Inhibition of Soluble Recombinant Furin by Human Proteinase Inhibitor 8*

Furin is a ubiquitous prototypical mammalian kexin/subtilisin-like endoproteinase that is involved in the proteolytic processing of a variety of proteins in the exocytic and endocytic pathways, with cleavage occurring at the C terminus of the minimal consensus furin recognition sequence Arg-Xaa-Xaa-Arg. In this study, human proteinase inhibitor 8 (PI8), a widely expressed 45-kDa ovalbumin-type serpin that contains two sequences homologous to the minimal sequence for recognition by furin in its reactive site loop, was tested for its ability to inhibit a recombinant soluble form of human furin. PI8 formed an SDS-stable complex with furin and inhibited its amidolytic activity via a two-step mechanism with ak assoc of 6.5 × 105 m −1 s−1 and an overallK i of 53.8 pm. Thus, PI8 inhibits furin in a rapid, tight binding manner that is characteristic of physiological serpin-proteinase interactions. PI8 is not only the first human ovalbumin-type serpin to demonstrate inhibitory activity toward furin, but it is also the first significant inhibitor of furin identified that is not a serpin reactive site loop mutant, either naturally occurring or engineered.

The mammalian serine proteinase inhibitors, or serpins, are a superfamily of proteins that regulate proteolytic events in a wide variety of physiological processes including but not limited to blood coagulation, viral and parasite pathogenicity, intracellular proteolysis, and tumor suppression (1). Serpins inhibit their target proteinases by forming a 1:1 stoichiometric complex with the active site of the proteinase, which is in most cases resistant to denaturants (2). Serpins are composed of three ␤-sheets surrounded by eight ␣-helices and a reactive site domain that is highly divergent among serpin family members and exists as a stressed loop with a canonical conformation that confers the optimal conformation for high affinity association with the substrate binding cleft of the cognate proteinase (3)(4)(5)(6).
The P 1 -P 1 Ј 1 residues of the reactive site domain determine the inhibitory specificity of the serpin and act as a pseudosubstrate for the target proteinase (6). Unlike a typical substrate, the serpin has the ability to form a tight complex that may be essential for inactivation of the proteinase (4,(7)(8)(9). Most serpins can interact with more than one proteinase in vitro, but the affinity of such interactions must be determined to suggest physiological relevance.
Ovalbumin represents the parent prototype of a unique family of serpins whose members include plasminogen activator inhibitor-2 (PAI-2), 2 an elastase inhibitor isolated from monocyte-like cells, squamous cell carcinoma antigen (SCCA), maspin, proteinase inhibitor (PI) 6, PI8, PI9, bomapin, and SCCA2 (10,11). PI6, PI8, and PI9 are unique among the mammalian ovalbumin-type serpins in that they contain a cysteine residue in the P 1 Ј position within the reactive site domain, which is also present in the viral serpin CrmA (12). Ovalbumintype serpins lack a typical cleavable N-terminal signal sequence but have been found to reside intracellularly (13,14) or both intracellularly and extracellularly (15)(16)(17)(18). Therefore, it can be inferred that the functions of members of the ovalbumin family of serpins may not be strictly confined to the cytoplasm. It has been demonstrated that individual mammalian ovalbumin-type serpins can inhibit a variety of prototypic serine proteinases by distinctly different mechanisms using a variety of kinetic parameters (19). Although most members of the ovalbumin family of serpins exhibit defined proteinase inhibitory activity, the true physiological targets of these serpins have not yet been identified.
