusion was constructed by inserting a 1.3-kilobase pair SacII-Sal1 fragment of the fur cDNA from PUN-fur-595 into vector pUR278 (34). The resulting plasmid encoded a fusion protein (P-galactosidase-N-furin) with 366 residues ofthe furin subtilisin-and P-domains (30) fused to the carboxyl terminus of P-galactosidase. The fusion protein was purified from Escherichia coli inclusion bodies as described and used to immunize New Zealand White rabbits (21). Anti-furin antibodies were affinity purified as described (35).
Samples to be analyzed by immunoblotting were subjected to SDS-PAGE (10% gel) a t 40 volts overnight. Gels were transferred for 3 h a t room temperature at 40 volts onto HAY nitrocellulose (Schleicher & Schuell) in 20 mM Tris-HCl, pH 7.6,360 mM glycine, and 30% methanol. Nonfat dry milk (5% w/v) in Tris-HC1 buffered saline (TBS = 50 mM Tris-HC1, pH 7.5, 150 mM NaCI) was used to block nonspecific sites on the nitrocellulose. Transfers were incubated overnight a t room temperature using a 1 5 0 dilution of primary antiserum in TBS plus 1% nonfat dry milk and 10 mM NaN,. Blots were washed with TBS, incubated for 30 min with goat anti-rabbit IgG coupled to horseradish peroxidase (Amersham; 1 mg/ml) diluted 1:3,000 in TBS plus 1% nonfat dry milk, and then washed extensively with TBS. Immunodetection was performed using the ECL chemiluminescence detection procedure ( h e rsham) and XAR-5 x-ray film (Kodak). Densitometry was performed using Image Quant software (Molecular Dynamics, Inc.).

RESULTS
High Level Production of ss-Furin Using Recombinant Baculovirus-Full-length cDNA encoding human preprofurin was obtained from a KB cell library by a two-step cloning procedure (see Fig. 1 and "Experimental Procedures"). To produce ssfurin, the cDNA was modified by introducing a translational stop codon after codon 595. The ss-furin coding sequence thus retained all sequences conserved in the Kex2 family of processing enzymes, the subtilisin domain and the contiguous P-domain (21, 22, 30). A recombinant baculovirus (hereafter, bac: fur-595), in which the ss-furin cDNA was placed under the control of the viral polyhedrin promoter, was isolated and plaque purified. Stocks of wild type (bac:wt) and bac:fur-595 virus were used to infect Hi5 cells, an ovarian cell line derived  from the cabbage looper moth T ni. Cells were grown in serumfree medium, and samples of media collected a t times after infection were analyzed by SDS-PAGE (Fig. 2). A 57-kDa Coomassie Blue-stained polypeptide that accumulated only in the medium of cells infected with bac:fur-595 virus (Fig. 2 A ) reacted specifically with affinity-purified polyclonal anti-furin antibodies (Fig. 2B), confirming the identity of this species as ss-furin. Appearance of this band coincided with the appearance of a calcium-dependent proteolytic activity that cleaved the fluorogenic peptide substrate Boc-RVRR-MCA (Fig. 2 0 . Activity was first detected at 40 h after infection and increased linearly until a t least 89 h after infection. Purification of ss-Furin from Insect Cells-Due to both the high level of expression of ss-furin driven by the bac:fur-595 virus and the ability to passage Hi5 cells in serum-free medium, the specific activity of ss-furin in the medium of cells infected with the recombinant virus (Table I) was 20-30-fold higher than obtained by expression in cultured mammalian

Purification of ss-furin from conditioned medium of bacfur-595-infected Hi5 cells
from the Q-Sepharose column were measured after preincubation of enzyme for 5 min at 37 "C in reaction buffer. This led to a 3-fold increase in Activity of the dialysate was measured in the supernatant fraction after centrifugation a t 100,000 x g. Activities in the fractions and the pool activity. cells? (17,18). As a result, purification of ss-furin from the medium of Hi5 cells infected with bac:fur-595 virus required only a &fold enrichment. Dialysis of the medium was required to remove a low molecular weight contaminant that prevented binding of the enzyme to Q-Sepharose. Selective elution of ssfurin from Q-Sepharose depended on increasing the calcium concentration from 1 to 10 mM (Fig. 3A). By SDS-PAGE, the predominant polypeptide in the peak of activity migrated a t 57 kDa (Fig. 3B) and reacted specifically with anti-furin antibodies (Fig. 3C). N-Glycanase digestion of the 57-kDa polypeptide resulted in a loss of 2 kDa from the polypeptide, indicating that one of the three potential N-linked oligosaccharide signals is utilized (data not shown). The enzyme in fraction 4 was judged to be 292% pure by Coomassie Blue staining. Amino acid analysis was consistent with the expected amino acid composition of ss-furin.
