Single Amino Acid Substitution in the PC1/3 Propeptide Can Induce Significant Modifications of Its Inhibitory Profile toward Its Cognate Enzyme*

The proprotein convertase PC1/3 is synthesized as a large precursor that undergoes proteolytic processing of the signal peptide, the propeptide and ultimately the COOH-terminal tail, to generate the mature form. The propeptide is essential for protease folding, and, although cleaved by an autocatalytic process, it remains associated with the mature form acting as an auto-inhibitor of PC1/3. To further assess the role of certain residues in its interaction with its cognate enzyme, we performed an alanine scan on two PC1/3 propeptide potential cleavable sites (50RRSRR54 and 61KR62) and an acidic region 65DDD67 conserved among species. Upon incubation with PC1/3, the ensuing peptides exhibit equal inhibitory potency, lower potency, or higher potency than the wild-type propeptide. The Ki values calculated varied between 0.15 and 16.5 nm. All but one mutant exhibited a tight binding behavior. To examine the specificity of mutants, we studied their reactivity toward furin, a closely related convertase. The mutation of certain residues also affects the inhibition behavior toward furin yielding propeptides exhibiting Ki ranging from 0.2 to 24 nm. Mutant propeptides exhibited against each enzyme either different mode of inhibition, enhanced selectivity in the order of 40-fold for one enzyme, or high potency with no discrimination. Hence, we demonstrate through single amino acid substitution that it is feasible to modify the inhibitory behavior of propeptides toward convertases in such a way as to increase or decrease their potency, modify their inhibitory mechanisms, as well as increase their selectivity.

One of the most common methods used by cells to diversify the pool of their biologically active molecules is protein processing. Indeed, numerous secreted proteins are synthesized first as an inactive precursor, which is rendered biologically active upon cleavage at clusters of basic residues. Members of a family of proteins named proprotein convertases (PCs) 3 primarily perform this cleavage. To date, seven members were described: furin, PC1/3, PC2, PACE4, PC4, PC5/6, and PC7/ PC8/lymphoma proprotein convertase. Some of them are ubiquitously expressed such as furin, PACE4, and PC7, whereas others exhibit a more restricted expression pattern such as PC1/3 and PC2, which are solely present in endocrine and neuroendocrine tissues, and PC4, which is expressed only in germ cells. However, all of them belong to the larger family of serine proteases and are structurally related to bacterial subtilisin and yeast kexin (reviewed in Ref. 1). In terms of biological activities, numerous transfection experiments using recombinant enzymes and substrates, generation of knock-out animals as well as human cases of convertase deficiency pointed out the importance of convertases in crucial biological processes such as patterning during embryogenesis, angiogenesis, prohormone processing, tissue remodeling, and complement activation (2). Furthermore, some members are also implicated in many disease states, because they are able to activate various bacterial toxins and to process viral envelope glycoproteins needed for cell penetration (3). For all these reasons convertases represent attractive therapeutic targets.
Structurally, they share common features such as the presence of (i) a signal peptide guiding the protein to the secretory pathway, (ii) a propeptide (alternatively called prosegment, prodomain, or proregion), implicated in enzyme folding and inhibition, (iii) a catalytic domain that possesses the classic catalytic triad Asp, His, and Ser residues conserved among serine proteases, (iv) a P domain, which appears to regulate Ca 2ϩ and pH dependence of the enzyme, and finally, (v) a COOH-terminal tail often defining the localization in organelles and cells. Such a topological distribution of structural regions makes a convertase an efficient self-controlled molecule whereby activation of its zymogen into its active form requires sequential removal of particular segments. This characteristic is best exemplified by the complex activation of PC1/3, which requires removal of the signal peptide, removal of the propeptide, and further cleavage of the COOH-terminal tail.
Among these excised regions, the propeptide plays a pivotal role in enabling efficient functional expression of convertases. Indeed, not only does it prevent undue and untimely activation, but also it is proposed to act as an intramolecular chaperone. Well documented with degradative subtilisins, this feat relies on intimate protein-protein interactions between the propeptide and its cognate enzyme. Hence, concomitantly with the interaction of the COOH terminus of the propeptide where the primary cleavage site is located and the catalytic site, the propeptide could provide the enzyme with a template for acquiring the correct active conformation. However, this role though firmly established with subtilisins has not been widely documented in the case of convertases with the exception of results obtained from expression of chimeric enzymes (4,5) and mutagenesis of some furin and PC2 propeptide residues (6,7). According to Fu et al. (8), the convertase propeptides are best classified as type I propeptides through their role in folding and inhibition of the enzyme. Following the correct folding of the active site, an autocatalytic process leads to cleavage of the propeptide, which can remain associated with the mature form, thus acting as an auto-inhibitor (9,10). Based upon the structural relatedness of proconvertases and prosubtilisins, it is predicted that a secondary cleavage site might exist within the propeptide. Indeed, such a secondary cleavage is present in the case of profurin (11), prokexin (12), pro-PC1/3 (13), and pro-PC2 (7). Proteolytic cleavage at this secondary site is responsible for disassembly of the enzyme⅐inhibitor complex and for preventing further inhibition of the cognate enzyme by an otherwise intact propeptide. In most instances, these cleavages are sufficient to generate fully active enzyme, an exception being the activation of pro-PC1/3, which requires further cleavage in the COOH-terminal region.
