THE JOURNAL 0 1991 by The American Society for Biochemistry and Molecular Biology,

OF BIOLOGICAL CHEMISTRY Inc.

Vol . 266, No. 25, Issue of September 5, pp. 16954-16959,1991 Printed in U. S. A.

Functional Expression of Furin Demonstrating Its Intracellular Localization and Endoprotease Activity for Processing of Proalbumin and Complement Pro-C3*

(Received for publication, May 10, 1991)

Yoshio Misumi, Kimimitsu Oda, Toshiyuki Fujiwara, Noboru Takami, Kosuke TashiroS, and Yukio Ikeharas From the Department of Biochemistry and the Radioisotope Laboratory, Fukuoka university School of Medicine, Jonan-ku, Fukuoka 814-01 and the #Department of Zoology, Faculty of Science, University of Tokyo, Bunkyo-ku, Tokyo 113, Japan

We have cloned a rat cDNA encoding furin which is structurally related to yeast Kex2 protease. Products of 88 and 94 kDa were obtained by in vitro transcrip- tion/translation of the cDNA in the absence and pres- ence of microsomes. When the cDNA was transfected into COS-1 cells, furin was expressed as a major gly- cosylated form of 94 kDa, accompanied by a minor proteolytic form of 86 kDa, and found to be localized in the Golgi complex.

Proalbumin and complement p r o 4 3 a r e intracellu- larly processed into their mature forms by cleavage at the dibasic residues Arg-Arg, a common cleavage sig- nal found in many pro-type precursors. In cells trans- fected with the cDNA of C3 or albumin alone, only about half of each proform expressed was processed by an endogenous activity of the cells. When furin was co- expressed, the proforms of both C3 and albumin were completely processed into their mature forms. In ad- dition, co-expression of rat al-protease inhibitor mu- tant (Met3“+Arg) resulted in inhibition of the endog- enous and exogenous processing activities, as observed for the naturally occurring mutant Pittsburgh which has been identified as a specific inhibitor for the proc- essing enzyme. Taken together, these results indicate that furin is an endoprotease localized to the Golgi complex and capable of processing proalbumin and p r o 4 3 into the mature forms.

Many peptide hormones and neuropeptides are initially synthesized as larger precursors which are then cleaved into biologically active mature forms prior to secretion (1-3). Similar proteolytic events also take place in other systems, as observed for proalbumin (4, 5) and complement pro-C3l (6) in hepatocytes. The amino acid sequences most frequently used as endoproteolytic cleavage signals are the paired basic residues Lys-Arg and Arg-Arg. Many efforts have been made

* This work was supported by grants-in-aid for Scientific Research from the Ministry of Education, Science and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

5 To whom correspondence should be addressed: Dept. of Biochem- istry, Fukuoka University School of Medicine, Jonan-ku, Fukuoka 814-01, Japan. Tel.: 092-801-1011 (ext. 2878); Fax: 092-865-6032.

The abbreviations used are: C3, the third component of comple- ment; Endo H, endo-P-N-acetylglucosaminidase H; aI-PI, a,-protease inhibitor; al-PIM/R, a,-PI mutant (MeP-Arg); PAGE, polyacryl- amide gel electrophoresis; SDS, sodium dodecyl sulfate; WGA, wheat germ agglutinin.

to isolate and characterize endoproteases involved in the processing of these precursors (3, 7, 8). However, there are few instances where it has been established that the specificity of cleavage and subcellular localization of the enzyme are consistent with a physiological role.

Yeast Kex2, a Ca2+-dependent serine protease, is known to be involved in proteolytic processing of precursors of a-factor and killer toxin (9, 10). The enzyme also properly processes proinsulin ( l l ) , pro-opiomelancortin (12), and proalbumin (13), indicating its ability to cleave at both Lys-Arg and Arg- Arg sequences. The primary structure of Kex2 predicted by its DNA sequence reveals that it contains an active-site domain homologous to the bacterial subtilisins (14). Recently, a similar subtilisin-like domain has been demonstrated in two novel proteins predicted by cloning and sequencing of their cDNAs. One is furin (15-17) encoded by the fur gene (for the upstream region of the c-feslfps proto-oncogene) (18). The other is PC2 from a human insulinoma (19). A characteristic difference between furin and PC2 is that furin has a highly hydrophobic segment, corresponding to a membrane-span- ning domain, near the COOH terminus, while PC2 has no such domain.

In the present study we have examined the biosynthesis, intracellular localization, and proprotein-processing activity of furin expressed in COS-1 cells transfected with its cDNA. Furin expression was found to be localized to the Golgi complex and to process proforms of albumin and complement C3 co-expressed in the cells.