Active-site Titration ofPurified ss-Furin-Although most serine proteases do not exhibit rate-limiting cleavage of the acylenzyme with amide substrates, Kex2 protease was found to do so with a peptidyl methylcoumarin amide, permitting active-site titration by measuring the pre-steady-state burst of AMC release (32). ss-Furin also exhibited a pre-steady-state burst of AMC release in rapidly quenched reactions containing a saturating concentration of Boc-RVRR-MCA (400 PM) (Fig.  4A ). With 10.4 pmol of ss-furin, as determined by quantitative amino acid analysis (data not shown), extrapolation of the linear phase of hydrolysis (500 ms-60 s) revealed a 6.6-pmol burst, indicating that the enzyme in fraction 4 was -60% active.
Like preparations of furin derived from expression in mammalian cell culture (17, 181, cleavage of Boc-RVRR-MCA by ss-furin was Ca2+-dependent. ss-Furin exhibited a K,, of 18 PM and a Kc,, of 0.6 s-l for Boc-RVRR-MCA yielding a K,,JK,,, of 3.3 x lo4 M" s-l (Fig. 4B ),33-fold lower than observed with purified Kex2 protease on the same substrate and 330-fold lower than Kex2 on its best peptidyl-MCA substrate, acetyl-Pro-Met-Tyr-Lys-Arg-MCA (32). No hydrolysis was observed upon prolonged incubation (-1 h) with ss-furin of acetyl-Pro-Met-Tyr-Lys-Arg-MCA at substrate concentrations up to 360 PM (data not shown). Conservatively, release of 5 pmol of AMC could have been observed, indicating an upper limit of 30 M" for KJK, for ss-furin on this substrate.
Cleavage of the Human Insulin Proreceptor by ss-Furin-Initial attempts to assess the ability of ss-furin to cleave hIPR were made using medium from cells infected with the bac:fur-595 virus as a source of enzyme. CHO-IR cells, Chinese hamster ovary cells that stably express the hIR+11 isoform a t a level 100-fold higher than the endogenous receptor, were the source of hIPR. Clarified extracts were prepared from CHO-IR  cells by detergent solubilization followed by centrifugation at 12,000 x g. Affinity-purified polyclonal antibodies (anti-IRP antibodies) that recognize both hIPR and the mature p-subunit detected only two polypeptides in this extract, the 190-kDa insulin proreceptor and the 95-kDa mature P-subunit (Fig. 5, lane 1 ). This pattern was essentially unaltered by incubation at 37 "C for 30 min with buffer alone or with medium from bac: wt-infected cells (Fig. 5, lanes 2 and 3). Incubation of the CHO-IR extracts with medium from bac:fur-595-infected cells ( 85-kDa species recognized by the anti-IRP antibodies (Fig. 5, The difference between the molecular mass of the endogenous P-subunit and the product of cleavage by purified ss-furin could be due to differences in the oligosaccharyl modifications of the two species. To examine this possibility, the N-linked oligosaccharides of the proreceptor, endogenous P-subunit, and ss-furin cleavage product were probed by digestion with Nglycanase (Fig. 5, lanes 6 and 7) and Endo H (Fig. 5, lanes 9 and  10). The proreceptor band was completely sensitive to Endo H, demonstrating that proreceptors were present only in the endoplasmic reticulum or cis-Golgi. The ss-furin cleavage product was also completely sensitive to Endo H, consistent with a precursor-product relationship between this 85-kDa species and the proreceptor. As expected, N-glycanase digestion increased the mobility of both the proreceptor and ss-furin cleavage product by about the same amount as Endo H treatment. In contrast, the endogenous P-subunit band exhibited only slight sensitivity to Endo H digestion (36), indicating extensive modification of N-linked oligosaccharides in the medial Golgi or later compartments. This interpretation was confirmed by observing much more extensive digestion of the endogenous P-subunit band with N-glycanase (37), an enzyme that, unlike Endo H, cleaves oligosaccharides of the complex type.