Propeptides in the various convertases represent the first 80 -100 residues following the signal peptide. Unlike the very well conserved catalytic and P domains, they show very little sequence conservation except in their COOH-terminal portion. They contain multiple basic residues gathered in two regions namely, the region between positions 45 and 55 and the extreme COOH terminus (Fig. 1). 4 The COOHterminally located basic cluster is strongly conserved and represents the site of the first autocatalytic cleavage. In all other convertases, with the exception of PC7, basic residues containing regions represent the site of the second cleavage leading to the degradation of the propeptide and the release of the fully enzymatically active enzyme. In vitro and overexpression studies showed that mutations or removal of the COOH-terminal basic residues abolish the production of mature enzyme and prevent inhibition of the active protease by the propeptide. Studies conducted by various groups revealed that peptides derived from the middle portion (near the second cleavage site) and the NH 2 -terminal domain of the propeptide were weak inhibitors, whereas peptides derived from the first cleavage site are very potent competitive inhibitors, often in the nanomolar range (11, 14 -16). Moreover, it was also shown that the longer the COOH-terminally derived peptides the better was their inhibitory capacity (11,16,17). Fugère and colleagues (18,19) reported similar results upon investigating the inhibitory potency of all convertases propeptides against furin, PC5/6, and PC7. All these studies taken together indicate that the COOH-terminal portion of the propeptides confers a strong inhibitory potency but appears not to be very discriminative toward its cognate enzyme. Conversely, the remaining NH 2terminal part of the propeptide appears to be responsible for improved selectivity as well as ensuring the tight fit necessary to form the propeptide⅐enzyme complex as seen with furin and pro-PC1/3.
Herein we investigate the ability of certain residues within the propeptide of PC1/3 to confer specificity for its cognate protease. To do so, we performed individual alanine substitution of residues contained within two potential PC1/3 convertase-cleavable clusters as well as substitution of three negatively charged residues, two being specific to PC1/3. We tested the generated mutants in inhibitory assays against PC1/3 and furin. Our results show that the generated mutants can be subdivided in three categories, the first one exhibiting identical inhibition properties to the wild-type propeptide, the second having a diminished inhibition, and finally the third one exhibiting increased inhibition, with K i values ranging from 0.15 to 24 nM. Some of these mutations appear to be advantageous in increasing the inhibitory potency of the propeptide whereas some others clearly can affect the mechanism of inhibition thus conferring selectivity. Finally, we confirm through mutagenesis the location of the second cleavage site in the PC1/3 propeptide.

Expression and Purification of Recombinant mPC1/3 and Human
Furin-Recombinant murine PC1/3 is produced using the baculovirus expression system in Sf9 insect cells and whole larva as recently described (20). Modifications to the original recombinant virus described in Boudreault et al. (21) include substitution of the signal peptide of mPC1/3 by the one of the viral glycoprotein 67 (gp67) to enhance the secretion of the recombinant enzyme into the medium of Spodoptera frugiperda cells. Once expressed, the enzyme is recovered and purified as previously described (20,21). The recombinant soluble (COOH terminus truncated) human furin is obtained from the medium of Sf9 insect cells (13). The enzymatic activity of each recombinant convertase is assayed routinely by fluorometric assays using a fluorogenic substrate (22).
Cloning, Mutagenesis, Expression, and Purification of Recombinant mPC1/3 Propeptides-All the enzymes used for cloning as well as all oligonucleotide primers (Table 1) used in mutagenesis were purchased from Invitrogen. The bacterial expression vector pET24b (ϩ) (Novagen, WI) was modified as previously described (23). The cDNA corresponding to the wild-type (WT) mPC1/3 propeptide was amplified by PCR using the NAD000 and NAD001 primer nucleotides (listed in Table 1) and ligated into the vector between the BamH1 and NotI sites. Similarly, the propeptide mutants were generated by site-directed mutagenesis using the Stratagene QuikChange site-directed mutagenesis kit, as described in the manufacturer's protocol. All propeptide cDNA sequences were verified by DNA sequencing. In addition to the 83 residues of mPC1/3 propeptide, the expression construct contains two extensions at the NH 2 and COOH termini, MASMTGGQQMGRDP and SVQMAAALEHHHHHH, respectively, to facilitate cloning and purification. Each mPC1/3 propeptide, was expressed as a 112-residue His-tagged polypeptide in Escherichia coli strain BL21(DE3) (Novagen) following induction by the addition of 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h at 37°C. Following this period, the cells were harvested through centrifugation.
Propeptide Purification and Chemical Characterization-The bacterial cells were lysed by repeated sonications in the presence of 100 g/ml lysozyme, and the resulting suspension was loaded unto a Ni ϩ2 -Sepharose column (Amersham Biosciences). Following extensive washings of the column, the propeptide was eluted using 1 M imidazole. The eluate was dialyzed against 0.1% acetic acid, and the proteins were subsequently further purified on an analytical Vydac-C 4 RP-HPLC column (25 ϫ 0.46 cm, The Separation Group, Hesperia, CA) using a Varian 9010/9050 chromatography system. The aqueous phase consisted of 0.1% trifluoroacetic acid (v/v) in water, and the elution was carried out first isocratically at 10% organic phase (acetonitrile containing 0.1% trifluoroacetic acid) followed by a 1%/min linear gradient of organic phase to 65% with a flow rate of 1 ml/min. The elution was monitored by measuring absorbency at 225 nm. Typically, all propeptides eluted between 40 and 43% of the organic phase. A specific polyclonal antibody was obtained in rabbits following repeated immunizations using the complete WT-mPC1/3 propeptide (as described above) bacterially produced and purified; according to procedures developed at the Sheldon Biotechnology Center (McGill, Montreal, Canada). Hence, RP-HPLC fractions were analyzed for propeptide content using this antibody through dot blotting. The content of individual immunoreactive fractions was analyzed by SDS-PAGE followed by coloration and Western blotting and subsequently pooled and kept at Ϫ20°C. Routinely, starting with a 1-liter culture, ϳ5-10 mg of each propeptide can be recovered following purification.