PI8 is a 45-kDa serpin that contains the sequence Arg 336 -Asn 337 -Ser 338 -Arg 339 at the P 4 -P 1 positions in the reactive site domain, as well as the sequence Arg 339 -Cys 340 -Ser 341 -Arg 342 at the P 1 -P 3 Ј positions, both of which conform to the minimal sequence required for efficient processing by furin, Arg-Xaa-Xaa-Arg (20). Additionally, PI8 was recently demonstrated to be a potent inhibitor of the Bacillus subtilis dibasic endoproteinase subtilisin A (21). A number of mammalian convertases have been identified that demonstrate a high degree of functional and structural similarity to yeast kexin and bacterial subtilisin, of which furin is the prototype (reviewed in Ref. 22). Furin is a ubiquitously expressed, membrane-associated, calcium-dependent serine endoproteinase that cleaves a wide variety of precursor proteins in both the exocytic and endocytic pathways. Furin cleaves at the C terminus of Arg-Xaa-Xaa-Arg motifs and is involved in the proteolytic processing of the von Willebrand factor precursor, pro-factor IX precursor, the low density lipoprotein receptor-related protein, pro-␤-nerve growth factor, viral superantigens, and diphtheria toxin (22). In addition, several viral coat proteins, including influenza virus hemagglutinin, measles virus fusion protein, and HIV-1 gp160 (22), are cleaved and activated by furin, a process crucial in the establishment of viral infectivity. In the present study, we demonstrate that PI8 inhibits furin in a rapid, tight binding manner that is characteristic of physiological serpin-proteinase interactions.

EXPERIMENTAL PROCEDURES
General Kinetic Methods-Recombinant human furin was prepared from a vaccinia virus construct (20). Recombinant human PI8 was prepared as described (21). The K m for furin and the fluorogenic substrate Pyr-Arg-Thr-Lys-Arg-4-methylcoumaryl-7-amide (pERTKR-MCA, Bachem) as well as the catalytically active concentration of furin were determined as described previously (23). The K m for furin and pERTKR-MCA was determined to be 3.2 M. Active site-titrated furin was used to determine the amount of PI8 needed for a 1:1 molar binding stoichiometry for the determination of kinetic constants. Furin (1.25 nM) was mixed with increasing amounts of PI8 in a total volume of 180 l in 100 mM HEPES (pH 7.5) containing 0.5% Triton X-100 and 1 mM CaCl 2 . Reactants were incubated for 30 min at 37°C, and pERTKR-MCA was added to a final concentration of 50 M. The enzymatically released 7-amido-4-methylcoumarin was then detected at 25°C using an SLM Instruments SLM-8000 spectrofluorimeter with an excitation wavelength of 370 nm and an emission wavelength of 460 nm. The data were used to plot the enzymatic rate of substrate hydrolysis against the amount of PI8 used in the reaction. Linear regression to the x axis was used to calculate the amount of PI8 required for a 1:1 molar binding stoichiometry with furin.
Slow Binding Inhibition Kinetics-Inhibition progress curves were obtained under pseudo-first order conditions by incubating the reactants in 0.3 ml of the same buffer used for the titration of PI8. Reactions were started by the addition of enzyme to a solution containing the fluorogenic substrate and the appropriate inhibitor concentration. Reactions for each experiment were started within 30 s, and the enzymatic production of 7-amido-4-methylcoumarin was detected as described earlier. The final concentrations of the reactants were 2 nM furin, 100 M pERTKR-MCA, and 4, 8, 12, and 16 nM PI8. Spontaneous substrate hydrolysis was measured in separate experiments and determined to be negligible. The reactions were allowed to proceed until steady-state velocity was attained, and the data were fitted to the integrated rate equation for slow binding inhibition (24) by nonlinear regression using UltraFit 3.0 software (Biosoft) to obtain values for the initial velocity (v o ), the steady-state velocity (v s ), the initial fluorescence (A o ), and the apparent first order rate constant (kЈ) for the establishment of steady-state equilibrium of the proteinaseinhibitor complex. The data obtained from nonlinear regression analysis were then used in various graphical transformations (25)(26)(27)(28)(29) to obtain the inhibition and rate constants for the interactions of PI8 with recombinant human furin. Detection of SDS-Stable Furin-PI8 Complexes-Antibodies against recombinant human PI8 were generated in rabbits (30), and the IgG fraction was purified by protein A-Sepharose column chromatography. Furin and PI8 were incubated for 15 min at 37°C, and the reaction mixtures were subsequently subjected to 10% SDS-PAGE under reducing conditions (31) and electrophoretically transferred to a nitrocellulose membrane in 10 mM CAPS (pH 11) buffer containing 10% methanol. The membrane was blocked with 1% nonfat dry milk in Trisbuffered saline containing 0.02% azide, and complexes were detected by incubating the membranes with rabbit anti-PI8 IgG, followed by incubation with 125 I-labeled protein A and autoradiography.