Effects source of hIPR, were treated as described below and then separated on 8% SDS-PAGE and immunoblotted using anti-IRP antibodies. Lane 1, no incubation. Incubation for 60 min at 37 "C with either buffer alone (100 mM Tris-Cl, pH 7.5,l mM CaCI,, 0.01% Triton X-100; lanes 2,5 and 8), medium from bac:wt- infected cells (lanes 3, 6, and 9), or medium from bac:fur-595-infected cells (18 unitsheaction; lanes 4, 7, and 1 0 ) in a total volume of 100 pl. Reactions were terminated by trichloroacetic acid precipitation and the precipitates acetone-washed before SDS-PAGE. Samples in lanes 6 and 7 were digested with N-glycanase and in lanes 9 and 10 with Endo H (9). products, N-glycanase digestion of the endogenous p-subunit appeared to be incomplete, although the most rapidly migrating species approached the mobility of the ss-furin cleavage product. To assess more directly the specificity of the ss-furin cleavage of the solubilized proreceptor, the effects of mutations in the proreceptor processing site were determined.
Substitutions of alanine codons for each arginine codon in the proreceptor cleavage site (sequences and designations in the one-letter amino acid code were: R W , , , P,Ala; R&,,R, & l a ; &,,,KRR, P4Ala) were made in sequences encoding the hIR+11 isoform. Reconstructed cDNAs were subcloned under the control of the constitutive hybrid SV40 early promoter in the mammalian expression vector SRa (38). Mutant and wild type pSRaIR+11 plasmids were expressed by transient transfection of COS cells, and detergent-solubilized cell extracts were prepared. The COS cell extracts were incubated at 37 "C for 30 min with purified ss-furin (62 units of pooled fractions 3-5) in standard reaction conditions containing either 1 mM CaCl, or 10 mM EDTA. As seen previously in the case of human proreceptors in CHO-IR extracts, wild type ( Fig. 6 A , lane 3 ) and Pfia (Fig. 6 A , lane 7) proreceptors in the COS cell extracts were cleaved by ss-furin to generate an 85-kDa species recognized by the anti-IRP antibodies. Cleavage was blocked by Ca2+ chelation (compare Fig. 6A lanes 3 and 4 with lanes 7 and   8), consistent with the known Ca2+ dependence of Kex2 protease (32) and of furin purified from mammalian cells (17,18). Under identical assay conditions, ss-furin failed to cleave the P,Ala and P4Ma proreceptors (Fig. 6 A , lanes 5 and 9). This result is consistent with the absolute requirement a t furin cleavage sites for Arg a t both P, and P, and demonstrates that purified ss-furin cleavage of the hIR+11 occurs at the correct processing site. The 85-kDa band is, therefore, authentic hIR p-subunit containing immature N-linked oligosaccharide.
ss-Furin Exhibits the Same Specificity as the Endogenous hZPR-processing Enzyme-Previously, a naturally occurring substitution of Ser for Arg at PI in the hIPR processing site was found to block proteolytic maturation, resulting in severe insulin-resistant diabetes in two sibling human patients (11,12). Subsequent mutational analysis of the processing site demonstrated a requirement for Arg a t both PI and P4 for processing in CHO cells (16). In these studies, substitution of Ala for Lys at P3 or for Arg at P, had no effect on cleavage. In the current  6. Furin specificity in h P R cleavage. Panel A, purified ssfurin (50 nM) was incubated with detergent extracts (200 pg as protein) of COS cells transfected with either wild type or mutant hIPR constructs. Otherwise standard reactions (50 pl) contained either 1 mM CaCI, (+) or 10 mM EDTA(-) as indicated and were incubated for 30 min at 37 "C. Processing of proreceptors was assessed by SDS-PAGE followed by immunoblotting using anti-IRp antibodies. Panel B, COS cells were detergent solubilized after transfection with either wild type or mutant hIR expression plasmids, and cellular processing of proreceptors was assessed by immunoblotting using anti-IRP antibodies after SDS-PAGE. experiments, the endogenous COS cell-processing enzyme exhibited similar specificity ( Fig. 6B; also -Calcium lanes in Fig.  6A). Wild type hIR+11 and P&a mutant proreceptors were cleaved by the cellular enzyme to generate mature 95-kDa 0-subunit (Fig. 6B, lanes 1 and 3). However, no mature 0-subunit was observed with the P,Ala and P,Ala mutants (Fig. 6B,  lanes 2 and 4 ) . A faint 95-kDa band observed in anti-IRPstained immunoblots of both P,Ala-and P,Ala-transfected cells was also seen in the case of mock-transfected cells (data not shown) and therefore corresponds to endogenous COS cell insulin receptor 0-subunit. Thus, the IPR-processing enzyme endogenous to both CHO and COS cells exhibits the same specificity as purified ss-furin does in maturation of hIPR.