The peptide purity and concentration were determined for each mutant by quantitative amino acid analysis following 18 -24 h hydrolysis in the presence of 5.7 N HCl in vacuo at 110°C on a Beckman autoanalyzer (Model 6300) with a postcolumn ninhydrin detection system coupled to a Varian DS604 integrator/plotter. The NH 2 -terminal amino acid sequence, corresponding to ASMTGGQQMGRDPKRQFVNE-(W)AAEI (the underlined sequence indicates the NH 2 terminus of the mature mPC1/3) was determined through automated Edman degradation using an Applied Biosystems Procise 494cLC sequencer (Foster City, CA). Molecular mass determination and mass spectral analysis were done on a Voyager DE-Pro matrix-assisted laser desorption ionization time-of-flight instrument (PerSeptive Biosystems, Cambridge, MA); the propeptide in 0.1% trifluoroacetic acid was mixed with a freshly prepared solution of ␣-cyano-4-hydroxycinnamic acid (10 mg/ml) in 50% (v/v) acetonitrile-0.3% trifluoroacetic acid and 1 l deposited on the sample plate. Alternatively, a sample corresponding to 0.3 g of each propeptide was directly injected unto a Zorbax SB-C 18 column (0.3 ϫ 250 mm, Phenomenex, Torrance, CA) connected to a -Liquid chromatograph coupled to a QSTAR-XL hybrid LC/MS/MS Mass spectrometer (Applied Biosystems, Foster City, CA). The data generated were analyzed with the Analyst TM -QS V1.1 software (Applied Biosystems/MDS-Sciex).
Enzymatic Assays and Kinetic Analysis-All enzymatic assays of recombinant mPC1/3 and human furin were performed using initial rate determinations at room temperature on a SpectraMax Gemini EM spectrofluorometer (Molecular Devices, Sunnyvale, CA). The assays were done in a final volume of 100 l in black 96-well flat bottomed plates (Corning Life Sciences, Acton, MA) using 100 M of the fluorogenic substrate pGlu-Arg-Thr-Lys-Arg-MCA (Peptides International, Louisville, KY,USA). For mPC1/3, the buffer consisted of 100 mM sodium acetate at pH 6.0 containing 10 mM CaCl 2 . Prior to use, the purified recombinant enzyme was incubated in the presence of Ca 2ϩ for ϳ6 h or until the release of AMC was determined as linear, to allow conversion into the fully active 71-kDa form. For human furin, the final assay conditions were 100 mM Tris-HCl buffer, pH 7.0, with 1 mM CaCl 2 . The fluorescence of the released 7-amino-4-methylcoumarin (AMC) was measured using an excitation and an emission wavelength of 370 and 460 nm, respectively. All the assays were started by the addition of the enzyme (corresponding to an activity measured as 15 M AMC-released/h), and the data points were collected every 30 s. The evaluation of the various inhibition parameters was done as previously described (24). Briefly, the progress curves obtained for the inhibition of mPC1/3 by its propeptide has been shown to follow a tight binding character and, hence, can be defined by the equation (25), where P is the product formed (AMC released), v i is the initial velocity, v s is the steady-state velocity, t is the time, and k is the apparent rate constant for inhibition. Progress curves were submitted to non-linear regression curve analysis using the software Grafit 4 (Erithacus Software, Horley, Surrey, UK), which allows the determination of the individual parameters v i , v s , and k for each curve. The obtained k values were then plotted against inhibitor concentrations. However, competitive tight binding inhibitors can display two different behaviors. The first one corresponds to a single-step process (Mechanism A) whereby the enzyme and the inhibitor combine to form a stable complex. In this case the plot k versus [I] is linear and best described by Reaction 1, where K ϭ k 2 /k 1 . The second one (Mechanism B) is a two-step process whereby the enzyme⅐inhibitor complex (E⅐I) undergoes a conformational change. In this case, the plot k versus [I] saturates with increased amounts of inhibitor deviating from a straight line, and the K i value can be determined following non-linear regression analysis through curvefitting using Reaction 2.
In the case of classic competitive inhibition, the various kinetic parameters are evaluated using the Enzyme Kinetic V1.0 module (SigmaPlot 2000 for Windows V6.1, SPSS Inc., Chicago, IL). In most cases, the GTTCAGCCCACGTCACACGAGCATCATCAGATAACCTC a S denotes the sense strand, whereas A denotes the antisense strand.
computed results were in close agreement with the equations exhibiting overall fit exceeding 0.990.

Propeptide Iodination and Cleavage by Recombinant mPC1/3 and Human Furin-
The purified mutant and WT propeptides were chemically labeled with radioactive iodine. For this purpose, 2.5 g of each propeptide was dried then resuspended in 0.05 M sodium phosphate buffer, pH 7.4. After the addition of 250 Ci of [Na 125 I] (Amersham Biosciences), the reaction was started by the addition of 50 g of chloramine-T in phosphate buffer. The reaction was stopped with 100 g of sodium metabisulfite. The volume of the reaction was made up to 1 ml with 0.1% trifluoroacetic acid (v/v) in water, and the sample passed trough a Sep-Pak C 18 cartridge (Millipore, Billerica, MA) as described in the manufacturer's protocol. The iodinated peptide was recovered by elution with 60% acetonitrile (v/v) in 0.1% trifluoroacetic acid-water, and the radioactivity present in the elution fractions was determined using an automatic Gamma Counter (LKB-Wallac model 1272). The cleavage reaction was carried out in a total volume of 100 l containing 2.5 ϫ 10 5 cpm of radiolabeled propeptide (ϳ2 nM, based on protein content), the sodium acetate buffer specific for each enzyme as above described and the respective enzyme preparation (50 M AMC-released/h). After a 30-min incubation period, the reaction was stopped with 10 l of glacial acetic acid. The sample was dried, reconstituted, and boiled in Laemmli buffer, and its content was analyzed by electrophoresis on a 15% SDS-polyacrylamide gel. The separated peptides were subsequently electro-transferred overnight unto an Immobilon-P membrane (Millipore). The radioactivity on the membrane was measured using a Storm model 860 Imaging system (Amersham Biosciences) with PhosphorImager capability and using the ImageQuaNT TL software.
Model Building and Modeling of mPC1/3, Human Furin, and the Various mPC1/3 Propeptides-The model of the catalytic domain of mPC1/3 was built upon the atomic coordinates obtained from the crystal structure of the proprotein convertase furin (26) available using the accession code 1P8J in the Brookhaven Protein Data Bank (PDB). Based on the extensive amino acid sequence alignment previously reported (27,28), the furin sequence was mutated into the mPC1/3 sequence one amino acid at a time. The backbone dihedral angles and the side chains of each amino acid were adjusted until an acceptable low energy conformation was obtained. Similarly, each mutant propeptide was modeled using the atomic coordinates derived from the NMR solution structure of the mPC1/3 (29,30) also deposited in the PDB under the accession number 1KN6. The variation of Gibbs free energy was computed following 1000 steps of structure minimization. All calculations were carried out using SYBYL version 6.91 software (Tripos Associates, St. Louis, MO) on an IBM-PC platform as described previously (31).