EXPERIMENTAL PROCEDURES

Material~-[~’SS]Methionine (X300 Ci/mmol) was purchased from DuPont-New England Nuclear. mCAP@ mRNA capping kit was obtained from Stratagene; endo-P-N-acetylglucosaminidase H (Endo H) from Seikagaku Kogyo (Tokyo, Japan); Dulbecco’s modified Earle’s medium, Eagle’s minimum essential medium, and fetal calf serum from Nissui Seiyaku (Tokyo); maleimide-activated keyhole limpet hemocyanin from Pierce Chemical Co. Goat anti-rat C3 anti- serum was obtained from Cappel Laboratories; peroxidase-conjugated sheep Fab of anti-rabbit IgG from Biosys Corp. (Compiegne, France); peroxidase-conjugated wheat germ agglutinin (WGA) from E-Y Labs. Anti-rat albumin IgG was raised in rabbits as described (20).

Preparation of Anti-furin Peptide ZgG-The synthetic peptide Thr- Gln-Met-Asn-Asp-Asn-Arg-His-Gly-Thr-Arg-Cys was used for rais- ing antibodies in rabbits. The sequence corresponds to the active-site His-containing domain in the predicted furin sequences (positions 187-198) (16). The synthetic peptide (3 mg) was conjugated with maleimide-activated keyhole limpet hemocyanin (6 mg) (21). The conjugates were injected into three rabbits every 3 weeks; 300 pg of the antigen with Freund’s complete adjuvant at the first injection and then 150 pg each with incomplete adjuvant. Antisera obtained from the rabbits after five injections were subjected to affinity chro- matography through a Sepharose column coupled to the furin peptide.

16954

Proprotein-processing Endoprotease Furin 16955

Construction n f Expression P/a.smid.s-Full-length cDNAs encoding rat furin (If), e,-protease inhihitor (nl-PI) (22). e , - P I mutant (Met'"'+Arg) (cr,-PIM/R) ( 2 3 ) , and complement C3 (24) were pre- pared as descrihed. The rat alhumin cDNA was provided hy Dr. Y. Kaneda (Osaka University). These cDNAs were digested with I.:roRI, and inserted into the I ~ J R I site of the pSG5 expression vector (25). The insert orientations of these plasmids were confirmed hv restric- tion endonuclease mapping. Each construct was prepared hv the alkaline-SDS lysis method (26) and purified hy equilihrium centrif- ugation in CsCl/ethidium hromide.

Transfection and Ano/ysis of Expressed Proteins-Each plasmid (20 pg) was transfected into 5 X 10'' COS-1 cells using an electropor- ation apparatus (Gene Pulser. Rio-Rad) as descrihed previously ( 2 3 , 2 7 ) . The transfected cells were cultured in Dulhecco's modified Earle's medium containing 10% fetal calf serum in 10-cm dishes for 2 days. The cells, unless otherwise indicated, were labeled for 4 h a t 37 "C with [."S]methionine (50 pCi/dish) in 5 ml of methionine-free Eagle's minimum essential medium. Cell lysates and media were prepared, and aliquots (20 pl each) were suhjected to SDS-PAGE (7.5% gels for C3 and furin, and 10% gels for alhumin), followed hv fluorography. When indicated. the cell lysates and media were suhjected to immu- noprecipitation with lgG monospecific for each antigen. The immu- noprecipitates were also analyzed hy SDS-PAGE or polyacrylamide gel electrophoresis (pH 5-8) (28) .

The biosynthesis of furin was also examined hy pulse-chase exper- iments. The transfected cells were pulse-laheled with [ '"Sjmethionine (100 pCi/dish) for 20 min and then chased in fresh Eagle's medium containing unlaheled methionine. A t the indicated times of chase, cells were harvested and the lysates suhjected to immunoprecipitation with anti-furin peptide IgG. The immunoprecipitates hefore and after treatment with Endo H (0.2 unit/ml at pH 5.5 and 37 "C for 16 h) were analyzed hy SDS-PAGE/fluorography as ahove.

In Vitro Transcription/Trans/ation-'I'he insert of the furin cDNA clone was cleaved with I h R I and inserted into the EroRI site of the vector pGEM3. In vitro transcription of the furin mRNA with T7 RNA polymerase was carried out using the m C A P mRNA capping kit in accordance with the manufacturer's protocol. In vitro transln- tion (0.5 pg of mRNA in a reaction mixture of 20 pl) was performed in a reticulocyte lysate system with [:"'S]methionine (20 pCi) in the presence or absence of dog pancreas microsomes (0.1 A?w unit) as descrihed previously (29). The products, when indicated, were di- gested with trypsin and chymotrypsin (50 pg each/ml) at 0 "C for 1 h in the presence or ahsence of 0.5% Triton X-100 (30).