It is interesting to note that the level of mature 0-subunit seen in cells transfected with the P&la mutant was variable and that in the case of all three mutants but not the wild type, a n additional proreceptor band was observed at -200 kDa, just above the major proreceptor band. This species probably corresponds to proreceptor molecules that have undergone extensive modification ofN-linked oligosaccharides, suggesting that processing of the P&a proreceptor is somewhat delayed in vivo compared with wild type hIPR. Comparative Kinetics of Cleavage of Wild l)pe and Mutant Proreceptors-To estimate the degree of discrimination by ssfurin between the wild type and mutant proreceptors, the kinetics of cleavage were examined. Fig. 7A demonstrates that with the amount of ss-furin used in the experiment in Fig. 6A (58 units of pooled fractions 3-5), cleavage of wild type hIR-11 proreceptor was actually complete by 5 min of incubation. Incubation with 6-fold less enzyme (9.6 units) resulted in 11% cleavage in 5 min. Therefore, the time course of cleavage for wild type and mutant proreceptors was examined with the lower amount of ss-furin activity. Rates of cleavage were assessed for both the hIR-11 and hIR+11 isoforms of wild type proreceptor. Fig. 7B shows the time course of cleavage of wild type hIR-11 proreceptor by ss-furin at the reduced concentration of enzyme (9.6 units). Densitometric analysis of these data (Fig. 7C) demonstrated a correlation between disappearance of hIR-11 proreceptor and appearance of 0-subunit in vitro. The rate of disappearance of the hIR-11 proreceptor appeared to be first order, suggesting that proreceptor concentration was significantly below K,, for furin cleavage (see "Discussion"). Initial rates of cleavage of the hIR-11 and hIR+11 forms of wild type proreceptors, as measured by the appearance of 0-subunit, were approximately the same (Fig. 7 0 ). The rate of cleavage of P+la proreceptor was also similar to that of the wild type forms. In contrast, cleavage of the P,Ala and P,Ala proreceptors was undetectable a t 60 min, whereas both the wild type hIR-11 and hIR+11 and P$la mutant proreceptors were cleaved to completion by that time (Fig. 70).
Approximately 20% of both the hIR-11 and hIR+11 wild type proreceptors was cleaved by the first time point (3.5 min). Because cleavage of 5% of the P,Ala and P,Ala proreceptors would have been easily detectable a t 60 min, the initial rate of cleavage of wild type proreceptors, proportional to k,,jK,,,, was at least 70 times that of the P,Ala and P,Ala mutants. Because cleavage of these two mutant proreceptors was undetectable under even more vigorous reaction conditions (i.e. with a &fold higher enzyme concentration as shown in Fig. 6A), the difference in kc,,jKm between the wild type and the P,Ala and P,Ala mutant proreceptors is likely to be considerably greater than 7O-fold.

DISCUSSION
The results presented here demonstrate that human furin cleaves the human insulin proreceptor at the physiological processing site with the same pattern of specificity as the cellular processing enzyme. First, incubation with ss-furin converts the Endo H-sensitive 190-kDa human insulin proreceptor in cell extracts to an Endo H-sensitive 85-kDa polypeptide that cross-reacts with anti-IRB antisera in immunoblots (Fig. 6). The difference in mobility between the mature, Endo H-resistant 95-kDa 0-subunit of hIR and the 85-kDa species produced by ss-furin digestion in vitro is accounted for by differences in N-linked oligosaccharide modifications. Second, complete inhibition of cleavage by mutation of either the PI or P, Arg in the proreceptor processing site demonstrates directly that ss-furin cleaves the proreceptor at the correct site. Moreover, the inhibitory effects of these mutations and the lack of a substantial effect of mutation of the P, Arg to Ala are entirely consistent with the reported specificity both of the cellular insulin proreceptor-processing enzyme and of furin.
This direct biochemical evidence for furin as the physiological insulin proreceptor-processing enzyme is complemented by the analysis of two tissue culture lines found to be defective in maturation of proproteins a t furin-like cleavage sites. Moehring and co-workers (39) found that a CHO cell line (RPE.40 cells) with a pleiotropic defect for maturation of precursors of bacterial exotoxins and viral envelope glycoproteins also exhibited a defect for insulin receptor maturation which was corrected by expression of mouse furin cDNA. The Arg-Xaa-Xaa-Arg cleavage motif is conserved in proreceptors structurally related to the insulin proreceptor including the insulin-like growth factor-I receptor, the insulin receptor-related receptor, and the hepatocyte growth factor receptor (6)(7)(8). A human colon carcinoma LoVo cell line incapable of processing the hepatocyte growth factor proreceptor was shown to have a point mutation in the fur gene (40). This defect was complemented by transfection of LoVo cells with mouse fur cDNA (41).