RESULTS
Production of Recombinant Propeptides-Because the major caveat in inhibiting convertases by their propeptides points to the lack of specificity due to the redundancy of the inhibitory COOH-terminal region, we tried to identify some other residues that might confer increasing specificity without modification of the inhibition properties. Furthermore, we and others have previously shown that synthetic peptides of various lengths did not exhibit any significant selectivity (14,17,19) nor did the isolated propeptides when assayed against a variety of convertases (19). However, single amino acid substitutions were shown to exhibit a profound effect on both inhibition and activation of convertases. Indeed, single amino acid substitution in the furin propeptide rendered the activation process inoperative and thus yielded no enzymatically active furin (6). However, the propeptide being presented in a cis fashion, i.e. being part of the zymogen, this study could not address the specificity aspect between convertases. The sequence alignment of the convertases propeptides shows very weak sequence similarities between the seven members of the family, except at the extreme COOH-terminal portion (Fig. 1A). Nevertheless, the similarity of certain key residues in the propeptide of mPC1/3 and prosubtilisin, the secondary structure might be conserved. Analysis of the secondary structures of the propeptides of PC by CD measurements appears to agree with such an assumption (19). We therefore speculated that certain structural features of the propeptide could be modified as to confer increased specificity. To test our hypothesis, we carried out site-directed mutagenesis of 10 individual residues located in two convertase potential cleavable sites ( 50 RRSRR 54 and 61 KR 62 ) and in an acidic region ( 65 DDD 67 ) uniquely conserved among species thus disrupting the ␣2 helix (Fig. 1B). Alanine residues, chosen because of their least structurally disrupting effect (32), replaced these residues and the resultant mutant propeptides expressed in E. coli BL21 cells, because the propeptide structure does not reveal any potential sites of glycosylation or sulfation. In addition, as mentioned under "Materials and Methods," our propeptide was elongated at both the NH 2 and the COOH termini to help in cloning and purification. The added sequence had no observable detrimental effect in inhibition properties. The propeptides were first purified from bacterial extracts using classical His-affinity chromatography and then further purified by HPLC and immunoreactive fractions were pooled together (representative data are shown in Fig. 2). Each propeptide was analyzed by mass spectrometry, and its molecular mass was found to be within 1 Da of the computed mass: for example, the mass of WT propeptide is 12,733.34 (average). This expression system allows us to obtain ϳ5-10 mg of each purified propeptide mutant per liter of culture.
Effect of the Propeptide Mutations upon mPC1/3 and Human Furin Activity-We assessed the inhibitory potency and selectivity of the generated mutants through processing of a small fluorogenic substrate pERTKR-MCA by recombinant mPC1/3 and human furin. We determined the IC 50 value of the inhibition of the WT propeptide against mPC1/3 and human furin as being ϳ20 nM (data not shown). Using this concentration, each mutant, numbered M1 to M10, was assayed against enzymatically active mPC1/3 (Fig. 3A) and human furin (Fig. 3B). The results indicate that, irrespective of the enzymes, the generated mutants belong to three categories namely, mutants exhibiting lower, higher, or equal inhibitory potency to the WT propeptide (Fig. 3).
In the case of mPC1/3, it is readily seen that M1, M2, M5, and M10 are better inhibitors than the WT (Fig. 3A). Interestingly, the first three mutants correspond to the substitution of arginine residues in the RRSRR sequence and hence were expected to play a role in the interaction with the enzyme. On the other hand, the last one contains a mutation of an aspartate residue at position 67 not likely to interact either with the active site or with the enzyme. The computed K i values (see below) also confirm this result, as they are 6 -30 times more potent than the WT (Table 2). Even more intriguing is the fact that the M9 mutation corresponding to an Ala for Asp at position 66 results in a significant diminution of inhibition contrary to an identical substitution at position 65 in M8. The mutation of the Ser 52 and Arg 53 (M3 and M4), present in the second potential cleavable site, to alanine also had opposite effects, in terms of inhibition, but much less than the other substitutions in this cluster. Finally the substitution at the other potential cleavable site Lys 61 and Arg 62 (M6 and M7) had very little effect on the inhibition of the active enzyme. It is noteworthy to mention that again, substitutions in the ␣2 helix where the three aspartate residues are located, led to the most significant changes in inhibition properties hinting that this region, while not likely to interact with the active site, does play an important role in the interaction of the propeptide with the enzyme (see "Discussion").
When we tested the same mutants on a baculovirus preparation of human furin, a similar but different scenario appeared. First, the mutations induce either an increase or a decrease in the inhibitory potency of the PC1/3 propeptide toward human furin (Fig. 3B). In marked contrast to mPC1/3, most of the mutants generated are more potent than the WT. Thus, all substitutions in the second cleavage site, namely mutants M1, M2, M3, M5, and to a much lesser extent, M4, proved beneficial in terms of human furin inhibition ( Table 2). It is worth noting in Fig. 3B that the inhibition afforded by M3, although competitive (see below), appears to diminish with time. Indeed, human furin appears to recover enzymatic activity hinting that M3 is rapidly cleaved by the enzyme thus neutralizing its inhibition. Substitution of the two amino acids occupying the third putative cleavage site led to mixed results, as M6 (K61A) appeared to have minimal effect whereas M7 (R62A) yielded a much improved inhibitor. Finally, similarly to what we observed with mPC1/3, mutations of the three aspartate residues led to significant changes in inhibition profile. Indeed, substitution at position 66 (M9), diminishes significantly the inhibitory potency of this mutant. On the other hand, the mutants M8 and M10 are potent inhibitors of human furin. Most interestingly and irrespective of their mode of inhibition (see below), two inhibitors (M7 and M8) revealed a much improved selectivity as, solely in terms of K i , they are 40 times more potent toward human furin than mPC1/3. On the other hand, M10, also a very potent inhibitor of either enzyme, displays no selectivity whatsoever as it inhibits human furin and mPC1/3 with the same K i .