Immunoc~tochemi.s~~-Cells grown on coverslips were rinsed in phosphate-buffered saline and fixed with 4% paraformaldehyde in 0.1 M phosphate huffer (pH 7.4) for 1 h a t room temperature. After fixation, cells were permeahilized hy treatment with 0.05% saponin in phosphate-buffered saline for 15 min (31). The cells were incuhated with anti-furin peptide IgG (35 pg/ml) or anti-rat alhumin IgG (25 pg/ml) in permeahilizing huffer for 1 h a t room temperature. Cover- slips with cells were hriefly washed with the same solution, and incuhated with peroxidase-conjugated sheep Fah of anti-rahhit IgG (10pglml) for 1 h. After heing washed with phosphate-hufferedsaline, the cells were incuhated for 10 min at room temperature with a substrate medium containing diaminohenzidine (0.5 mg/ml) and 0.01% H,O, in 0.05 M Tris-HCI huffer (pH 7.0) ( 3 2 ) . In separate experiments cells which had heen fixed and permeahilized as ahove were also incuhated with peroxidase-conjngated WGA (50 pg/ml) in permeahilizing buffer for 2 h a t room temperature, followecl hy the enzyme reaction as ahove. All coverslips with attached cells were then mounted on glass slides with glycerin. Specimens were examined with a Nikon light microscope.

RESULTS

Structural Characteristics of Furin-The cDNA clone pcHF3 with a 4.7-kilobase insert was initially isolated from a human liver cDNA library and found to contain the same open reading frame as that recently reported for human furin (15). We then isolated a corresponding cDNA clone (pcRF104, 4.3 kilobases) from a rat liver cDNA library using as a probe a BgIII-RarnHI fragment (1.5 kilobases) of pcHF3. The cDNA insert contained an open reading frame that encodes a 793- residue polypeptide with a calculated size of 86,662 (16). T h e entire sequence of rat furin shows 93.7 and 96.7% identity with those of the human (15) and mouse (17) proteins, re-

spectively. The predicted sequence contains a domain homol- ogous to the subtilisin family, a cysteine-rich domain, and three potential N-linked glycosylation sites (Fig. 1). In the subtilisin-like domain, residues Asp"", His"", and Ser"" of furin correspond to the respective active-site residues of suh- t i l isin RPN' (33), and the sequences surrounding these resi- dues are also highly conserved in yeast Kex2 (14) and human PC2 (19) and PC3 (34). A hydropathy plot demonstrated the presence of two major hydrophohic domains; the NH:,-termi- nal domain may represent a signal peptide, while the COOH- terminal domain corresponds to a transmemhrane domain (Fig. 1). Taken together, these structural features suggest that furin is a suhtilisin-like serine protease associated with the membrane in the COOH-terminal region.

In Vitro Transcription and Translation-The rat furin cDNA was suhjected to in vitro transcription and translation (Fig. 2). A control experiment, in which a pGE":l plasmid devoid of the furin cDNA insert was used for transcription, did not show production of any detectahle protein (Ianr 1 ) . The plasmid containing the cDNA insert directed the svnthe- sis of mRNA that was translated into a protein of 88 kI)a (lane 2), which is consistent with the predicted mass of furin (87 kDa). Translation of the mRNA in the presence of micro- somes produced another form of 94 kDa, in addition to the 88-kDa precursor (Ianr 3 ) . Digestion of the two products with mixed proteases (trypsin and chvmotrypsin) yielded a single protein of 86 kDa (Ianr 4 ) , which was aholished when the sample was digested in the presence of detergent (Ianr 5 ) . All three forms were immunoprecipitated with anti-(furin pep- tide) IgC (data not shown). Rased on the predicted structure of furin (Fig. l ) , these results could he interpreted as follows.

Subtlllron lakc dornam

71: 736

FIG. 1. Schematic representat ion of rat fu r in s t ruc tu re p re - dicted by the cDNA sequence. , < / I , putative signal peptide; I). / I . and S , putative active-site Asp. His, and Ser. respectively, which correspond to the ratalvtir triad of serine proteases; ( ' - r i ch , ('ys-rirh domain; 7". transmernhrane domain. 12il\efi "/rd/ipops" reprrsmt potential N-linked glycosylntion sites. rVtrrnhrrs indirnte pc~siticlns from the NH:! terminus. 'The undvrlincd sequenre inrludinc the active-site His (*) indicates a peptide u s ~ d for raising nntihodies.

M I 2 3 4 5 LO.

115- - 77.

66- - 56. - " -

[fn,,r,*, - - + + +I Tritonl-lW) - - - - +I mRNA - + + + +

(prolea5es - - - + +

FIG. 2. Analysis of i n vitro transcription/translntion prod- ucts. The p(;F:M:1 plasmid hearing the furin rl)S'A WIS trnnsrrihwl. The transcripts were then suhjerted to in vitro t r:lnsl;lt ion n s d f s .

scrihed under "Experimental I'roretlrlres." The Inheled prodrlrts, I)+ fore and after proteolytic digestion, were nnalyzetl hy Sl)S-llA(;F: (7.5% gels), followed hv fluorography. Innr. I . rontrol w i t horlt the transcripts; lanes 2-.5, translated with the transcripts in thr nhsenre (lane 2) or presence of clog pancreas mirrosomes (Innr3.q %.5 ). fr~llc~wctl hy digestion with trypsin/rhymotr>Tsin in the ahsenre ( / o w . I ) or presence of 0.5';' Triton S-IO0 (lonr- 5 ) . /,ant, .\I. moleruhr mnss markers containing the (115 k l h ) and d ((Ti k I h ) rh;lins o f complement C 3 , transferrin (7'7 k I h ) . nntl rr1-protease inhihitnr 1%; k I h ) which had hem metaholirally Inheled with [ Slmrthir~ninr.

