Functional Analysis of the Cucumisin Propeptide as a Potent Inhibitor of Its Mature Enzyme*

Cucumisin is a subtilisin-like serine protease (subtilase) that is found in the juice of melon fruits (Cucumis melo L.). It is synthesized as a preproprotein consisting of a signal peptide, NH2-terminal propeptide, and 67-kDa protease domain. We investigated the role of this propeptide (88 residues) in the cucumisin precursor. Complementary DNAs encoding the propeptides of cucumisin, two other plant subtilases (Arabidopsis ARA12 and rice RSP1), and bacterial subtilisin E were expressed in Escherichia coli independently of their mature enzymes. The cucumisin propeptide strongly inhibited cucumisin in a competitive manner with a Ki value of 6.2 ± 0.55 nm. Interestingly, cucumisin was also strongly inhibited by ARA12 and RSP1 propeptides but not by the subtilisin E propeptide. In contrast, the propeptides of cucumisin, ARA12, and RSP1 did not inhibit subtilisin. Deletion analysis clearly showed that two hydrophobic regions, Asn32–Met38 and Gly97–Leu103, in the cucumisin propeptide were important for its inhibitory activity. Site-directed mutagenesis also confirmed the role of a Val36-centerd hydrophobic cluster within the Asn32–Met38 region in cucumisin inhibition. Circular dichroism spectroscopy revealed that the cucumisin propeptide had a secondary structure without a cognate protease domain and that the thermal unfolding of the propeptide at 90 °C was only partial and reversible. A tripeptide, Ile35-Val36-Tyr37, in the Asn32–Met38 region was thought to contribute toward the formation of a proper secondary structure necessary for cucumisin inhibition. This is the first report on the function and structural information of the propeptide of a plant serine protease.

Proteases play key roles in diverse processes regulating plant growth, development, and responses to environmental stimuli. They are necessary for protein turnover, strict protein quality control, and degrading specific sets of proteins. Comparative genomics analyses could provide valuable insights into the abundance and roles of various plant protease families. For example, the Arabidopsis thaliana genome has over 550 prote-ase sequences corresponding to almost 3% of the proteome, representing all five catalytic types: serine, cysteine, aspartic acid, metallo, and threonine (1,2). Of these, serine proteases appear to be the largest class of plant proteases, although protease activity has been demonstrated only by a few of them.
Cucumisin (EC 3.4.21.25) is an extracellular thermostable alkaline serine protease that is expressed at high levels in melon fruits (Cucumis melo L.). It comprises more than 10% of the total juice protein and is synthesized in the central parts of the fruits (3). Cucumisin is synthesized and accumulated only in melon fruits, and a cis-regulatory enhancer element in the cucumisin promoter regulates fruit-specific expression of the cucumisin gene (4). We have determined the complete nucleotide sequence of a cucumisin cDNA, the first sequenced plant serine protease, and found that cucumisin is a member of the subtilisin (EC 3.4.21.62) superfamily characterized by a catalytic triad of three amino acids: Asp, His, and Ser (5). The primary structure of cucumisin deduced from the cDNA sequence revealed that it is synthesized as a precursor consisting of four functional domains: a possible signal peptide (22 amino acid residues); NH 2 -terminal prosequence (88 residues); 54-kDa protease domain (505 residues), which is the active enzyme domain of the 67-kDa native cucumisin; and 14-kDa COOHterminal polypeptide (116 residues), which arises by limited autolysis of the 67-kDa native cucumisin (3,5). The optimal pH and temperature of the caseinolytic activity of cucumisin were found to be 10.5 and around 70°C, respectively (3), and its substrate specificity was reported to be fairly broad (6).
Since the cloning of the cucumisin cDNA, many other plant cDNAs for subtilisin-like serine proteases (subtilases) have been cloned. Subtilases constitute the S8 family within the SB clan of serine proteases (7) and are subdivided into six families based on their sequence similarities. Most plant subtilases are grouped into the pyrolysin family, which is characterized by a large insertion between the stabilizing Asn and the reactive Ser and/or long COOH-terminal extensions (8). In Arabidopsis, 56 genes predicted to encode functional subtilases have been annotated (9). Plant subtilases are involved in many physiological processes, such as microsporogenesis, symbiosis, hypersensitive response, signal transduction and differentiation, senescence, and protein processing (for reviews, see Refs. 2 and 10). For instance, SDD1 and ALE1 are involved in stomatal development or cuticle formation and epidermal differentiation, respectively. AtSBT1.7 (also termed ARA12) is involved in the maturation of the seed coat (11), and AtSBT6.1 is implicated in stress-induced processing of a membrane-associated transcription factor, thus inducing the expression of stress response gene (12).