In an attempt to relate the effect of the mutation upon the structure of each propeptide, we decided to model each mutant on the solution structure of pro-PC1/3 as previously determined (29,30). This was accomplished by computing the change in free energy resulting from each mutation. As presented in Table 2, no clear correlation is readily apparent. However, those mutants found to have the strongest effect (as seen by an increase in free energy) are M7, M8, and M9, mutants that displayed significant modifications in inhibition properties as compared with the WT. Interestingly, these mutants correspond to substitution at amino acids not residing in the major cleavage sites nor thought to interact with the active site.
Mechanism of Inhibition of PC1/3 Propeptide Mutants toward mPC1/3 and Human Furin-Synthetic inhibitory peptides against convertases, engineered from the COOH-terminal part, presented a competitive pattern of inhibition supporting the fact that this portion of the propeptide interacts directly with the active site (14,19). Upon elongating these peptides toward the NH 2 terminus, a change in inhibitory behavior was observed, because the initial competitive inhibition changed with the length of the peptide to mixed-inhibition and even FIGURE 1. Alignment of mammalian convertase propeptide sequences and PC1/3 mutagenesis sites. A, the amino acid sequence alignment of the seven mammalian convertase propeptides was performed using the ClustalW program. Residues in black boxes are conserved, whereas those in gray boxes represent common or chemically similar amino acids. Gaps were introduced to maximize the degree of similarity. B, alanine scan was performed on two potential convertase cleavable sites and a region (␣ 2 helix) encompassing highly conserved residues amongst PC1/3 of various species. The PC1/3 propeptide sequence is shown together with its secondary structure elements identified in the NMR solution structure (29). The various mutants used in this study and the putative sites of cleavage are indicated.
non-competitive inhibition (14). In the case of the propeptide of the closely related Kex2p enzyme, the propeptide behaved as a mixed inhibitor (33). When we tested the complete propeptides against various convertases, we and others found that most of them behaved as tight binding inhibitors of their cognate enzyme (7,13). Overall, these results indicate that sites other than the extreme COOH terminus on the propeptide play important roles in regulating enzyme activity by interacting with the enzyme outside of the active site cleft.
To better document the nature of the inhibition mechanism, we performed on-line assays using fixed concentrations of enzyme and substrates but varying amounts of each mutant. Mutants M3 and M9 displayed inhibition curves typical of purely competitive behavior (Fig. 4A), confirmed by further analysis through linear regression of data plotted using classic representation (data not shown). On the other hand, all the other mutants exhibited inhibition curves more characteristic of slow tight binding inhibition and identical to the curve observed with the WT-propeptide (13). However, slow tight binding inhibition can fit a one-or a two-step mechanism depending upon the pathway used to form the stable inhibitory complex (24). In the single-step process herein referred to as mechanism A, the inhibitor binds tightly to the enzyme without inducing any conformational change. By contrast, mechanism B is best explained by first initial tight binding to the enzyme (step 1) followed by a conformational change (step 2) leading to enhancement of the stability of enzyme⅐inhibitor complex. To differentiate between these two mechanisms, it is best to measure observed K obs at different inhibitor concentrations. Using the progress curves shown in Fig. 4 (B and D), a plot of K obs as a function of inhibitor concentrations yields a straight line in the case of a single-step inhibition mechanism (Fig. 4C) or an hyperbolic curve for two-step inhibition (Fig. 4E). Illustrated for the mutant M10, similar curves are obtained for the other mutants with the exception of mutant M2 whose inhibition followed the two-step mechanism. As indicated in Table 2, the computed K i values ranged from 150 pM up to 16.5 nM when assayed against mPC1/3. The computed K i for the WT-propeptide (4.4 nM) was in good agreement with the one measured originally (6 nM) (13).
In the present study, the slow tight binding kinetic though still present was much less apparent though the tight binding characteristic of the various propeptides with the exception of M9 was clearly observed. A K i value for inhibition of furin by WT-PC1/3 propeptide was computed as 1.1 Ϯ 0.3 nM, a value well in agreement with the value previously reported of 1.6 nM (19). Indeed as shown, in Fig. 5A for the mutant M3, the velocity does not decrease in linear fashion with the inhibitor concentration as would be expected if the enzymatic activity is nullified by the formation of an inactive inhibitor⅐enzyme complex. This behav- ior is not observed in the case of M9, because the decrease in velocity was related to the inhibitor concentration in linear fashion. In the case of M9, further analysis demonstrated its full competitive inhibition (Fig.  5B) yielding a K i of 24 nM ( Table 2). The K i values of the other inhibitors were computed through non-linear regression according to a tight binding model (25). It is important to note that the M3 mutant (S52A) exhibits a competitive inhibition profile toward PC1/3 but is a tight binding inhibitor of human furin and is the only one to do so.
Cleavage of Propeptide Mutants by mPC1/3 and Human Furin-It is well established that the propeptides of convertases, just like the ones of the yeast Kex2p (34) and the related subtilisins (10,35,36), are cleaved at the primary cleavage site through an intramolecular process. However, to release the full enzymatic activity and prevent further inhibition by the bound propeptide, further internal cleavages are necessary if not mandatory. Whereas this internal cleavage appears essential for furin (11) and PC1/3 (13), internal cleavage of the PC2 propeptide is not obligatory for full activation of PC2 (7). To assess whether the enzyme is still capable of cleaving the mutant propeptides and whether this cleavage can be related to inhibitor potency, we decided to incubate the different mutants with an excess of PC1/3 or furin. To follow the extent of cleavage, the propeptide mutants were labeled with 125 I, because three tyrosine residues are conveniently present at positions 28, 41, and 76 within the propeptide sequence.