16956 Proprotein-processing Endoprotease Furin

1) The 94-kDa protein is a form that is co-translationally transported into the microsomes, processed by signal pepti- dase, and further modified to have a higher molecular mass by the addition of oligosaccharide chains. 2) Upon treatment with the proteases, the 88-kDa form that had not heen seg- regated into the microsomes was completely digested, while the 94-kDa form was converted into the 86-kDa form by proteolytic removal of its COOH-terminal tail beyond the transmembrane domain. The 86-kDa form remained pro- tected by the membrane.

Transfection, Expression, and Localization of Furin-The plasmid pSG5 containing the furin cDNA insert was trans- fected into COS-1 cells. After 2 days of culture, the cells were pulse-labeled with [%]methionine and then chased, followed by immunoprecipitation with anti-(furin peptide) IgG. SDS- PAGE/fluorography demonstrated that under these condi- tions furin was synthesized as two forms, a major one of 94 kDa and a minor of 86 kDa (Fig. 3A) . When treated with Endo H, the 94-kDa form underwent reduction in its molec- ular mass to 88 kDa, while the 86-kDa form shifted to 81 kDa (Fig. 3R, lane I). The results demonstrate that the difference in molecular mass between the two forms is not due to differences in extent of glycosylation. Although both forms apparently did not alter their molecular weights during the chase times (Fig. 3 A ) , the Endo H treatment revealed that their oligosaccharides are finally processed into Endo H- resistant forms (complex type) a t a later time of chase (Fig. 3R, lane 4 ) . No protein immunoreactive with anti-(furin pep- tide) IgG was detected in the medium even after a prolonged time of chase (up to 12 h) (not shown), supporting the predic- tion that furin is a membrane bound protein.

We then examined the intracellular localization of furin thus expressed in the transfected cells. In control COS-1 cells without transfection, no significant reaction with peroxidase- conjugated anti-(furin peptide) IgG was observed (Fig. 4A). In contrast, cells transfected with the furin cDNA were in- tensely stained with immunoperoxidase, demonstrating that the reaction products are localized a t a juxtanuclear region (Fig. 4R). The staining profiles of these cells are quite similar t o those stained for albumin in cells transfected with the albumin cDNA (Fig. 4C) and also with those stained for WGA binding (Fig. 4 0 ) . Albumin is known to be concentrated at the Golgi complex prior to its secretion (27,31), and the lectin WGA specifically binds to oligosaccharides of proteins con- centrated in the Golgi complex (35, 36). Thus, the immuno- cytological observations suggest that expressed furin is local-

M 1 2 3 4 1 2 3 4 A) -Endo H B)+Endo H

kDa

115- - 77- - 66- - 56- - "" "

36-

FIG. 3. Biosynthesis of furin in transfected cells. COS-1 cells which had heen transfected with the furin cDNA were laheled with [""Slmethionine ( I 0 0 pCi/dish) for 20 min at :I7 "C and then chased. At the indicated times of chase, cells were separated from the medium, lysed, and suhjected to immunoprecipitation with anti-(furin peptide) IgG. The immunoprecipitates hefore ( A ) and after ( H ) treatment with Endo H were analyzed hv SDS-PAGE (7.5";; gels)/Iluorography. Ianm 1-4 in each panel represent samples ohtained at 0, 1, 3 , and 5 h, respectively. of the chnse time. Ianc M , molecular mass markers.

A B

D

8

w

FIG. 4. Intracellular localization of furin expressed in COS- 1 cells. COS-1 cells hefore (pnnrls A and I ) ) or nfter transfection with the furin cDNA ( p o n d R ) or the alhumin rDNA ( p o n d ( ' ) were cultured for 2 days. The cells grown on coverslips were fixed in 4p;

paraformaldehyde, permeahilized, and suhjected to immunochemiral (peroxidase-coupled lg(;) or cytochemical (peroxidase-iVGAI reac- tions as clescrihed under "Experimental I'rocedures." I'onrls A ant1 H , anti-(furin peptide) IgG;ponrl(', anti-alhumin 1gG;pnnrl I ) . iV(;A. The photograph of pond A was slightly overexposed compared with those for other panels. Hor, 10 pm.