Despite the prevalence and importance of plant subtilases, information on their enzyme activities and structures is very limited. Recently the x-ray diffraction analysis of a tomato subtilase (SBT3) has been reported (13). The primary structures of cucumisin and other plant subtilases suggest that they are secretory enzymes synthesized as inactive preproproteins and targeted to the endoplasmic reticulum (ER) 2 by signal peptides. The NH 2 -terminal amino acid sequence of the mature enzyme, which was first analyzed for cucumisin, is conserved among most plant subtilases. Amino acids at positions ϩ1 and ϩ2 are both Thr, and those at positions ϩ3 and ϩ4 are Arg/His and Thr/Ser, respectively (14). This suggests a common mechanism for the propeptide processing of plant subtilase precursors. Detailed mechanisms of subsequent processing and activation of plant subtilase precursors are unknown except for a recent report demonstrating that the prodomain cleavage of a tomato subtilase 3 (SlSBT3) occurs autocatalytically and that the zymogen maturation is an intramolecular process (15).
For bacterial and mammalian subtilases, much work has focused on the subsequent processing of the zymogens and its relevance for enzyme maturation. The prodomains of bacterial subtilisins are autocatalytically cleaved at their junction with the catalytic domains. They remain non-covalently bound and act as specific inhibitors of proteolytic activity (16,17). Also, subtilisin propeptides can act as intramolecular chaperones assisting the correct folding of the mature enzyme (18,19). Prodomain function and processing have also been investigated in detail for kexin-like mammalian proprotein convertases. For instance, the cleavage of the prodomain of furin at its junction with the catalytic domain occurs in a rapid intramolecular reaction in the ER, and this is necessary for the protein to fold into its native state (20). To date, however, no information is available on the roles of the prodomains of plant serine proteases, and the biochemical characterization of propeptides remains to be elucidated.
Here, we describe the strong inhibitory activity of the cucumisin propeptide against mature cucumisin and the rela-tionship between the secondary structure and the inhibitory activity of the cucumisin propeptide. This is the first report demonstrating that the propeptide of a plant serine protease acts as a tightly binding competitive inhibitor of the mature enzyme and that the secondary structure of the propeptide is indispensable for its inhibitory activity.

EXPERIMENTAL PROCEDURES
Reagents-Restriction and modification enzymes were obtained from New England Biolabs, Roche Applied Science, and Promega Corp. Glutaryl-L-alanyl-L-alanyl-L-prolyl-Lleucine p-nitroanilide (Glt-Ala-Ala-Pro-Leu-pNA) was purchased from Peptide Institute (Osaka, Japan). All other commonly available reagents were of analytical grade.
Subcloning of cDNAs for Propeptides of Several Subtilases and Expression of Recombinant Peptides in E. coli-General DNA manipulations were carried out using standard procedures (21). Complementary DNAs for cucumisin, ARA12 (termed AtSBT1.7 in A. thaliana subtilase code), and RSP1 were described in our previous studies (5,14). Subtilisin E cDNA was a gift from Dr. Hiroshi Takagi (22). Each cDNA was amplified by PCR using the cucumisin cDNA as a template and expressed in E. coli as His 6 -tagged proteins of the cucumisin propeptide, designated cuc-pro, and its short peptides, designated cuc-pro⌬N9, cuc-pro⌬N16, cuc-pro⌬C7, and cuc-pro⌬C14. The synthesized oligonucleotide primers are listed in Table 1. The primer sets used for PCR were as follows: P-1 and P-2 for cucpro, P-2 and P-3 for cuc-pro⌬N9, P-2 and P-4 for cuc-pro⌬N16, P-1 and P-5 for cuc-pro⌬C7, and P-1 and P-6 for cuc-pro⌬C14.
After digesting the PCR products with NheI and HindIII, the DNAs were subcloned into the corresponding restriction sites of pET28a (Merck) and introduced into E. coli Rosetta (DE3) (Merck). The nucleotide sequences of the resulting subclones were confirmed on both strands by sequencing using an automated sequencer (model 4000L, LI-COR Biosciences, Inc., Lincoln, NE). For the expression of wild-type cucumisin propeptide (cuc-pro-WT) that has no extra amino acids in the NH 2 terminus, such as His 6 tag, the nucleotide sequence was amplified using the primers P-7 and P-2 after which it was ligated into NcoI-HindIII sites of pET28a. For cDNA amplification of three propeptides, ARA12, RSP1, and subtilisin E, the primer sets used were P-8 and P-9, P-10 and P-11, and P-12 and P-13, respectively. Each PCR product was ligated into BamHI-Hin-dIII, NheI-HindIII, and NheI-HindIII sites in pET28a, respec- GGGGAAGCTTTCAGTGTAGCTCGTAA P-10 GGGGGCTAGCTCACGCAAGCTGTACATA P-11 GGGGAAGCTTTCACGCCGTCCTGTACCTC P-12 GGGGGCTAGCGCCGGAAAAGCAGTACA P- 13 GGGGAAGCTTCAATATTCATGTGCAATATG tively. To express recombinant proteins, transformed cells were cultured in LB medium containing 50 g/ml kanamycin at 37°C until an absorbance of 0.6 at 600 nm was reached. Recombinant proteins were induced by adding 1 mM isopropyl ␤-Dthiogalactopyranoside for 16 h at 37°C.