Inhibition of Human Furin by PI8 -Preliminary
studies indicated that the interaction between furin and PI8 obeyed slow binding inhibition kinetics, as the amidolytic activity of furin inhibited by PI8 attained steady-state equilibrium and the data were successfully fitted to Equation 1. On average, ten PI8 molecules were required to form a stable inhibitory complex with one molecule of furin, as determined by titration. The kinetic characterization of the inhibition of furin by PI8 was performed using PI8 concentrations ranging from two to eight times the molar concentration of furin. A family of inhibition progress curves representative of the interaction between furin and PI8 at the chosen PI8 concentrations is shown in Fig. 1. As expected, steady-state equilibrium was achieved more readily as the concentration of PI8 in the reaction mixture increased. Data obtained from the inhibition progress curves were fitted to Equation 1 by nonlinear regression analysis, and the results indicated that the initial velocity, v o , was inversely proportional to the concentration of PI8 for each set of progress curves. This suggests that the slow onset of the inhibition of furin by PI8 follows the two-step mechanism P L | ; where a loose proteinase-inhibitor (PI) complex is rapidly formed, followed by a slow isomerization to the tight PIЈ complex (25). This observation was confirmed by plotting v max /v o against the PI8 concentration, which indicated a linear relationship with a positive slope (data not shown). The K i for the formation of the initial loose complex was calculated from the slope of the line using the , and was estimated to be 7.2 Ϯ 1.2 nM (n ϭ 4). In addition, the apparent first order rate constant kЈ was found to increase as PI8 concentration increases, which is consistent with the proposed mechanism. The overall second order association rate constant k assoc was determined by plotting log( To determine the rate constant for the reverse isomerization step k Ϫ2 of the furin-PI8 tight complex, a plot of kЈ against v o /v s was generated. This plot was linear (data not shown), and k Ϫ2 was calculated directly from the slope of the line to be 2.5 Ϯ 0.3 ϫ 10 Ϫ5 s Ϫ1 (n ϭ 4). Using the relationship t1 ⁄2 ϭ 0.693/k Ϫ2 , a half-life of 7.7 h was estimated for the reverse isomerization of the tight complex to the loose complex. The value of the rate constant for the formation of the tight complex k 2 was determined by fitting a plot of kЈ versus PI8 concentration (data not shown) to the hyperbolic equation (24) kЈ ϭ k Ϫ2 ϩ k 2 by nonlinear regression analysis. Using this method, k 2 was estimated to be 3.3 Ϯ 0.2 ϫ 10 Ϫ3 s Ϫ1 (n ϭ 4). To verify that the interaction of furin and PI8 occurs by the suggested mechanism and to justify the use of a hyperbola to describe the relationship between kЈ and [PI8], a double-reciprocal plot of 1/(kЈ Ϫ k Ϫ2 ) versus 1/[PI8] was generated (Fig. 2) using the values obtained from Equations 1 and 2 that is linear and crosses the positive y axis at a point approximately equal to 1/k 2 for the mechanism suggested earlier (29). The y intercept of the plot in Fig. 2 was used to determine a value for k 2 of 2.1 Ϯ 0.8 ϫ 10 Ϫ3 s Ϫ1 (n ϭ 4), which is reasonably close to the value of k 2 determined by Equation 2. More importantly, the plot in Fig. 2 justifies the manipulation of data and determination of kinetic constants according to the suggested mechanism.