A E Cell lysates Medium 1 2 3 4 1 2 3 4

"" 1 2 3 4

C M C M C M C M

.4-

""

FIG. S. Processingof pro-C3 by furin co-expressed in trans- fected cells. COS-I cells transfected with the indicated plasmids were incuhnted with [,""S]methionine at 37 "c' for 4 h, followed by preparation of cell lysates and media. Aliquots of each sample were directly analyzed by SDS-PA(;E (7..5f'L gels)/fluorography ( p o n d A I . The remaining samples were suhjerted to immnnoprecipitation with a n t i 4 3 IgG, and the immunoprecipitates were ~ I S I I analyzed as ahove ( p o n d R ) . The transfected plasmids were those fnr ( ' 3 and a , - l ' l (Ionc I ) ; C.? and nl-I'lM/H (lone 2 ) ; C3, t r , - P I . and furin ( Innr ; ] I ; and C.?. nl-PIM/R, and furin (lone 4 ) . C' and M represent the cell lysates and medium, respectivelv. Pro, n, and if denote the precursor and suhunits, respectively. of C3. An nrrotrhmd indicates the pogition of t r l - P I or ~ v , - P l M / R .

ized mainly in the Golgi complex. Furin Cleaves Proforms of C3 and Albumin into Thcir

Mature Forms"Pro-C3 has the structure NH,,-((f subunit)- Arg-Arg-Arg-Arg-(a subunit)-COOH and is processed into the suhunits with disulfide linkage prior to secretion (6). A possible function of furin in processing of pro-C3 was exam- ined by transfection experiments in COS-1 cells (Fig. 5 ) . The serine protease inhibitor al-PI and its mutant aI-PIM/R were also co-expressed, since the mutant nl-PIM/R (47) is consid- ered to be a specific inhihitor of the proalhumin-processing enzyme (13, 23, 38). The three secretory proteins C3, al-PI, and al-PIM/R were co-expressed well and secreted into the medium (Fig. 5A, right panel). Analysis of samples ohtained by immunoprecipitation more clearly demonstrated the proc- essing of pro-C3 (Fig. 5 R ) . Pro-C.7 expressed alone (not shown, see Ref. 23) or co-expressed with t r , -PI (Fig. 3 3 , I m p

I C ) was processed into the a and /3 subunits, hut a consider- able amount of pro-C3 was also secreted ( lam IM). Thus, it is likely that the COS-1 cells possess an endogenous proteo- lytic activity which is not sufficient for complete processing of pro-C3 expressed under these conditions. T h i s proteolytic

Proprotein-processing Endoprotease Furin 169.57

activity was completely inhibited by co-expressed nl-PIM/R (Fig. 5R, lane 2 M ) . In cells co-transfected with furin cDNA, pro-C3 was completely processed into two smaller molecules which correspond to the (Y and ,3 subunits (lane 3 M ) , suggest- ing that furin properly cleaved pro-C3 as did the endogenous protease. Co-expression of a,-PIM/R resulted in a significant but not complete inhibition of processing (lane 4 M ) .

The conversion of proalbumin into the mature form is accomplished by proteolytic removal of the NH2-terminal extension Arg-Gly-Val-Phe-Arg-Arg from the proform (4, 5). The two forms were indistinguishable on SDS-PAGE (Fig. 6A, right panel) but clearly separated by gel electrofocusing (Fig. 6R) . Essentially the same results were obtained for the processing of proalhumin in cells with or without co-transfec- tion of furin and/or nl-PIM/R cDNAs. Proalbumin was com- pletely converted into the mature form by co-expressed furin (Fig. 6R, lane 3 M ) , but its processing was not completely blocked by n,-PIM/R (Fig. 6H, lane 4 M ) .

A difference in the expression level of furin was also con- firmed. Furin expressed by co-transfection was detected a t a comparable level (Fig. 6C, lanes 3 and 4 ) to that in cells transfected with furin cDNA alone (Fig. 2). However, the corresponding protein(s) was not identifiable in the cells without transfection of furin cDNA (Fig. 6C, lanes 1 and 2) . The activity exerted by such a high level of furin could be sufficient for complete processing of the proforms (Figs. 5R and 6R, each lane 3 M ) but insufficient to be inhihited by the mutant inhibitor nl-PIM/R (Figs. 5R and 6R, each lane 4 M ) .

A 0 Cell lysates Medium 1 2 3 4 1 2 3 4 1 2 3 4

"" C M C M C M C M _ _ , -

"

FIG. 6. Processing of proalbumin by furin co-expressed in the transfected cells. COS-1 cells transfected with the indicated plasmids were incul)tlted with [.'"S]methionine a t 37 "C for 4 h, followed hy preparation of cell Ivsates and media. Aliquots of each sample were analvzed hv SDS-PACE (10";. gels)/fluorography (panrl A ). The remaining samples were suhjected to immunoprecipitation with anti-all)umin IgG (pnnrl H ) or anti-(furin peptide) Ig(; (panrl ('), and the immunoprecipitates were analvzed l ~ v gel electrofocusing (pH 5-8) (panrl H ) o r hv SDS-PAGE (pnnrl C ) , followed hv fluorog- raphy. The transfected plasmids were those for albumin and cr,-l 'I (lnnr I); alhumin and ~ r i - I ' I M / R (lnnr 2); albumin, tri-PI, and furin (lnnr 3 ) ; and alhumin, ni-I'IM/R, and furin (lnnr 4 ) . C and M represent the cell Ivsates and medium, respectivelv. An arrow and an arrorchwf indicate the positions of proahlmin (+albumin) and t r I -

1'1 (or t r l - I ' I M / l ~ ) , respectivelv. /'A and .SA represent proalbumin and alhumin, respectively. A long nrrorcl shows a pH gradient from pH 8 to 5.