Site-directed Mutagenesis of Recombinant Cucumisin
Propeptide-Site-directed mutagenesis was used to introduce amino acid substitutions using a QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's protocol. Oligonucleotide primers used for the site-directed mutagenesis are listed in Table 2. All cDNA sequences used for mutated propeptides were verified by DNA sequencing.
Purification of Recombinant Propeptides-Purification of recombinant propeptides was performed at 4°C. Transformed cells were harvested by centrifugation at 8,000 ϫ g for 10 min, suspended in buffer A (50 mM sodium phosphate buffer, pH 7.5 containing 0.3 M NaCl and 5 mM ␤-mercaptoethanol), and homogenized with a supersonic wave using a UD-200 ultrasonic disruptor (TOMY Co. Ltd., Tokyo, Japan) with output 2 for a total of 5 min on ice. After centrifugation at 12,000 ϫ g for 30 min, the pellet was solubilized in buffer B (buffer A containing 8 M urea and 20 mM imidazole) with sonication. The solution was incubated for 16 h at 4°C. Insoluble debris were removed by centrifugation at 12,000 ϫ g for 30 min, and the supernatant was filtered through a 0.45-m nitrocellulose membrane filter (ADVANTEC Co. Ltd., Tokyo, Japan). The His 6 -tagged propeptides were purified using a nickel-Sepharose HP and HiTrap Chelating HP column (0.7-cm inner diameter ϫ 2.5 cm, GE Healthcare) equilibrated with buffer B. After washing the column with buffer B, the protein was eluted stepwise with every 5 ml of 50 -200 mM imidazole in the same buffer. The fractions containing recombinant proteins confirmed by 15% SDS-PAGE were pooled. The concentration of purified proteins was adjusted to about 50 g/ml with buffer B, and the solution was dialyzed against buffer C (50 mM sodium phosphate buffer, pH 7.5 containing 0.2 M NaCl) for 24 h. To purify the cucumisin propeptide without the His 6 tag (cuc-pro-WT), the propeptide was solubilized in buffer B and dialyzed against buffer D (50 mM sodium phosphate buffer, pH 7.5) for 16 h. After centrifugation at 12,000 ϫ g for 30 min, the supernatant was filtered through a 0.45-m nitrocellulose membrane filter and put on a DEAE-Sepharose column (0.5-cm inner diameter ϫ 5 cm) equilibrated with buffer D. After washing the column with buffer D, the protein was eluted stepwise with every 5 ml of 25-150 mM NaCl in the same buffer. The yields of wild-type and mutant propeptides were 30 and 70 -90%, respectively. The fractions containing recombinant proteins were pooled and used for further analysis.
Purification of Cucumisin-Prince melons (C. melo L. cv. Prince) were cultivated at the experimental farm attached to the Faculty of Agriculture, Kobe University from April to August. Fruits were tagged upon pollination, and developing fruits were harvested between 15 and 20 days after pollination. Purification of cucumisin was performed at 4°C as described previously (5) with slight modification. The central parts of the fruits were separated from the sarcocarp and washed with buffer E (50 mM sodium acetate buffer, pH 5.0 containing 0.3 M NaCl). This wash was combined with the juice from the central parts of the fruits and used as the crude extract. After centrifugation at 8,000 ϫ g for 15 min, solid ammonium sulfate was added to the supernatant to 60% saturation, and the precipitate was collected by centrifugation at 12,000 ϫ g for 20 min. The proteins were dissolved in a small volume of buffer E and put on a HiPrep 16/60 Sephacryl S-200 HR column (1.6-cm inner diameter ϫ 60 cm; GE Healthcare) equilibrated with the same buffer. The eluted fractions containing protease activity were pooled and precipitated with 60% saturated ammonium sulfate. The proteins were dissolved in a minimum volume of buffer E, and gel filtration using the HiPrep 16/60 Sephacryl S-200 HR column was repeated in the same manner. The protease fractions were collected by ammonium sulfate precipitation, dissolved in a small volume of buffer F (50 mM sodium acetate buffer, pH 5.0), and dialyzed against the same buffer for 16 h. The protein solution was put on a CM-Sepharose Fast Flow (GE Healthcare) column (1.6-cm inner diameter ϫ 10 cm) equilibrated with buffer F. After washing with buffer F, protease was eluted in the same buffer with a liner gradient of 0 -200 mM NaCl. The protease fractions that were eluted as a single peak were pooled and precipitated with ammonium sulfate. The pre-

Oligonucleotide sequence (5 to 3) Mutants
cipitate was dissolved in a small volume of buffer F and dialyzed against the same buffer for 16 h. The purified enzyme was confirmed to be homogeneous based on SDS-PAGE and staining of the gel with Coomassie Brilliant Blue R-250. Protein Measurement-Protein concentration was determined spectrophotometrically using molar absorption coefficient constants (E 280 ) computed by the ProtParam program (23). The concentrations of the mutated cucumisin propeptide were also determined using the Bio-Rad Protein Assay kit using wild-type cucumisin propeptide as a standard.