Formation of a PI8-Furin SDS-Stable Complex-Complex formation between furin and PI8 was visualized by Western blotting using rabbit anti-human PI8 IgG. As shown in Fig. 3, incubation of furin with PI8 resulted in the formation of an SDS-stable complex that migrated with an apparent molecular mass of ϳ225 kDa following reduction with 2-mercaptoethanol. In the absence of reducing agent, the complex migrated with an apparent molecular mass of ϳ200 kDa (data not shown). Additionally, no complex was observed following incubation of furin with a PI8 preparation that had been previously heat denatured (data not shown). The formation of a tight, SDSstable complex between furin and PI8 is also consistent with the inhibition mechanism described earlier. DISCUSSION In the present study, we have performed a detailed kinetic analysis of the inactivation of human furin by PI8. PI8 inhibited furin via a two-step mechanism characterized by the rapid formation of an initial loose complex followed by a slow isomerization to a tight, stable complex that was visualized by SDS-PAGE followed by Western blotting. The apparent molecular mass of the complex was approximately 225 kDa, which is significantly higher that the predicted mass of ϳ140 kDa. Although the reason for the anomalously high molecular mass of the complex is unknown and will require further investigation, the aberrant migration may arise either from aggregation of furin-PI8 complexes or incomplete denaturation of the complex by SDS. It is unlikely, however, that this apparently higher molecular mass occurs as a result of an alternative binding stoichiometry, because furin has only one catalytic center and PI8 has only one reactive site loop to facilitate the interaction. A second immunoreactive species migrating at approximately 42 kDa was seen in lane 2 of Fig. 3, which most likely represents PI8 cleaved as a result of the interaction with furin. The overall inhibition constant for the inactivation of furin by PI8 was 53.8 pM, indicating that PI8 is a potent inhibitor of this proteinase. The initial loose complex of furin and PI8 had a K i of 7.2 nM, which is similar to the K i values of 8 and 6.6 nM for the inhibition of plasmin and chymotrypsin by ␣ 2 -antiplasmin, respectively (26). The furin-PI8 loose complex is converted to the tight complex at a rate of 3.3 ϫ 10 Ϫ3 s Ϫ1 , which is comparable with the rates reported for chymotrypsin-␣ 2 -antiplasmin and plasmin-␣ 2 -antiplasmin complexes (26). In addition, the k assoc for furin and PI8 was determined to be 6.5 ϫ 10 5 M Ϫ1 s Ϫ1 , which was lower than the rate of inhibition of subtilisin A by PI8 (1.2 ϫ 10 6 M Ϫ1 s Ϫ1 ) (21) but exceeded the rates of inhibition of plasma kallikrein by C1-inhibitor (6.9 ϫ 10 4 M Ϫ1 s Ϫ1 ) (32) and human thrombin and coagulation factor Xa by PI8 (1.0 ϫ 10 5 M Ϫ1 s Ϫ1 and 7.5 ϫ 10 4 M Ϫ1 s Ϫ1 , respectively) (21), as well as the rate of inhibition of granzyme B by CrmA (2.9 ϫ 10 5 M Ϫ1 s Ϫ1 ) (33). These comparisons indicate that the kinetic constants for the inhibition of furin by PI8 are of physiological significance. Furthermore, PI8 is the only ovalbumin-type serpin, as well as the only naturally occurring intracellular human serpin not associated with a disease state, demonstrated to be a significant inhibitor of furin. Previously, only peptide chloromethylketones and an ␣ 1 -antitrypsin (␣ 1 -AT) variant have been described as significant inhibitors of furin. ␣ 1 -AT Portland is an engineered variant of ␣ 1 -AT Pittsburgh (34) that carries an additional Ala 355 3 Arg mutation in its reactive site domain to provide the minimal consensus sequence for efficient recognition and processing by furin (35). PI8 contains the sequence Arg 336 -Asn 337 -Ser 338 -Arg 339 at the P 4 -P 1 positions in the reactive site domain that, based upon sequence alignment, is presumably recognized by the substrate binding cleft of furin in this interaction. Interestingly, PI8 contains a second sequence Arg 339 -Cys 340 -Ser 341 -Arg 342 at the P 1 -P 3 Ј positions in the reactive site domain, which also may be involved in the interaction of PI8 with furin or another mammalian convertase. The precise sequence in the PI8 reactive site domain involved in the interaction between furin and PI8 is presently unknown.