DISCUSSION

Davidson et al. (39) recently characterized two distinct site- specific endoproteases for proinsulin cleavage in insulinoma secretory granules, demonstrating that these activities (types I and 11) are Cay+ dependent and cleave on the COOH side of Arg-Arg and Lys-Arg, respectively. T-ype I activity was found to process proalbumin into the mature form (40). A similar Ca"-dependent proteolytic activity has also heen detected in Golgi vesicles prepared from liver in which proalhumin is processed (38,41). Knowledge of these endoproteases includ- ing their molecular structures remains to he determined.

COS-1 cells derived from monkey kidney are considered to have only the constitutive pathway for secretion. The cells have often heen used for transfection and expression of var- ious secretory proteins. Although precursors of insulin (42). opiomelanocortin (43), and cholecystokinin (44) are not proc- essed a t all, those of somatostatin (45), von Willebrand factor (46), complement C3 (23), and alhumin (27) are significantly processed into mature forms in transfected cells. These oh- servations indicate that the selective processing of precursors in COS-1 cells is not simply explained hy differences in secretory pathways (regulated or constitutive) and of dihasic residues at the cleavage sites. Cleavahle dihasic pairs such as Arg-Lys (for somatostatin), Lys-Arg (von Willebrand factor), and Arg-Arg (albumin and C.7) are also found in precursors that are uncleavable in COS-1 cells. Despite this difficulty, COS-1 cells appear to have an endogenous activity for proc- essing these selected precursors, although at a level insuffi- cient for the complete processing of the overexpressed suh- strates. The incomplete processing found in C O S 1 cells was rather favorable in the present study for comparison of the endogenous activity with that exhibited hv furin exogenously introduced.

The in vitro transcription/translation experiments revealed that furin is synthesized as a primary precursor of 88 kDa, which is co-translationally processed into a 94-kDa form in the presence of microsomes. The same 94-kDa form was ohtained in the transfected cells and found to have oligosac- charides detected by the response to Endo H treatment. The presence of three N-linked glycosylation sites in the predicted structure also supports the increase in molecular mass on addition of oligosaccharides. An unexpected finding was that the transfected cells synthesized an additional minor form of 86 kDa which was precipitahle with anti-(furin peptide) IgC. Pulse-chase experiments with or without Endo H treatment excluded the possibility that the difference in molecular mass between the two forms is due to heterogeneity and/or different maturation of their oligosaccharide chains. Thus, it is likely that the 86-kDa form may he proteolytically derived from the major 94-kDa form, although the cleavage site remains to be determined. Since the smaller form was already detertahle in the pulse-label period, conversion should have occurred im- mediately after protein synthesis. I t is known that the proform of haptoglobin having a single Arg at the cleavage site is processed into (Y and [j subunits in the endoplasmic reticulum immediately after its synthesis (47, 48). I t is also of interest to note that the precursor of yeast Kex2 protease undergoes a rapid proteol-ytic removal of an NH,-terminal propeptide of about 9 kDa before reaching the Golgi (49). This event is considered to he related to the mechanism hy which the bacterial subtilisin precursor, structurally similar to both Kex2 and furin, is activated by autoproteolytic removal of the NH,-terminal extension (50, 51 ).

Furin expressed in COS-1 cells was found to he localized in the Golgi complex and ahle to cleave hoth pro-C3 and proal- humin which are believed to he processed in Golgi and related

16958 Proprotein-processing Endoprotease Furin

vesicles of hepatocytes (4-6, 28, 41). The products cleaved from pro-C3 and proalbumin migrated to positions corre- sponding to the respective mature forms on SDS-PAGE or gel electrofocusing. Although the precise cleavage sites remain to be determined, the results suggest that furin properly processes the proforms, because the endogenous protease works in COS-1 cells as well as in hepatocytes (6, 23, 27,41).

The use of protease inhibitors may prove valuable for characterizing the enzyme involved in the processing of pro- forms. The al-PI/Pittsburgh mutant was found in a patient with a bleeding disorder in which a considerable amount of proalbumin was detected in the plasma (37). The mutant inhibitor has the single substitution Met358+Arg at the reac- tive site and was confirmed to inhibit the conversion of proalbumin into albumin catalyzed by the Golgi enzyme (38) and by Kex2 (13) in vitro. The mutant aI-PIM/R used in the present study corresponds to human al-PI/Pittsburgh (23), and indeed completely inhibited the endogenous processing activity for proalbumin and pro-C3 in COS-1 cells. The inhib- itor, however, did not completely inhibit the proform-process- ing activity of the cells in which furin had been expressed. The incomplete inhibition observed could be explained by the fact that furin is overexpressed and accumulates in the Golgi complex to such an extent as to be easily detectable by immunoprecipitation and immunostaining, in contrast to that of the endogenous activity. The mutant inhibitor, although also being overexpressed, is a secretory protein and seems not to form a covalent complex with furin, as demonstrated for the inhibition of Kex2 with the inhibitor (52).