The Endo H sensitivity of the 190-kDa proreceptor in cell extracts indicates that this species is either an endoplasmic reticulum or early Golgi form (Fig. 5, lanes 7 and 101, support-ing the conclusion that proreceptor processing ordinarily occurs after modification of N-linked oligosaccharide in the medial and trans-Golgi. This is consistent with localization of furin, a t B. A. least when overexpressed, to the trans-Golgi network (24). Cleavage of the endoplasmic reticulum form of insulin proreceptor by purified ss-furin demonstrates that oligosaccharyl modification does not regulate recognition of the cleavage site. Indeed, previous studies showed that an 85-kDa species is produced from the 190-kDa insulin proreceptor when cultured adipocytes are treated with low concentrations of the cation ionophore monensin that block late glycosyl modifications without blocking proreceptor cleavage (42). ss-Furin cleaved both the hIR-11 and hIR+11 insulin proreceptor isoforms with indistinguishable kinetics (Fig. 7D). Thus, the presence or absence of the 12 residues encoded by exon 11, which lie just upstream from the processing site in hIR+11 proreceptor, does not affect recognition by the protease. In Rat-1 cells transfected with expression constructs of both insulin receptor isoforms, hIR-11 was shown to bind insulin with a 2-fold greater affinity than hIR+11 (43). Thus, the 12 residues encoded by exon 11 may influence the affinity of the a-subunit for insulin, but not the specificity of proreceptor processing.
Engineering of a secreted, soluble form of furin for expression in insect cells was based on a similar approach employed in amplification and purification of ss-Kex2 protease from yeast (321, which has also been used by others to secrete furin from mammalian cells (17,18). A comparison of the properties of purified ss-Kex2 and ss-furin is instructive, because despite substantial conservation of primary structure (44% identity and 63% similarity in the subtilisin domain), the two enzymes exhibit not only significant similarities but also differences in their catalytic activity and specificity. Both enzymes exhibit selectivity for Arg a t P,. However, whereas residues other than Lys or Arg at P, substantially reduce the kc,,lK, for Kex2 protease (the effect is almost exclusively on&,), substitution ofAla for Arg in the P, position of the insulin proreceptor processing site had little effect on processing by ss-furin. Instead, furin appears to exhibit a high degree of specificity for Arg at P4. Kex2 protease does not exhibit specificity for P4 substrate residues, although aliphatic residues or methionine tend to be found in this position in natural substrates for the enzyme.
ss-Furin exhibits a substantially lower kc,, (-0.6 s-l) for Boc-RVRR-MCA than does Kex2 (25 s-l). One possibility is that the tetrapeptide is a poor substrate for furin, in which case furin may require more extensive interactions for an optimal kc,,. A second possibility is that the catalytic cycle of furin is intrinsically slower than that of Kex2. Because the transit time of substrates through the mammalian secretory pathway is nearly 10-fold slower than in yeast, the encounter time for furin with its substrates in vivo is likely to be considerably longer than between Kex2 and its substrates. Thus furin may be as fast as it needs to be.
It is striking that both Kex2 and furin exhibit burst kinetics in cleaving MCA substrates, which most likely means that cleavage of the acylenzyme is rate-limiting. An interesting hypothesis is that the extensive primary structure interactions between Kex2 or furin with substrate peptides somehow favor formation of the acylenzyme over its hydrolysis.
Our direct biochemical studies, in combination with coexpression studies, provide strong evidence that furin is the cellular enzyme responsible for activation of the insulin proreceptor and most likely the precursors of other members of the IR family of receptors as well. The glycoprotein precursors of numerous lipid-enveloped viruses, including HIV-I 160-kDa glycoprotein and fowl plague virus hemagglutinin, require cleavage by a cellular protease at an Arg-Xaa-Xaa-Arg sites for activation of their fusogenic potential, which is necessary for entry into host cells. Inhibitors of furin activity have been designed (44) and shown to prevent syncitium formation and secondary infection of HIV virions released from cell cultures (26). Efforts that target furin for antiviral drug development must take into account the important cellular roles of the enzyme in the host. For example, although inhibition of insulin proreceptor processing briefly during acute viral infection may be possible, toxicity due to long term inhibition may be unacceptable.