Assays for Protease and Protease Inhibitor and Kinetic Measurements-Cucumisin activity was assayed at pH 7.5 and 30°C using Glt-Ala-Ala-Pro-Leu-pNA as a substrate. The assay was started by adding 20 l of cucumisin in 10 mM sodium phosphate buffer, pH 7.5 to 220 l of substrate-propeptide mixture in assay buffer (150 mM sodium phosphate buffer, pH 7.5 containing 0.2 M NaCl). After incubating for 10 min at 30°C, the reaction was stopped by adding 240 l of assay buffer containing 6.8 M guanidine hydrochloride. The mixture was centrifuged at 12,000 ϫ g for 10 min at room temperature, and the release of p-nitroaniline was measured by absorbance at 405 nm. All assays were performed in triplicate. K m and K i values were calculated by non-linear regression using GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA). The K i values were determined by curve fitting of the data using GraphPad Prism 5.0 to Equations 1 and 2 (24).
In these equations, [E], [I], and [S] denote the enzyme, inhibitor, and substrate concentration, respectively. V i and V max denote the rate of pNA release in the presence and absence of the inhibitor, respectively. Circular Dichroism (CD) Spectrum Measurements-CD spectra of the cucumisin propeptide and its mutants placed in a fused silica cuvette of 0.1-mm path length were recorded from 250 to 190 nm or from 250 to 210 nm in the presence of urea at 20°C using a Jasco J-720C spectrometer. Protein concentration was about 1 mg/ml, and the buffer used was 10 mM sodium phosphate buffer, pH 7.5 containing 0.2 M NaCl with or without urea. For measuring thermal denaturation, the temperature was increased by 0.5°C/min from 20 to 90°C. Data were monitored for CD absorbance at 220 nm. For deconvolution, CD spectra were subjected to multicomponent secondary structure analysis using the CONTIN/LL algorithm (25).

RESULTS
Sequence Alignments among Subtilase Propeptides-The amino acid sequence of the cucumisin propeptide was compared with those of several plant subtilases and bacterial subtilisin E (Fig. 1). The cucumisin propeptide had 25-45% identities with those of the other plant subtilases. Although the sequences of plant subtilase propeptides had low identities (Ͻ20%) with that of subtilisin E, they contained conserved motifs N1 and N2, which were previously identified within bacterial subtilisins and appeared to be critical for protease domain folding (26). In the subtilisin E propeptide, hydrophobic residues located within N1 and N2 motifs (Val 41 , Phe 43 , Ile 59 , Val 66 , Leu 80 , Val 85 , Leu 88 , Val 94 , and Val 97 ) constituted a hydrophobic core. Among plant subtilase propeptides, most of these hydrophobic residues, including Val 41 , Leu 80 , Leu 88 , Val 94 , and Val 97 , located within N1 and N2 motifs of subtilisin E were well conserved. It is assumed that these conserved hydrophobic amino acids are important for the functions of the cucumisin propeptide.
Most propeptides of bacterial subtilisins are known to be removed by autoprocessing. In the case of plant subtilases, it was recently reported that the cleavage of the tomato SlSBT3 propeptide also occurred by autoprocessing (15). The NH 2 -terminal residues of mature regions of plant subtilases, including SlSBT3 (Thr-Thr-(Arg/His)-(Ser/Thr)), are well conserved, suggesting that cleavages of most plant subtilase propeptides are also likely to occur autocatalytically as with the SlSBT3 propeptide.
