Arg15-Lys17-Arg18 turkey ovomucoid third domain inhibits human furin.

Turkey ovomucoid third domain with Leu18 in its reactive site is a potent inhibitor of many serine proteinases: subtilisins, chymotrypsins, and elastases. Previous studies showed that an L18K mutation made it a moderately strong inhibitor of trypsin, while an L18E mutation made it a strong inhibitor of Glu-specific Streptomyces griseus proteinase (GluSGP). For human furin substrates the consensus optimal sequence is RXKR decreases. Therefore the A15R, T17K, and L18R mutations were made in turkey ovomucoid third domain. The mutant inhibits human furin with a Ka of 1.1 x 10(7) M-1. As human furin catalyzes an obligatory step in human immunodeficiency virus proliferation, this inhibitor, along with the others already available, deserves further study.


studies showed that an LlSK mutation made it a moderately strong inhibitor of trypsin, while an L18E mutation made it a strong inhibitor of Glu-specific Streptomyces griseus proteinase (GluSGP). For human furin substrates the consensus optimal sequence is RXKRJ.
Therefore the A15R, T17K, and L18R mutations were made in turkey ovomucoid third domain. The mutant inhibits human furin with a K , of 1.1 x lo' M -~. As human furin catalyzes an obligatory step in human immunodeficiency virus proliferation, this inhibitor, along with the others already available, deserves further study.
Furin is a member of the recently characterized class of subtilisin-related proprotein convertases or SPCsl (for a review see Steiner et al. (1992); see also Chan et al. (1992) for nomenclature). Of the described mammalian convertases furin is probably most broadly distributed among various tissues. Like yeast Kex2 (Fuller et al., 1988) it has a COOH-terminal transmembrane domain and is an integral membrane protein. Thus a COOH-terminally truncated fragment has been developed and used for detailed studies on furin in vitro (Molloy et al., 1992). A broad consensus has been reached that the sequence RXKFtJ is the optimal sequence for furin (the 1 indicates the scissile bond) (Hosaka et al., 1991;Bresnahan et al., 1993). However, a detailed analysis (Molloy et al., 1992) suggests that the RXXRJ is sufficient for efficient processing of at least some GM 10831 (to Purdue University), DK 44629 (to G. T.), AG 10462 (to * This work was supported by National Institutes of Health Grants S. A,), and GM 08339 (to K. R.), National Research Service Award DK08703 (to S. S. M.), and National Science Foundation Grant 9018707 (to S. A,). 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. ' The abbreviations used are: SPC, subtilisin-related proprotein convertases; HIV, human immunodeficiency virus; GluSGP, glutamic acidspecific S. griseus proteinase. furin substrates in vitro. This is in contrast to the more typical SPCs, where K R J and RRJ sequences appear to be the principal determinants of specificity.
Furin appears to be responsible for maturation of many endogenous proproteins. Moehring et al. (1993) showed that mutated Chinese hamster ovary cells with elevated resistance to several viruses and bacterial toxins lose this resistance when transfected with cDNA for mouse furin. This strongly implicates furin as the enzyme responsible for maturation of precursor viral membrane glycoproteins. Cleavage of glycoprotein gp160 to a heterodimer composed of gp120 and gp41 is an essential step in human immunodeficiency virus (HIV) proliferation (McCune et al., 1988). Hallenberger et al. (1992) strongly suggest that human furin is the enzyme responsible for this cleavage since the cleavage is inhibited by the use of two designed small molecule inhibitors (Stieneke-Grober et al., 1992), decanoyl RAKR chloromethylketone and decanoyl REKR chloromethylketone.