Our results confirm that native WT propeptide is cleaved by PC1/3 generating a lower molecular weight band with an apparent mass of ϳ7.5 kDa (the computed mass of fragment Ϫ13 to 54 is 7481 Da) corresponding to cleavage at the internal site originally reported (13). However, as seen in Fig. 6A, all mutations of the arginine residues occupying positions 50 (M1), 51(M2), 53 (M4), or 54 (M5) proved detrimental to efficient cleavage in that region. Considering that the preferred cleavage site in substrates of PC1/3 is recognized as being one whereby arginine residues are occupying (just as for furin) position P 1 and P 4 , 5 mutation at any position within that sequence would be expected to perturb the cleavage. This confirms our prior identification of the identity of the PC1/3 propeptide internal cleavage site as being 50 RRSRR 54 in agreement with the convertase recognition motif RXXR. Furthermore, it seems that the mutations found in M2 and M5 have a more pronounced effect on inhibition than M1 and M4 hinting that the former might be the favored motif. On the other hand, cleavage at the other potential site, namely 61 KR 62 , could not be detected contrary to what has been observed with a similarly located pair of basic residues in the mPC2 propeptide sequence cleaved by mPC2 (7).
The PC1/3 WT-propeptide is cleaved by human furin in an identical fashion to what we observed with PC1/3 yielding a single band of apparent molecular mass of 7.5 kDa. However, in certain samples, we observed an additional band exhibiting an apparent molecular mass of 8.5 kDa as indicated in Fig. 6B. Interestingly, mutations of the Arg 51 as well as the Arg 54 yields exclusively the band of 8.5 kDa indicating that furin is not able to cleave the identical motif formed by Arg 50 and Arg 53 . In this situation, human furin prefers to cleave at the other possible site namely the 61 KR 62 yielding a fragment of computed mass of 8387 Da (Ϫ13 to 62). Indeed, whenever the sequence 51 RXXR 54 is present, one can see the appearance of this band though in the case of the WT and M3 it is faint. As mentioned previously in conjunction with Fig. 3B, M3 though a potent inhibitor appears to be very rapidly cleaved possibly explaining the absence of the 8.5-kDa band. More surprising, however, is the observation that mutating either Lys 61 or Arg 62 fails to completely abolish cleavage at that site, although it can be seen clearly impaired. The latter mutation, if it does not strongly influence cleavage, does lead to a 2.6-fold increase in K i when compared with M6 and more importantly to a considerable increase in specificity over PC1/3.

DISCUSSION
The activation of subtilisin and subtilisin-like serine proteases is a multistep process requiring that the propeptides play a dual role. First, as an intramolecular chaperone, it is assisting the folding of the catalytic domain. This feat is accomplished by lowering the transition state energy allowing the conversion of a collapsed metastable intermediate to a native enzyme (38). The importance of this role was first revealed for prosubtilisin and pro-␣-lytic protease (reviewed in Ref. 10), and further results led to proposal of the concept of "protein memory" (39). In this context, an identical protein sequence can give rise to different conformations through the folding with a mutated chaperone. Application of this concept through propeptide engineering led to the production of new proteases exhibiting altered stability, substrate specificity, and activity (40). Second, as an inhibitor of its cognate enzyme, the propeptide prevents undue activation both in terms of location and time in such a way as to allow full enzymatic activity until the active enzyme is needed and properly located. However, it is worth noting that these two functions are not absolutely linked as in the case of subtilisin E (41). The latter study actually prompted us to initiate the present study.
Indeed, the introduction of point mutations in the peptidase-propeptide interface seriously compromised the propeptide inhibition potential when added in trans without affecting its folding capacity of the propeptide. Hence, using the pro-PC1/3 as a model, we wanted to know a This value was obtained following energy minimization as described under "Materials and Methods." b The site 2 is hereby defined as corresponding to the 50 RRSRR 54 , whereas the site 3 is defined as being the 61 KR 62 pair. c These K i values were obtained using a fully competitive model, whereas the others for PC1/3 were based upon a slow tight binding model and those for human furin were based upon a tight-binding model. whether by substituting certain amino acids, we could modify the inhibitory potency, the inhibitory mechanism, and the selectivity of the propeptide. This objective also resulted from the previously reported difficulties in developing synthetic peptides of various sizes based upon the propeptide sequences, which could prove potent and selective. Thus, small peptides mimicking the COOH-terminal portion of the propeptide were synthesized, and their properties were assayed. Even though these peptides could be made into very potent inhibitors in the nanomolar range, they are mostly competitive and nonselective, which restrict their use to structural characterization of the enzymes. Such peptides have been synthesized and assayed with numerous PCs, including furin, PC1/3, PC5/6, and PC7 (14, 19, 42). As mentioned above, there exists in the propeptide sequence a second potential cleavable site known to be important for the activation and the secretion of PC5/6, PC2, and furin. In PC1/3, mutations in the second cleavable region ( 51 RSRR 54 to 51 SSGR 54 ) had no effect on PC1/3 processing in Xenopus egg system (43). When synthetic peptides derived from the second cleavable site were tested against PC1/3, PC7, and furin, they were found to be either non-inhibitory or generally weak inhibitors, exhibiting K i in the micromolar range. In any case, no synthetic peptide was able to duplicate mechanistically the inhibition generated by the complete propeptide when presented in trans. Indeed, the inhibition afforded by the propeptide leads to the formation of a stable complex as previously shown with various convertases. This is explained by important interactions between the propeptide and sites close or remote of the convertase active site as visualized in the crystal structure of the propeptide of subtilisin BPNЈ, E, and ␣-lytic protease in complex with their cognate enzyme (44 -46). Similar though distinct contacts were also inferred in the case of the convertase propeptides with their cognate enzymes following the determination of the NMR solution structure of pro-PC1/3 (30). Nevertheless, the entire propeptides, independent of the way they are presented to the enzyme be it in cis or in trans, lack selectivity, because many convertase propeptides inhibit their own protease and other related convertases with variable potency. Conversely, swapping of the propeptide has been shown, in certain cases (for example, PC1/3 and furin), to yield enzymatically active proteinases (5). Hence, it can be proposed that selectivity might not be derived from the basic architecture of the propeptide, which is reported to be composed of fourstranded antiparallel ␤-sheets and two ␣-helices, but more so by the localized interactions of certain residues at the interface of the propeptide⅐enzyme complex. Thus we tried herein by site-directed mutagenesis to explore the possibility of modulating the extent of the interactions between the propeptide and its enzyme. To do so, ten mutants were generated, and their interactions with two enzymatically active convertases were determined.