Expression and Purification of Recombinant Cucumisin Propeptide-The bacterial subtilisin propeptide plays an important role in inhibiting the active domain and folding of the mature enzyme (17)(18)(19). Because the amino acid sequences of plant subtilase propeptides have weak similarities with that of bacterial subtilisin and several hydrophobic amino acids are well conserved between plant subtilases and bacterial subtilisin as described above, it was assumed that plant subtilase propeptides have functions similar to those of the subtilisin propeptide. To evaluate this assumption, cDNAs for the propeptides of cucumisin, Arabidopsis ARA12, rice RSP1, subtilisin E, and several cucumisin propeptide mutants were subcloned into expression vectors in E. coli as His-tagged peptides with 23 extra amino acids at the NH 2 terminus. Then, the inhibitory activities of purified recombinant propeptides against mature cucumisin were assayed. As all of the recombinant proteins were insoluble in aqueous solution, they were dissolved in a buffer containing 8 M urea and then purified by affinity column chromatography using a nickel-Sepharose column after which the urea was removed by dialysis as described under "Experimental Procedures." Because the cucumisin propeptide without extra amino acids at the NH 2 terminus (cuc-pro-WT) expressed in E. coli was also insoluble, it was dissolved in the urea-containing buffer and then dialyzed to remove urea after which an additional DEAE-Sepharose column chromatography step was carried out for thorough purification. The homogeneities of the purified propeptides were confirmed by SDS-PAGE. Each purified recombinant propeptide migrated as a single major protein band (Fig. 2).
Inhibitory Activities of Recombinant Cucumisin Propeptide and Related Polypeptides-Recombinant cuc-pro-WT strongly inhibited cucumisin (Fig. 3A). Preincubation of cucumisin with cuc-pro-WT for 10 min before the enzyme assay did not affect the remaining activity of cucumisin (data not shown), suggesting that the propeptide acted as a typical rapid equilibrium inhibitor. Using 16 nM cucumisin and 2.1 mM Glt-Ala-Ala-Pro-Leu-pNA as a substrate, the IC 50 value for inhibition by cucpro-WT was ϳ20 nM, and the K i value determined was 6.2 Ϯ 0.55 nM (Fig. 3A). These results indicated that the cucumisin propeptide was a tightly binding inhibitor of cucumisin. As shown in Fig. 3B, the IC 50 values for inhibition of cucumisin by cuc-pro-WT increased linearly in proportion to the substrate concentration, clearly demonstrating a competitive type of inhibition.
As some subtilase propeptides have been shown to inhibit not only their cognate proteases but also other homologous proteases (27), we examined whether the propeptides of two plant subtilases, ARA12 and RSP1, and subtilisin E could inhibit mature cucumisin. Interestingly, ARA12 and RSP1 propeptides inhibited cucumisin with the K i values of 62.0 Ϯ 11 and 100 Ϯ 12 nM, respectively, but that of subtilisin E did not inhibit cucumisin (Fig. 3C). In contrast, subtilisin Carlsberg was not inhibited by the propeptides of three plant subtilases (cucumisin, ARA12, and RSP1) but was strongly inhibited by the subtilisin E propeptide as reported previously (data not shown) (18). These results suggested the compatibility with and the inhibitory specificity of plant subtilase propeptides for their cognate enzymes.
Important Region(s) within Cucumisin Propeptide for Inhibition of Mature Enzyme-To evaluate the important region(s) within the cucumisin propeptide for the inhibition of mature enzyme, we expressed NH 2 -or COOH-terminal truncated propeptides in E. coli as His 6 -tagged proteins and measured the inhibitory activities of these recombinant propeptides against mature cucumisin (Fig. 4). The 9 NH 2 -terminal amino acidtruncated propeptide (cuc-pro⌬N9) still had strong inhibitory activity against mature cucumisin (K i ϭ 14.7 Ϯ 0.72 nM) comparable with that of cuc-pro (K i ϭ 7.1 Ϯ 0.37 nM), whereas a 16-amino acid-truncated propeptide (cuc-pro⌬N16) showed a much weaker inhibition (K i ϭ 5.5 Ϯ 0.73 M). This suggested that the region from Asn 32 to Met 38 (NIYIVYM) was important for the inhibitory activity. Similarly, the 7 COOH-terminal amino acid-deleted propeptide (cuc-pro⌬C7) showed strong inhibition (K i ϭ 52.6 Ϯ 6.5 nM) (i.e. about 6 times less inhibition than cuc-pro), whereas the 14 COOH-terminal amino aciddeleted propeptide (cuc-pro⌬C14) resulted in no inhibition against cucumisin, suggesting that the region from Gly 97 to Leu 103 (GVVSVFL) was also important for the inhibitory activity. It should be noted that these two important regions, NIYIVYM and GVVSVFL, have hydrophobic characteristics. Collectively, these results indicate that each Asn 32 -Met 38 and Gly 97 -Leu 103 region has no inhibitory activity by itself, and the possible cooperation between these two hydrophobic regions is likely necessary for the inhibitory activity of the propeptide.