During preparation of this paper, the functional expression of human and mouse furins has been reported by other groups (17, 53-55). Precursors examined as a substrate of furin are those of von Willebrand factor (53,54), p-nerve growth factor (55), and renin (17), all of which have the same dibasic residues Lys-Arg at the cleavage site. The former two precur- sors were found to be processed into the mature forms by co- expressed furin (53-55), whereas prorenin was not cleaved at all (17). Taken together, these results indicate that furin processes precursors having Arg/Lys-Arg residues at the cleavage site, although its substrate preference is not simply restricted to dibasic residues alone, as observed for prorenin. The intracellular localization of furin also supports its func- tion in the Golgi complex where a variety of pro-type precur- sors are processed.

Acknowledgments-We wish to thank Dr. Y. Sakaki (Kyushu Uni- versity) for preparation of oligonucleotides, M. Sohda for excellent technical assistance, and Drs. K. Ohkubo and A. Fujita for help in some parts of the studies.

REFERENCES 1. Docherty, K., and Steiner, D. F. (1982) Annu. Reu. Physiol. 44,

2. Andrews, P. C., Brayton, K., and Dixon, J. E. (1987) Experientia

3. Fischer, J. M., and Scheller, R. H. (1988) J. Biol. Chem. 263,

4. Judah, J. D., and Quinn, P. S. (1976) Trends Biochem. Sci. 1,

5. Ikehara, Y., Oda, K., and Kato, K. (1976) Biochem. Biophys. Res.

6. Misumi, Y., Takami, N., and Ikehara, Y. (1984) FEBS Lett. 175 ,

7. Docherty, K., Carroll, R. J., and Steiner, D. F. (1983) Proc. Natl.

8. Parish, D. C., Tuteja, R., Alstein, M., Gainer, H., and Loh, Y. P.

9. Julius, D., Brake, A., Blair, L., Kunisawa, R., and Thorner, J.

625-638

(Basel) 43, 784-790

16515-16518

107-109

Commun. 72,319-326

63-67

Acad. Sci. U. S. A. 80, 3245-3249

(1987) J. Biol. Chem. 261 , 14392-14397

(1984) Cell 37, 1075-1089

10. Fuller, R. S., Brake, A., and Thorner, J. (1989) Proc. Natl. Acad. Sci. U. S. A. 8 6 , 1434-1438

11. Thim, L., Hansen, M. T., Norris, K., Hoegh, I., Boel, E., Forstrom, J., Ammer, G., and Fiil, N. P. (1986) Proc. Natl. Acad. Sci. U.

12. Thomas, G., Thorne, B. A., Thomas, L., Allen, R. G., Hruby, D. E., Fuller, R., and Thorner, J. (1988) Science 241 , 226-230

13. Bathurst, I. C., Brennan, S. O., Carrell, R. W . , Cousens, L. S., Brake, A. J., and Barr, P. J. (1987) Science 235,348-350

14. Mizuno, K., Nakamura, T., Ohshima, T., Tanaka, S., and Matsuo, H. (1988) Biochem. Biophys. Res. Commun. 156, 246-254

15. van den Ouweland, A. M. W . , van Duijnhoven, H. L. P., Keizer, G. D., Dorssers, L. C. J., and van de Ven, W . J. M. (1990) Nucleic Acids Res. 18 , 664

16. Misumi, Y., Sohda, M., and Ikehara, Y. (1990) Nucleic Acids Res. 18,6719

17. Hatsuzawa, K., Hosaka, M., Nakagawa, T., Nagase, M., Shoda, A., Murakami, K., and Nakayama, K. (1990) J. Biol. Chem.

18. Roebroek, A. J. M., Schalken, J. A., Leunissen, J. A. M., Onnek- ink, C., Bloemers, H. P. J., and van de Ven, W . J. M. (1986)

19. Smeekens, S. P., and Steiner, D. F. (1990) J. Biol. Chem. 265,

20. Ikehara, Y., Oda, K., and Kato, K. (1977) J. Biochem. (Tokyo) 81,1293-1297

21. Green, N., Alexander, J., Olson, A., Alexander, S. Shinnick, T. M., Sutcliffe, J. G., and Lerner, R. A. (1982) Cell 28. 477-487

22. Misumi, Y., Sohda, M., Ohkubo, K., Takami, N., Oda, K., and Ikehara, Y. (1990) J. Biochern. (Tokyo) 108,230-234

23, Misumi, Y., Ohkubo, K., Sohda, M., Takami, N., Oda, K., and Ikehara (1990) Biochem. Biophys. Res. Commun. 171,236-242