In summary, this study demonstrates that modifying a single amino acid within the structure of propeptide can have profound effect on its reactivity. Indeed, when compared with the native propeptide (WT), single residue mutation yielded propeptides (i) with different inhibition mechanism depending on the enzyme assayed (M3); (ii) that are weaker and displaying different mechanisms with both enzymes (M9); (iii) with identical inhibition mechanism but with enhanced selectivity (M7, M8); and (iv) that are very potent but not at all selective (M1, M2, M5, and M10).
The mutation of the arginine Arg 50 (M1), Arg 51 (M2), or Arg 54 (M5) in the RRSRR sequence all led to very potent inhibitors with K i in the low picomolar range for both enzymes. This validates the previous assumption that this sequence represents the site of the secondary cleavage of the propeptide for PC1/3. This site contains two possible RXXR motifs, 50 RRSR 53 and 51 RSRR 54 . As illustrated in Fig. 6, mutations at positions 50, 51, 53, and 54 severely impair the cleavage by PC1/3 hinting that PC1/3 can use either of the recognition sites. On the other hand, furin appears more selective as only the mutations at position 51 or 54 have an effect on cleavage. This obviously confirms the reported strong preference for substrates having both P 1 and P 4 residues as basic amino acids. However, in the context of the PC1/3 propeptide sequence, it thus seems that furin prefers much more the 51 RSRR 54 site than the 50 RRSR 53 even though both display the minimal furin recognition motif namely RXXR. Interestingly, the crystal structure of furin revealed that, in contrast to the yeast convertase Kex2, the former does not exhibit the strong requirement of the latter for a basic amino acid at the P 2 position (27,47,48). Nevertheless, our results using either enzyme favor a sequence containing basic residues at positions P 1 , P 2 , and P 4 . This observation is also consistent with our previous results using Barley serine proteinase inhibitor 2-derived cyclic peptides. In that study, we noticed an additive favorable effect in terms of substrate recognition when basic residues occupy positions P 2 and P 4 (31). The mutation of these residues, especially the arginine occupying position 54, severely impairs the cleavage of the propeptide and thus could contribute to an increase in inhibitor potency. This observation is especially true with PC1/3, because M5 exhibits a K i of 0.2 nM, and this could reasonably be explained by the fact that the enzyme, due to lack of cleavage, cannot dissociate from the propeptide. In the case of furin, the consequence of mutations in this region is not as significant, and this might be due to the presence of another furin-sensitive site in the propeptide (see below). Finally, the M1 mutant (Arg 50 ) if one considers the Arg 54 as the P 1 position corresponds to a P 5 substitution and is a very potent inhibitor of both enzymes, but only furin can cleave it efficiently. The S 5 site in furin offers a negatively charged environment (Glu 233 /Asp 235 ) in complete agreement with its preference for positively charged P 5 side chain; substitution by a small neutral residue such as alanine thus appears not detrimental. On the other hand, substitutions by acidic residues are much less tolerated at that position (49,50). The inability of PC1/3 to cleave this mutant is difficult to rationalize, because the S 5 site in PC1/3 is less negatively charged than found otherwise in other convertases such as furin, PACE4, PC5/6, and PC7. Hence, substituting the arginine at this position by an alanine should not be expected to prevent the cleavage by PC1/3 unless the arginine side chain plays an as yet undetermined role in the propeptide structure or in the propeptide⅐enzyme complex. However, it could also be that, in this particular instance, the arginine residue is essential for the interaction with Glu 251 and with Asn 244 , both residues being present in PC1/3 S 5 . In the solution structure of PC1/3 propeptide this residue resides in a solvent-accessible loop located between the third ␤-sheet segment and the second ␣-helix (30). Another plausible explanation might be that by introducing an alanine residue, one decreases the high concentration of positively charged residues in the loop and in turn, this prevents complete dissociation of the enzyme⅐inhibitor complex.
The mutation of the Ser 52 (M3) rendered the propeptide purely competitive toward PC1/3, whereas it did not change inhibitor's behavior against furin. This observation is interesting, because, in view of possible therapeutic applications, a tight binding inhibitor forming a stable complex is largely preferred to the high concentration of a competitive inhibitor required to abolish the activity. Essentially, this serine residue appears to play an important role in stabilizing or promoting the enzyme⅐propeptide complex. The replacement of this polar residue by an alanine could abolish some important weak interactions. Both enzymes cleave this mutant efficiently, and this is in agreement with biochemical and structural data reported for furin and PC1/3. Actually, the substrate-binding site S 3 in furin is highly permissive, because the x-ray structure does not reveal a discrete binding pocket (47). However, we have shown that one way to affect substrate recognition and cleavage by furin is to introduce at P 3 an acidic residue such as Glu (49). Based on the now available structure, this is explained by unfavorable contact with Glu 233 , which, though occupying the S 5 cleft, is surface-located and can possibly interact with a P3 residue (48). In PC1/3, a neutral Asn residue replaces this Glu residue; hence, it is not surprising that the only substitution reported to compromise PC1/3 substrate cleavage efficiency at this position is a Pro residue (15,49).