Inhibitory Activities of Point Mutants Derived from Cucumisin Propeptide-The hydrophobic amino acid residues Ile 35 , Val 36 , Tyr 37 , Val 98 , and Val 101 in Asn 32 -Met 38 and Gly 97 -Leu 103 regions within the cucumisin propeptide are well conserved among plant subtilases (Fig. 1). Hydrophobic amino acids Val 41 , Val 94 , and Val 97 in the subtilisin E propeptide, cor-  (40). The numbers of amino acid residues begin from the first Met. The first amino acid residues of propeptides were predicted using the SignalP 3.0 program (41). Well conserved amino acid residues within the propeptides (over 70% of the propeptides) and similar amino acid residues are shaded in black and gray, respectively. Gaps are denoted by dashes. Gray bars above the sequences represent the hydrophobic regions within the cucumisin propeptide. Two boxed regions show motifs N1 and N2. The secondary structures of the subtilisin E propeptide are shown below the sequences. The vertical arrow indicates the propeptide processing sites. ARA12 (14) and AIR3 (42)  responding to Val 36 , Val 98 , and Val 101 in the cucumisin propeptide, were reported to form a hydrophobic core (26,28). As it was expected that these hydrophobic residues in the cucumisin propeptide would be responsible for the inhibitory activity against cucumisin, site-directed mutagenesis of these residues for substitution to Ala were performed as described under "Experimental Procedures." The inhibitory activities of these recombinant propeptide mutants were measured. In addition, Ile 33 , located within the Asn 32 -Met 38 region but not a conserved hydrophobic residue among plant subtilases, was also substituted with Ala. The K i values of these mutants against cucumisin are listed in Fig. 4   V36A/V98A/V101A, showed only slightly weaker inhibition than V36A. These results suggest that Val 98 and Val 101 , which are conserved among many subtilases, are not too important for the inhibition and the possible cooperation between two hydrophobic regions.
CD Spectroscopy for Cucumisin Propeptide-Recombinant subtilisin propeptide is completely unfolded without a protease domain and is folded when it binds to the protease domain (29,30), although some subtilase propeptides are folded without protease domains (27,31). To determine whether the recombinant cucumisin propeptide had a stable conformation without a protease domain, CD spectroscopy was performed. Under non-denaturing conditions, the spectra of cuc-pro and cucpro-WT were very similar to each other and revealed some negative ellipticity at 208 and 222 nm (Fig. 5A), corresponding to the CD spectrum of random coil and ␣-helix, respectively. ␤-Sheet also seems to contribute to the spectrum around 215 nm. In contrast, the intensity around 215-230 nm remarkably decreased under denaturing conditions with 8 M urea, indicating the decrease of ordered secondary structures except for random coil. The deconvolution of the CD spectra suggests that cucpro-WT and cuc-pro contain 15.0 and 15.9% ␣-helix and 33.3 and 33.5% ␤-sheet, respectively (Table 3). These results indicate that both cucumisin propeptides, with or without 23 extra NH 2 -terminal amino acids, have secondary structures by themselves without the protease domain. As there were no significant differences of the K i values between cuc-pro (7.1 Ϯ 0.37 nM) and cuc-pro-WT (6.2 Ϯ 0.55 nM), the 23 NH 2 -terminal amino acids in cuc-pro were likely not to affect either the inhibitory activity or the secondary structure of the cucumisin propeptide. This validated the use of His 6 -tagged recombinant propeptides, such as cuc-pro and its mutants, for the following experiments.
Thermal Stability of Cuc-pro Conformation-For cuc-pro, the intensity of the negative ellipticity at 222 nm in its CD spectrum decreased with increasing temperature from 30 to 90°C, indicating the unfolding of cuc-pro by heat treatment (Fig. 5B). The transition temperature (T m ) for thermal unfolding was ϳ55°C. In contrast, when the temperature decreased from 90 to 30°C, the intensity of the negative ellipticity increased and returned to its level before heating (Fig. 5C). The intensity of the negative ellipticity at 222 nm also decreased with increasing urea concentration at 30°C (Fig. 5D). The decrease of intensity in 8 M urea, however, was significantly larger than that at 90°C without urea (Fig. 5, B and D). These results suggested that the thermal unfolding of the cucumisin propeptide, even at 90°C, was partial and reversible. The inhibitory activity of cuc-pro was also fairly stable up to 90°C for 10 min (Fig. 5E). Because the inhibition assay was performed at 30°C, the structure of cuc-pro was thought to have quickly recovered during the enzyme assay, demonstrating the reversibility of the thermal unfolding of the cucumisin propeptide.
CD Spectroscopy for Ala Substitution Mutants-If the tripeptide Ile 35 -Val 36 -Tyr 37 formed a hydrophobic core within the cucumisin propeptide, Ala substitution of these residues could affect the propeptide conformation. To verify this hypothesis, CD spectra of Ala substitution mutants were measured. The CD spectrum of the Y37A mutant, which strongly inhibited cucumisin, was very similar to that of cucpro (Fig. 6). In contrast, the CD spectra of V36A and I35A/ V36A/Y37A revealed that the ordered secondary structures of these mutants were significantly decreased. In particular, the decreasing intensity of the negative ellipticity at 222 nm was remarkable and correlated with the reduction of the inhibitory activity. The estimated content of ␣-helix in these two mutants was also significantly lower than that in cucpro, especially in I35A/V36A/Y37A (9.3%) ( Table 3). These results suggest that Ile 35 -Val 36 -Tyr 37 contribute to the for-mation of the proper secondary and, probably, the tertiary structure necessary to inhibit cucumisin.