We have a strong interest in designing inhibitors for serine proteinases based on the turkey ovomucoid third domain, OMT-KY3. OMTKY3 is a "standard mechanism" (Laskowski and Kato, 1980) protein proteinase inhibitor with Leu'' as its P1 residue (Schechter and Berger, 1967) (Fig. 1) and is a member of the Kazal family of protein inhibitors (Laskowski and Kato, 1980). It is a powerful inhibitor of many serine proteinases that show a preference for neutral residues at PI. When glutamic acid-specific Streptomyces griseus proteinase (GluSGP) (Yoshida et al., 1988) became available, Komiyama et al. (1991) converted OMTKY3, which is ineffective against GluSGP, into a powerful inhibitor by a simple L18E substitution at PI (K, = 5.5 x 1O1O M -~) . Similarly OMTKY3 is ineffective against trypsin, but an L18K mutation at P1 turned it into a moderately good inhibitor of trypsin (K, = 3.2 x lo8 M -~) . Therefore it seemed possible that the introduction of the optimal consensus sequence for furin into the reactive site region of turkey ovomucoid third domain by making the AER, T17K, and L18R replacements (see Fig. 1) might produce a furin inhibitor. At that time an expression system for ovomucoid third domain variants became available at Rutgers. When the mutant was expressed it was found to be a modest inhibitor of the truncated form of human furin (Molloy et al., 1992). I t may be worth pointing out that many other inhibitor frames such as, for example, bovine trypsin inhibitor (Kunitz) or soybean trypsin inhibitor (Kunitz) might have also been chosen as a frame for introduction of the furin consensus sequence into the reactive site region. The reason for choosing OMTKY3 was that it is a powerful inhibitor of subtilisin (Empie and Laskowski, 1982) while the other two inhibitors do not interact with subtilisin well because of the overall structures of their frames (Hirono et al., 1979). As SPCs and therefore furin are subtilisin homologs it may be that OMTKY3 is better suited as a furin inhibitor.

MATERIALS AND METHODS
Expression of Furin-Truncated human furin, terminating at Leu713 and thus omitting the transmembrane domain, was expressed in African Green monkey kidney cells BSC-40 from a vaccinia virus recombinant W:hFUR713t as described by Molloy et al. (1992). The final purification was on a Pharmacia LKl3 Biotechnology Inc. MonoQ 5/5 anion exchange column. The eluted 250-pl fractions were monitored by a furin enzymatic assay using a fluorogenic peptide substrate and SDS-polyacrylamide gel electrophoresis. The three fractions constituting the peak of activity served as furin source in this research.

RESULTS AND DISCUSSION
Addition of massive amounts of OMTKY3 (Fig. 1, top left) produced insignificant inhibition of furin activity. We estimate that K, for this inhibitor is 200 M -~ or less. Addition of the L18R variant (Fig. 1, middle right) in very large amounts caused small but noticeable inhibition. K, is approximately 450 M -~. However, when the A15R,T17K,L18R variant (Fig. 1, bottom  left) was employed clear inhibition was observed (Fig. 2). Fitting of these data caused some problems. Normally the operational molarity of our enzymes is measured prior to the experiment either by burst titration or by titration with a very strong inhibitor. Neither is available for human furin. Therefore, an iterative fitting procedure was employed to yield the molar furin concentrations given on the vertical axis of Fig. 2. The equilibrium constant K, is 1.1 x lo7 M -~. As can be seen the fit in Fig. 2 is quite good. A15R,T17K,L18R OMTKY3 is a moderate inhibitor of furin. We believe that in subsequent design iterations it can be improved as more and more is being published about the detailed specificity of furin. Attempts to improve the inhibitor might prove helpful for unraveling this specificity as well, because K, values are far more quantitative than most observations on cleavage. However, even with a K, of 1.1 x lo7 M -~ the inhibitor may be useful in various biological studies on the role of furin and on the HIV virus replication. Indeed tests on this inhibitor have begun. It may be worth noting that the fusion protein that was designed to facilitate the expression of OMTKY3 may in fact help to target the inhibitory variant to appropriate cells.
One of the uses of the new "standard mechanism" furin inhibitor is that operational normality of furin can now be determined, albeit in a cumbersome way. Should it turn out to be possible to improve the K, to >5 x 10' M -~ the operational normality determination will become quite straightforward by simple titration.