The basic amino acid pair, 61 KR 62 , a possible convertase-cleavable site present in PC1/3 propeptide and containing P 1 and P 2 basic motifs, is not cleaved by PC1/3. However, substituting either one or the other leads to the propeptide being among the weakest in this series toward PC1/3. Being located in the second ␣-helix segment, their replacement by a small neutral residue could potentially have an impact on the integrity of this helical segment. As previously mentioned, a similarly located pair of basic residues in the pro-mPC2 sequence is recognized and cleaved by mPC2 (7). This pair of basic residues is also recognized as a substrate by furin in the context of the pro-PC1/3 sequence. Interestingly, in the profurin sequence, a typical furin recognition sequence, namely RHSR, is similarly located. Moreover, the substitution of the Arg 62 residue leads to an increase in propeptide inhibitory force against furin suggesting at first sight that this residue is interacting elsewhere with the enzyme to emphasize the contact. Alternatively, the degradation of the propeptide by furin requires the cleavage at that site, in addition to the second cleavage. The ability of furin to cleave M7 (R62A) at two positions, namely at the RSRR and at the KA sites, complicates the interpretation. Indeed, one is left wondering whether furin itself can cleave on the COOH-terminal to a Lys residue. Obviously, looking at the results of the cleavage experiment of mutants M6 and M7, furin would be able to cleave after both Lys 61 and Arg 62 , because either mutant yields two lower molecular weight peptides. Interestingly, as observed with M2 and M5 mutation that removes either of the arginines occupying P 4 or P 1 , furin can utilize efficiently this basic pair to cleave the PC1/3 propeptide. Nevertheless, furin cleavage at a pair of basic residues flanked by a P 4 hydrophobic residue instead of the usual basic FIGURE 6. Cleavage of the propeptide mutants by mPC1/3 and human furin. A and B, each mutant as well as the native bacterially produced propeptide were iodinated and an aliquot (2.5 ϫ 10 5 cpm) incubated for 30 min with enzymatically active mPC1/3 (A) and human furin (B) (see "Material and Methods"). Following acidification of the incubation mixture, the peptides in the digest were separated using a 15% SDS-PAGE transferred to an Immobilon P membrane, and the radioactivity was detected through autoradiography.
residue as well as cleavage on the COOH-terminal side of Lys represents two uncommon and rarely seen types of cleavage. Indeed, concerning the former, we and others have shown that hydrophobic interactions at the S 4 -P 4 interface can contribute positively to the efficiency of cleavage by furin (49,51). Furthermore, such pair of basic residues can be found in a number of natural substrates, including one in the furin propeptide itself at the second cleavage site (11), although in these instances, it is not followed at P 1 Ј by an hydrophobic aliphatic residue, a feature less favorable to furin cleavage (52,53). Concerning the latter, namely the cleavage COOH-terminal to a Lys residue, the M7 cleavage pattern indicates that it is certainly not a favored one, although it is nonetheless happening. Indeed, there are very few examples of hormones or neuropeptides being processed at a Lys residue (for example, progastrin, proenkephalin, pro-thyrotropin releasing hormone, and prosomatostatin) and even less shown conclusively to be cleaved by furin (54,55). Actually, synthetic peptides mimicking the cleavage site of proalbumin were used to demonstrate that introduction of a P 1 Lys residue abolishes processing (52). On the other hand, other studies have indicated that, despite a diminished efficiency, furin can adequately perform such cleavage, the unfavorable replacement of P 1 Arg by Lys likely compensated by surrounding residues (19,48,51,56). Finally, though the KR sequence is not recognized by PC1/3 in terms of internal cleavage site, mutating of Arg 62 decreases significantly its inhibitory potency rendering M7 the worst inhibitor of PC1/3 in the series. On the other hand, for furin, M7 ranks among the best inhibitors.
Individual mutation of the three Asp residues located in the ␣2 helix exhibited significant and unique properties. A clear relationship between the levels of perturbation introduced in the propeptide structure by the various mutations was not established with the inhibitory potency (see Table 2). However, the three weakest inhibitors of PC1/3 correspond to M7, M8, and M9, and these mutations appear to be the most perturbing in terms of conformational stability. It is worth mentioning that these mutations were introduced at sites remote from the primary reactive site and unlikely to play a role in the formation of the enzyme⅐inhibitor complex. Similarly to the parallel drawn between the M2 and M5 mutants in terms of cleavage by PC1/3 or furin, a similar one can be identified here between M7 and M9, the latter exhibiting a more profound effect, because M9 is not able to form a tight complex with PC1/3. As shown in Fig. 7, examination of the pro-PC1/3 solution structure reveals that the Arg 62 (mutated in M7) and the Asp 66 (mutated in M9) are not only deeply buried within the structure but more importantly can interact through hydrogen bonding. Indeed, it can be seen that the Asp backbone atoms as well as those in the side chain can participate in six hydrogen bonds. Hence, mutating the Arg 62 leads to the loss of two of these bonds, but the Asp residue is still able to link with the Arg 68 . However, mutating this Asp will remove all those interactions except those arising from the main-chain atom, a condition likely to lead to severe conformational perturbations. Interestingly, whereas this modification is detrimental in terms of inhibiting pro-PC1/3 (mechanistically as well as in potency), the M7 mutant is a better inhibitor of human furin and exhibiting an increased selectivity. Clearly, even buried deeply in the structure and hence unlikely to actively participate in the molecular interactions between the enzyme and the propeptide, the Asp 66 appears very important in terms of the global folding of the propeptide. The furin propeptide possesses the equivalent Arg 62 and Asp 66 but not the Arg 68 , and it would thus be expected considering that PC propeptides are proposed to share similar structure (19,30) that mutations in furin propeptide at these positions could have significant effects. Considering that the side chain of Asp 66 is pointing inwards and is involved in hydrogen bonding, it can be safely assumed that it serves as an anchor point for this short helix, and hence the two other Asp are likely to point outwards and be solvent-accessible. Mutating these two residues is thus likely to perturb locally the ␣-helix but more importantly to decrease the negative surface potential. Interestingly, this does not have the same effect on PC1/3 and on furin, because M10 is the strongest inhibitor of either enzyme, whereas M8 is a much stronger inhibitor of human furin than of PC1/3.
In conclusion, this study highlighted the fact that, using the PC1/3 propeptide as a model structure, one can introduce small changes in the primary sequence that will affect its global structure, its sensitivity to proteolysis, and its inhibitor potency as well as mechanism.