Isolation and functional diversity of Bowman-Birk type serine proteinase inhibitors from Hyacinthus orientalis

Bowman–Birk inhibitors (BBI) are plant-derived serine proteinase inhibitors. Endogenously, they function as defense molecules against pathogens and insects, but they also have been explored for applications in cancer treatment and inflammatory disorders. Here, we isolated 15 novel BBIs from the bulb of Hyacinthus orientalis (termed HOSPIs). These isoinhibitors consisted of two or three chains, respectively, that are linked by disulfides bonds based on proposed cleavage sites in the canonical BBI reactive site loop. They strongly inhibited trypsin ( K i = 0.22 - 167 nM) and α-chymotrypsin ( K i = 19 - 1200 nM). Notably, HOSPI-B4 contains a six-residue reactive loop, which appears to be the smallest such motif discovered in BBIs to date. HOSPI-A6 and -A7 contain an unusual reactive site, i.e. Leu-Met at the P1-P1' position and have strong inhibitory activity against trypsin, α-chymotrypsin and elastase. Analysis of the cDNA encoding HOSPIs revealed that the precursors have HOSPI-like domains repeated at least twice with a defined linker sequence connecting individual domains. Lastly, mutational analysis of HOSPIs suggested that the linker sequence does not affect the inhibitory activity, and a Thr residue at the P2 site and a Pro at the P3' site are crucial for elastase inhibition. Using mammalian proteases as representative model system, we gain novel insight into the sequence diversity and proteolytic activity of plant BBI. These results may aid the rational design of BBI peptides with potent and distinct inhibitory activity against human, pathogen, or insect serine proteinases.


Introduction
Proteinase inhibitors (PI) are widely distributed in all domains of life and they are involved in the regulation of several biochemical processes. For instance, in animals PIs are important for blood coagulation regulation, immune system activation, and to tune the activity of proteinase processing of peptide hormones [1]. In plants, PIs are involved in maintaining physiological homeostasis and serve the plants' innate defense machinery against herbivore attacks or microbial infections [2]. Research on PIs in plants started with the discovery of a soybean trypsin inhibitor by Kunitz in 1947 [3]. Since then, many plant-derived small molecule and proteinaceous PIs have been discovered. Currently there are more than 6700 individual plant PIs known that are classified into at least 12 distinct types, such as serpins, phytocystatins, Kunitz inhibitors, Bowman-Birk-, α-amylase-trypsin-, mustard-type-, potato type-I-, potato type-II-, potato metallocarboxypeptidase-, squash-and cyclotide inhibitors [2,4]. Therefore these peptides PIs provide promising molecules for discovery and development of drug candidates or agrochemicals.
Bowman-Birk inhibitors (BBI) from soybean have been studied in clinical trials as anticancer agents for melanoma and oral cancers [5,6]. BBIs are cysteine-rich proteins typically consisting of 60-90 amino acids including 10-12 cysteines, which form a cysteine framework. BBI stability is a result of the disulfide bonds supporting a compact protein structure and a rigid conformation, which result into a considerable resistance against heat, acid and basic conditions (Figure S1A-C) [7][8][9].
They were originally isolated from species of the legume family of dicotyledonous plant [10]. The inhibitory domain (BBI domain) consists of an anti-parallel β-sheet motif with two inhibiting loops on opposite sides of the molecule [2,9], which is also referred to as double-headed inhibitor ( Figure   S1B) [11]. BBIs have also been isolated from monocotyledonous plants of the Poaceae family.
Intriguingly these BBIs contain only one functional reactive site located on inhibition loop in one fully functional motif, whereas the lack of key disulfide bonds in the opposite face result in a loss of proteinase inhibitory capacity in the second motif ( Figure S1A) [12,13]. The canonical reactive sites of BBIs are present in the stabilized inhibition loops by highly conserved internal disulfide bond, which contain 9 to 11 residues. The enzyme specificity of BBIs is linked to the sequence of the reactive site; in particular the P1 site appears to be important to guide substrate preference of the target proteinase [14,15].
We previously reported the amino acid sequences, gene structures and biochemical properties of BBIs from onion bulb (Allium cepa) as the first BBI found in Liliaceae spp. [16], which exhibit high sequence homology to leguminous plant PIs. In the onion-derived inhibitors, a particular cysteine residue near the second reactive site was absent leading to functional impairment of the second inhibitory module. However, we demonstrated that the onion BBI had not only in trypsin inhibition, but also in the inhibition of chymotrypsin [16]. Thus, the mechanism and selectivity of some BBIs In this study, 15 novel proteinase inhibitors from hyacinth bulb (Hyacinthus orientalis L., formerly known to belong to the Liliaceae plant family) were isolated and their primary structures were determined. Based on homology analysis they belong to the family of BBI and display a characteristic highly conserved cysteine residues. Since there is limited information of monocot plant PIs, we were interested to investigate the diversity of BBI in hyacinth. Interestingly, these novel inhibitors had unique sequence motifs in their inhibition loop. Therefore, we determined their inhibitory activity against representative mammalian serine proteinases. Accordingly, they were termed Hyacinthus orientalis serine proteinase inhibitors, HOSPIs. To gain valuable insight into the sequence-activity relationship of these interesting peptide PIs, a cDNA transcript analysis was performed. Cloning of six HOSPI genes from the bulb revealed two distinct types of genes.
Interestingly, the HOSPI-like domains were repeated at least twice in each gene with a signature linker sequence connecting the domains. To investigate the role of this linker motif, their sequence variations and the function of the different domains of the precursor, we established the expression system in E. coli of a fusion peptide with the linker and the mature HOSPI sequence derived from a representative HOSPI gene. These results suggested that the linker sequence does not affect the inhibitory activity. Further, a Thr residue at the P2 site and a Pro at the P3' site are crucial for elastase inhibition. Thus, the surrounding residues of the reactive site might contribute to the recognition and/or interaction with the enzymes. In this study we are able to provide a system to study the functional variety of plant BBIs for the inhibition mechanism of mammalian serine proteinases.