In order for PI8 to inhibit furin in vivo, PI8 must presumably enter the secretory pathway. The ovalbumin-type serpins PAI-2, SCCA, and maspin each lack a cleavable N-terminal signal sequence, but all can be found extracellularly. PAI-2 contains two hydrophobic regions proximal to the N terminus, designated as H 1 and H 2 , that are involved in its secretion through facultative polypeptide translocation. H 2 is also homologous to the uncleaved secretion signal in ovalbumin (17). PAI-2 constructs carrying a deleted H 1 or H 2 region transfected into Chinese hamster ovary cells demonstrated decreased glycosylation and secretion, whereas mutants displaying increased hydrophobicity in the H 1 or H 2 regions exhibited increased glycosylation and secretion. PI8 shares 62 and 65% amino acid sequence identity with the H 1 and H 2 regions in PAI-2, respectively (Fig. 4). However, if amino acid mismatches are allowed to be resolved by the presence of a hydrophobic amino acid in the PI8 sequence, as shown by an asterisk in Fig.  4, PI8 shares 69 and 100% identity with the H 1 and H 2 regions of PAI-2, respectively. These hydrophobic regions near the N terminus in PI8 may permit secretion of PI8 in a manner analogous to PAI-2 and, ultimately, the interaction of furin and PI8. Potential PI8 secretion may also be regulated at the level of transcription. In the case of PAI-2, both the cytosolic and secreted forms are encoded by a single mRNA as seen by Northern blot analysis (17). However, a Northern blot of poly(A) ϩ mRNA from a wide variety of human tissues revealed two PI8 transcripts of 3.8 and 1.4 kilobases (12). In this connection, yeast invertase and human gelsolin are two proteins that each have cytoplasmic and secreted forms. The cytoplasmic and secreted forms of invertase are encoded by two mRNA forms transcribed from a single gene through the use of different transcriptional initiation sites, where the transcript for the cytoplasmic form initiates within the invertase signal sequence (36). The cytoplasmic and secreted forms of gelsolin are also encoded by two mRNA forms transcribed from a single gene through the alternative use of two promoters. The 5Ј region of the mRNA coding for cytosolic gelsolin is derived from two exons that encode an untranslated sequence and translation starts at an internal exon common to both mRNAs, whereas the 5Ј end of the mRNA coding for secreted gelsolin is derived from a single unique exon that encodes an N-terminal signal sequence (37). Thus, the two distinct PI8 mRNA species may serve a purpose analogous to those of invertase and gelsolin.
The mammalian processing endoproteinase furin is actively involved in normal cellular processes and has also been linked to pathological situations. In this regard, furin is involved in the cleavage and activation of diphtheria toxin, a process that is important for cytotoxicity, and is also a processor and activator of the coat proteins of such viruses as influenza virus, measles virus, and HIV-1, a process essential for virus infectivity. The interaction of furin with PI8 results in the rapid formation of a tight complex with a relatively long half-life. Because PI8 contains two hydrophobic regions proximal to the N terminus nearly identical to the internal secretion signals in PAI-2, as well as two forms of PI8 mRNA, also seen with yeast invertase and human gelsolin, it is not unreasonable to suggest that a fraction of the PI8 synthesized may be secreted and interact with furin under normal conditions or in response to specific stimuli. Therefore, PI8 may regulate the activity of furin and, in turn, such events as pro-protein processing and virus infectivity by its secretion, either through facultative polypeptide translocation facilitated by hydrophobic interactions or by alternative transcriptional initiation to produce mRNA encoding PI8 that contains a cleavable N-terminal signal sequence.