The relative ease with which OMTKY3 was converted to a moderate inhibitor of furin raises the question: has nature already developed many such inhibitors? This question becomes more relevant when we recall that the combining regions of most protein proteinase inhibitors are hypervariable (Laskowski and Kato, 1980;Hill and Hastie, 1987;Laskowski et al., 1987Laskowski et al., , 1988Creighton and Darby, 1989). We have therefore searched sequence data bases of standard mechanism protein proteinase inhibitors for the RXXRJ sequence, where the arrow defines the position of the inhibitor's reactive site. Surprisingly, we found very few, not because R is rare at PI (there are many trypsin inhibitors) but because R is rare at P4. In the  , not of the connecting peptide extended third domains. As the extended third domain has been bering system. Thus NH,-terminal Val shown here is residue 6. The used in crystallographic studies it has become the basis for the num-"missing" NH, terminus has no effect on interaction with enzymes . The P,, P,' system of Schechter and Berger The 12 residues forming the consensus enzyme-inhibitor contact set (1967) is also shown. The arrow indicates the reactive site peptide bond.
this work are in black. (Apostol et al., 1993) are in boldface circles. The mutations introduced in Expression of OMTKY3 and Its Variants"OMTKY3 and its sitespecific mutagenesis variants were expressed as fusion proteins with a protein A fragment in Escherichia coli. The expression system developed at Rutgers will be the subject of another paper and was presented as a poster (Lu et al., 1992). The fusion protein was secreted into the periplasmic space where the disulfide bridges in the OMTKY3 domain form spontaneously. After osmotic shock it was isolated by affinity chromatography on a Pharmacia IgG column. The inhibitory domain was split off by CNBr cleavage in the linker (the only Met in the fusion protein) and isolated by size exclusion and ion exchange chromatography. The three proteins: 6-56 OMTKY3, Arg18 6-56 OMTKY3, and Arg15, Lys17, Arg18 6-56 OMTKY3 (see Fig. 1) were characterized by amino acid analysis, sequencing (Porton 2020) from residue 6 through 19 and by mass spectrometry (Westec) with electrospray injection. All of the analyses, sequences, and molecular masses were in accord with expectations based on Fig. 1.
Concentration and Activity Determinations-Molar concentration of the inhibitor variants was determined by injecting aliquots of the inhibitor solution onto a TSK G2000 analytical size exclusion column. The effluent was monitored at 206 nm by an LKB detector, and the area under the inhibitor peak was integrated. This was compared with the area under an equal size aliquot of standardized OMTKY3. As OMTKY3 is a powerful inhibitor of many serine proteinases it can be readily standardized.
The activity of human furin was determined by monitoring the hydrolysis of 1.2 x M (final concentration) of t-butyloxycarbonylarginylvalylarginylarginyl-4-methylcoumarin-7-amide. The fluorescence emission was measured at 460 nm (10-nm slit width) on a Perkin-Elmer LS50 spectrofluorometer using 370 nm wavelength (10-nm slit width) for excitation. The reactions were conducted in 0.1 M HEPES, 0.5% Triton X-100, 0.001 M CaCl,, 0.001 M p-mercaptoethanol, pH 7.50, at 30 "C. The enzyme and inhibitor were incubated together in the reaction medium for 2 h to reach equilibrium. Then substrate was added, and substrate hydrolysis was allowed to proceed for 16 h and then stopped by addition of a 60-fold excess of 1 m M ZnCl, solution. The fluorescence of these solutions was then measured.  (Fig. 1). The curve drawn through the points is the best fit for K , = 1.1 x lo7 M -~. pH is 7.5 and temperature is 30 "C (for buffer conditions see "Materials and Methods"). The molar concentrations of the total inhibitor present were measured directly, but the molar concentrations of free furin were obtained from these data by an iterative process, The dashed line is what would be expected if the inhibition were stoichiometric. large set of over 400 Kazal domain sequences at Purdue, we found only one with P4 R. This sequence also has P1 R and is chicken ovoinhibitor's fourth domain (Scott et al., 1987). In most inhibitor families we found no R X X R J sequences, but they are abundant in the squash inhibitor family. One of these, an inhibitor from watermelon Citrullus vulgaris trypsin inhibitor I (Otlewski et al., 1987) was available to us. Therefore we tested Citrullus vulgaris trypsin inhibitor I and entire chicken ovoinhibitor for inhibition of furin. Neither was effective. Both have . . .RCPRJI. . . sequences, and their failure to inhibit may be due to either Pro in the P2 position or to Ile in the P1' position or to more remote residues. The question of whether evolution avoids producing natural furin inhibitors is still open.