Materials and reagents
Hyacinthus orientalis L. bulbs were purchased at Home Center & Home Fashion Store, Nafco (Fukuoka, Japan). Proteinases were from Sigma Chem. Co. (St. Louis) and proteinase substrates (as outlined below) were purchased from the PEPTIDE INSTITUTE (Osaka, Japan).

Preparation of hyacinth extract and purification of inhibitors via chromatography
H. orientalis bulbs (626.2 g) were homogenized using a common household blender. The processed mixture was then centrifuged at 9000 rpm (KUBOTA 6500, KUBOTA Co., Tokyo, Japan) at 4°C for 40 min to obtain the aqueous supernatant containing the inhibitors of interest. The extract solution was thermally inactivated at 60°C for 30 min to remove precipitate proteins and as inactivation of enzymatic activity. The supernatant was mixed with saturated ammonium sulfate until made up final concentration 80%. The obtained precipitate was dissolved in 120 ml water, dialyzed and lyophilized to yield the crude inhibitor cocktail. Afterwards 2.0 g of crude inhibitor powder were dissolved in 50 mM Tris-HCl (pH 8.0) and applied to DEAE-Toyopearl 650 M ion exchange column (5.0 × 27 cm, TOSOH Co., Tokyo, Japan). The proteins were eluted with a gradient from 0 to 0.35 M NaCl in 50 mM Tris-HCl (pH 8.0). Finally, the proteinase inhibitors of each fraction were purified by reversedphase high performance liquid chromatography (HPLC) using a YMC-Pack C8 column (1.0 x 15 cm, YMC CO. LTD., Kyoto, Japan) with a linear gradient of solvents 0.1% TFA (eluent A) and acetonitrile in 0.1% TFA (eluent B) at the flow rate of 3.0 mL/min. Protein elution was monitored by absorbance at 230 nm.

Mass spectrometry
Molecular mass of each purified inhibitor was analyzed by a Voyager DE-STR matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometer (Applied Biosystems Japan, Tokyo, Japan). The analysis was performed by following the manufacturer's instructions. Briefly, approximately 3 g of each sample was dissolved in 0.1% TFA-50% acetonitrile containing saturated aqueous solution of sinapinic acid (Sigma Chem. Co., St. Louis) and spotted for analysis onto the target plate. The concentration of each purified inhibitor was determined using the molar extinction coefficients (M -1 cm -1 ) at 280 nm calculated by ProtParam tool on ExPASy server [17]. The  value of each inhibitor was listed in Supplementary Table S1. For each proteinase activity assay, the increase in absorbance at a wavelength of 410 nm was monitored to calculate the activity value. The apparent inhibition constant K i app was determined using a computational tool for curve fitting based on Morrison equation for tight binding of competitive inhibitors as shown below [18,19].

Proteinase inhibitory activity
The inhibition constant K i was calculated from the following equation [20].
K m values for trypsin, chymotrypsin, and elastase were determined experimentally using the Lineweaver-Birk plot [21].

SDS-polyacrylamide gel electrophoresis (SDS-PAGE)
SDS-PAGE was carried out on 12.5% gels by Laemmli's method [22]. Prestained XL-ladder marker kit of Apro Science (Tokyo, Japan) was used for the estimation of molecular weights. The sample after separation on SDS-PAGE gel was fixed with trichloroacetic acid/methanol/water (1:4:5, w/v/v).