DISCUSSION
Plant subtilases are thought to be synthesized as precursors containing NH 2 -terminal propeptides, but the functions of these propeptides are unknown. We found that purified recombinant cucumisin propeptide is a potent tightly binding competitive inhibitor of mature cucumisin. The K i value of cuc-pro-WT was 6.2 Ϯ 0.55 nM, suggesting that the enzymatic activity of the plant subtilase zymogen was regulated by the strong inhibitory activity of its propeptide. In this regard, the propeptides of bacterial subtilisin have also been reported to act as competitive inhibitors of their protease domains with inhibition constants in the nanomolar range (17,18). Besides plant subtilases, the propeptides of plant thiol proteases, such as papain and papaya proteinase IV, were also reported to inhibit their cognate proteases (32,33).
The proteolytic activity of cucumisin was also strongly inhibited by ARA12 and RSP1 propeptides but not by the subtilisin E propeptide (Fig. 3C). In contrast, the propeptides of cucumisin, ARA12, and RSP1 did not inhibit bacterial   subtilisin. The amino acid sequences of ARA12 and RSP1 propeptides are about 36% identical to that of the cucumisin propeptide (Fig. 1), and the K i values of cucumisin inhibition by ARA12 and RSP1 propeptides were about 20-fold higher than that of the cucumisin propeptide. These results show that the inhibitory activities of plant subtilase propeptides are dependent on their selectivity and compatibility with their cognate enzymes. In this regard, it has also been reported that propeptides of some bacterial subtilases inhibit other types of subtilase. For example, the aqualysin I propeptide, a thermostable subtilase synthesized by Thermus aquaticus YT-1, inhibits not only aqualysin I but also subtilisin BPNЈ (27). The amino acid sequence of the aqualysin I propeptide, however, is only 21% identical to that of the subtilisin E propeptide. In another astonishing example, Pleurotus ostreatus proteinase A inhibitor 1 (POIA1), which is not a protease propeptide, can inhibit subtilisin BPNЈ and can act as its intramolecular chaperone, although the amino acid sequence of POIA1 is only 18% identical to that of the subtilisin BPNЈ propeptide (34). Despite the low amino acid sequence similarity between POIA1 and subtilisin BPNЈ propeptides, the overall structural topology of the POIA1 propeptide is very similar to that of the subtilisin BPNЈ propeptide (35). These findings strongly support the idea that higher order structures of plant subtilases propeptides, rather than their primary structures, are important for their inhibitory activities against their cognate enzymes. Many bacterial subtilases are activated by removing their propeptides by autoprocessing, and the COOH termini of the propeptides are thought to inhibit enzymes by binding to their active site clefts (28). However, it is unknown whether the cucumisin precursor can be activated by removing its propeptide by autoprocessing. Our finding that a cuc-pro⌬C7 propeptide mutant (without 7 COOH-terminal amino acids) still inhibited cucumisin with a K i value of 52.6 Ϯ 6.5 nM (Fig. 4) demonstrated that the 7 COOH-terminal amino acids in a propeptide are not essential for its inhibitory activity. Jean et al. (31) also reported that a PfSUB-1 propeptide mutant without 11 COOH-terminal amino acids inhibited its cognate protease with about 14-fold larger K i value. Regarding the processing of the prodomain of plant subtilase precursor, Cedzich et al. (15) recently reported that the cleavage of the prodomain of tomato SlSBT3 occurs autocatalytically, and zymogen maturation is an intramolecular process.
The site of the processing of cucumisin prodomain and the conditions under which the propeptide-enzyme complex dissociates in planta are unknown. It has been reported that the processing of the prodomain of tomato SlSBT3 in the ER is a prerequisite for passage through the secretory pathway using a transient expression system in Nicotiana benthamiana leaves (15). As cucumisin is secreted and accumulated in the juice in melon fruits, it is also likely to be sorted along the secretory pathway after processing of the prodomain in the ER. Regarding the activation of the propeptide-enzyme complex, pH-regulated activation of furin in the secretory pathway and a pH sensor in the furin propeptide have been reported (36). Clarifying the site and timing of the processing of cucumisin prodomain and the mechanisms of the dis-sociation of propeptide-enzyme complexes will be major tasks for the future.