24. Misumi, Y., Sohda, M., and Ikehara, Y. (1990) Nucleic Acids Res. 18, 2178

25. Green, S., Issemann, I., and Sheer, E. (1988) Nucleic Acids Res. 16,369

26. Godson, G. N., and Vapnek, D. (1973) Biochim. Biophys. Acta

27. Oda, K., Takami, N., Fujiwara, T., Misumi, Y., and Ikehara, Y. (1989) Biochem. Biophys. Res. Commun. 163, 194-200

28. Oda, K., Misumi, Y., and Ikehara, Y. (1983) Eur. J. Biochem. 135,209-216

29. Ogata, S., Misumi, Y., and Ikehara, Y. (1989) J. Biol. Chem. 264 ,

30. Blobel, G., and Dobberstein, B. (1975) J. Cell Bwl. 6 7 , 835-851 31. Fujiwara, T., Oda, K., Yokota, S., Takatsuki, A., and Ikehara, Y.

32. Graham, R. C., and Karnovsky, M. J. (1966) J. Histochem.

33. Markland, F. S., and Smith, E. L. (1967) J. Biol. Chem. 242,

34. Smeekens, S. P., Avruch, A. S., LaMendola, J., Chan, S. J., and Steiner, D. F. (1991) Proc. Natl. Acad. Sci. U. S. A. 88, 340- 344

35. Tartakoff, A., and Vassalli, P. (1983) J. Cell Biol. 97, 1243-1248 36. Virtanen, I., Ekblom, P., and Laurial, P. (1986) J . Cell Biol. 8 5 ,

37. Owen, M. C., Brennan, S. O., Lewis, J. H., and Carrell, R. W .

38. Brennan, S. O., and Peach, R. J. (1988) FEBS Lett. 2 2 9 , 167-

39. Davidson. H. W.. Rhodes. C. J., and Hutton, J. C. (1989) Nature

S. A. 83,6766-6770

265,22075-22078

EMBO J. 5,2197-2202

2997-3000

299,516-520

3596-3601

(1988) J. Biol. Chem. 263, 18545-18552

Cytochem. 14 , 291-302

5198-5211

429-434

(1983) N. Engl. J . Med. 309 , 694-698

170

333,93-96 '

Chem. 264,14240-14245 40. Rhodes, C. J., Brennan, S. O., and Hutton, J. C. (1989) J. Bwl.

41. Oda, K., and Ikehara, Y. (1988) J. Biochem. (Tokyo) 104, 159- 161

6050

Boileau, G. (1987) J . Biol. Chem. 2 6 2 , 1876-1881

Chem. 263,13456-13462

42. Laub, 0.. and Rutter, W. J. (1983) J. Bwl. Chern. 258, 6043-

43. Noel, G., Zollinger, L., Lariviere, N., Nault, C., Crine, P., and

44. Lapps, W., Eng, J., Stern, A. S., and Gubler, U. (1988) J. Biol.

45. Warren, T. G., and Shields, D. (1984) Cell 39,547-555 46. Wise, R. J., Pittman, D. D., Handin, R. I., Kaufman, R. J., and

47. Hanley, J. M., Haugen, T. H., and Heath, E. C. (1983) J. Biol. Orkin, S. H. (1988) Cell 62, 229-236

Chem. 258,7858-7869

Proprotein-processing Endoprotease Furin 16959

48. Misumi, Y., Tanaka, Y., and Ikehara, Y. (1983) Biochem. Biophys.

49. Fuller, R. S., Wilcox, C., and Redding, K. (1990) J. Cell. Biochem.,

50. Wells, J. A., Ferrari, E., Henner, D. J., Estell, D. A., and Chen,

51. Power, S. D., Adams, R. M., and Wells, J. A. (1986) Proc. Natl.

52. Brennan, S. O., Peach, R. J., and Bathurst, I. C. (1990) J. Biol.

Res. Commun. 114,729-736

Suppl. 14C, 69 (abstr.)

E. Y. (1983) Nucleic Acids Res. 11 , 7911-7925

Acad. S C ~ . U. S. A. 83, 3096-3100

Chem. 265,21494-21497

53. van de Ven, W. J. M., Voorberg, J., Fontijn, R., Pannekoek, H., van den Ouweland, A. M. W., van Duijnhoven, H. L. P., Roebroek, A. J. M., and Siezen, B. J. (1990) Mol. Biol. Rep. 14,

54. Wise, R. J., Barr, P. J., Wong, P. A., Kiefer, M. C., Brake, A. J., and Kaufman, R. J. (1990) Proc. Natl. Acad. Sci. U. S. A. 87,

55. Bresnahan, P. A,, Leduc, R., Thomas, L., Thorner, J., Gibson, H. L., Brake, A. J., Barr, P. J., and Thomas, G. (1990) J. Cell Biol.

265-275

9378-9382

1 1 1,2851-2859