The NH 2 -teminal amino acid residues of mature regions (Thr-Thr-(Arg/His)-Ser/Thr) are well conserved among plant subtilases (Fig. 1) (14). Regardless of the sequence homologies of NH 2 -teminal amino acid residues of mature plant subtilases, the substrate specificities of plant subtilases that have so far been reported were quite different from each other. For example, cucumisin shows broad substrate specificity; ARA12 shows preference for Phe and Ala at the P 1 position and for Asp, Leu, and Ala at the P 1 Ј position (37); soybean C1 prefers Glu at the P 1 and Glu/Gln at the P 1 Ј position (38); and tomato SlSBT3 shows a preference for Gln and Lys at P 1 and P 2 positions (15). These findings suggest that the mechanisms for recognizing the propeptide processing site are different from that for substrate recognition during proteolysis by mature proteases.
The analysis of the important region(s) within the cucumisin propeptide for the inhibition of the mature enzyme using recombinant truncated propeptides and Ala-substituted mutants (Fig. 4) clearly showed that each of the two hydrophobic regions, Asn 32 -Met 38 (NIYIVYM) and Gly 97 -Leu 103 (GVVSVFL), had no inhibitory activity by itself. Thus, the possible cooperation between these two hydrophobic regions along with the formation of the higher order structure is likely necessary for the inhibitory activity of the propeptide. Indeed, the CD spectrum of the cucumisin propeptide revealed that it has a secondary structure by itself without the protease domain (Fig. 5A). In this respect, the propeptides of aqualysin I (27), PfSUB-1 (31), and human proprotein convertases (39) were also reported to form secondary structure by themselves. By comparison, the subtilisin BPNЈ propeptide has been reported to be unfolded by itself and is folded correctly only when it forms a complex with the protease domain (28,30). The structure of cucumisin prodomain also may be changed to some extent upon binding to the protease domain. Interestingly, the aqualysin I propeptide can inhibit subtilisin BPNЈ more strongly than the subtilisin BPNЈ propeptide (27). The mutants of subtilisin BPNЈ propeptide, which could have some secondary structures because of the introduction of amino acid replacements, had K i values of inhibition against subtilisin BPNЈ lower than that of a wild-type propeptide (30). These studies on bacterial subtilisins also support the idea that formation of the secondary structure is necessary for the inhibitory activity of the cucumisin propeptide against cucumisin.
Cuc-pro and Y37A had very similar CD spectra, suggesting that an Ala substitution at Tyr 37 did not affect the secondary structure of the cucumisin propeptide (Fig. 6). In contrast, the CD spectra of V36A and I35A/V36A/Y37A were different from that of cuc-pro especially with regard to the noticeable decreasing intensity of negative ellipticity at 222 nm, suggesting that the content of the ordered secondary structure in V36A and I35A/V36A/Y37A was decreased after the substitutions to Ala. The decreasing intensity of negative ellipticity at 222 nm for the ordered secondary structure of I35A/V36A/Y37A was more remarkable than that of V36A. K i values of both V36A (37.4 Ϯ 4.0 nM) and I35A/V36A/Y37A (200 Ϯ 27 nM) were higher than that of cuc-pro (7.1 Ϯ 0.37 nM), and I35A/V36A/Y37A inhibited cucumisin more weakly than did V36A (Fig. 4). The CD spectrum of I35A was also different from that of cuc-pro as the intensity of the negative ellipticity around 203 nm for the secondary structure of I35A was remarkably increased, similar to V36A (Fig. 6). For the CD spectrum around 222 nm, however, I35A was more similar to cuc-pro than to V36A. Because I35A had a strong inhibitory activity comparable with cuc-pro, the change of the secondary structure monitored around 203 nm for the Ala substitution of Ile 35 , which shows the increase of random coiled structure, was not likely critical for the inhibitory activity. As shown in Table 3, the estimated content of ␣-helix in V36A (14.4%) was lower than that of cuc-pro (15.9%). The decrease of the ␣-helix content in I35A/V36A/Y37A (9.3%) was remarkable. Collectively, the secondary structure monitored around 215-222 nm, primarily due to the contribution of ␣-helices and ␤-sheets, is suggested to be important for the inhibition of the protease domain. The random coiled structure observed for cuc-pro and its mutants may be converted to ordered secondary structures after docking to the cognate protease.
We described that the proper secondary structure along with the assistance of some hydrophobic residues was evidently important for the inhibitory activity of cucumisin propeptides. For the maturation of the cucumisin precursor, disabling the inhibitory activity of the propeptide and its degradation by the cognate or other protease activity prior to the activation of the cucumisin precursor were thought to be essential. The analyses of NH 2 -or COOH-terminal truncated mutants suggested that the degradation of the two hydrophobic regions in the cucumisin propeptide could easily weaken its inhibitory activity. To evaluate the detailed mechanisms of the inhibition by the propeptides of cucumisin and other plant subtilases, further structural studies, including x-ray analysis, will be necessary.