Determination of amino acid sequence
Each purified LOSPI (0.5-1.0 mg) was dissolved in 6 M guanidine hydrochloride, 10 mM EDTA, 0.5 M Tris-HCl (pH 8.6), and disulfide bonds were reduced with dithiothreitol (DTT) of 10 molar equivalents against the number of cysteine residues. Then, reduced-HOSPI was allowed to reacted with 4-vinylpyridine of 3 molar equivalent for DTT at room temperature for 90 min to yield Spyridylethylated proteins according to the procedure indicated in the reference [23]. By-product was eliminated by re-chromatography, and the modified proteins were purified. In this procedure, two or three peptide chains were detected in a separate by RP-HPLC after reductive reaction.
The pyridylethylated protein was digested with various proteinases (Arg-C, V8 proteinase, Asp-N), and fragmented peptides were separated by reverse phase HPLC using a YMC-Pack ODS C18 column (4.6 × 250 mm) column. The amino acid sequence of each peptide was determined by Nterminal Edman degradation using a protein automated sequencer PPSQ series (Shimadzu, Kyoto, Japan).

cDNA cloning and expression of HOSPI by Escherichia coli
The total RNA was extracted from the bulbs of H. orientalis according to a standard protocol [24].
The cDNA encoding HOSPI was cloned using a method of Deshimaru et al. [16]. Briefly, cDNA was synthesized by reverse transcription from RNA using an adaptor-linked oligo (dT) primer (5'-GGCCACGCGTCGACTAGTAC(T) 17 -3'). They were amplified by PCR using primers designed based on amino acid sequence of the purified natural inhibitor (Table S2). The primers LOSPI5-N

Isolation, purification and amino acid sequencing of LOSPIs
The aqueous extract from hyacinth bulb was applied to anion exchange chromatography (DEAE-Toyopearl), and each fraction was tested for inhibitory activity against trypsin ( Figure 1A). The active fractions were pooled in three fractions (named A, B and C), which were carried forward to further isolation and characterization. The isolated proteins responsible for the trypsin inhibition were named H. orientalis serine proteinase inhibitors, HOSPIs. The seven inhibitors in fraction A (HOSPI-A1 to -A7, Figure 1B), seven inhibitors in fraction B (HOSPI-B1 to -B7, Figure 1C), and one inhibitor in fraction C (HOSPI-C1, Figure 1D) were isolated using by reversed-phase HPLC. Amino acid sequences of proteinase inhibitors were determined by Edman degradation. As shown in Figure S2, full length sequences were obtained for HOSPI-A4, -A6, -A7, -B1, -B4 to -B7, and -C1, and partial sequences for HOSPI-A1 to -A3, -A5 and -B3 ( Figure S2). Most but not all of these novel inhibitors were homologous to soybean BBI, which is a prototypical dicotyledonous plant BBI [25]. Sequence comparisons exhibited that HOSPIs only share low sequence homology with BBIs (18-29% identity), however, the cysteine residues in these inhibitors are well conserved as well as the typical BBI domain ( Figure 2, Figure S1C). One of the disulfide bonds, usually embracing the second reactive site of soybean BBI, disappeared in HOSPIs. In a comparative analysis of amino acid sequences of HOSPIs, they were classified into two types: the 10 cysteine residues type and the 12 cysteine residues type. In addition, there are those of the N-terminus Gly (or Glu), or Trp-Pro-Pro-Val-Glu in which additional N-terminal extension are observed such as in HOSPI-A3 to -A7 and -B3 ( Figure 2).
Seven HOSPIs (HOSPI-A2, -A3, -A5, -B4, -B5, and -B6) consist of two chains, and HOSPI-B1 has three chains (Table S1). The chains were identified by peptide sequencing of HPLCderived S-pyridylethylated proteins after reduction of sulfhydryl group (refer to "P; the intact pyridylethylated protein before digestion by enzymes in Figure S2). These results suggested that the two or three chains were cross-linked by disulfide bonds. The amino acid sequences in the C-terminal region of these seven HOSPIs showed, Gly-Lys (Met or Thr)-Pro (His)-Leu (Met)-Tyr-Gln, corresponding to the second reactive loop in soybean BBI ( Figure 2). Another cleavage site observed in HOSPI-B1 and -B4 was detected between Arg and Ser on a first reactive loop. The amino acid sequences between HOSPI-B4 and HOSPI-B1 shows high homology, which can be distinguished by a loss/insert of only two residues (Leu and Ser) in cleavage site of the C-terminal region described above. However, HOSPI-B4 has two peptide chains without the cleavage site in the C-terminal region. The first reactive site loop of HOSPI-B1 and -B4 are made up of only 6 residues (Cys-Ala-Arg-Ser-Leu-Cys), which is the smallest reactive loop in BBI reported so far. In summary, we isolated We next focused on investigating the nine HOSPIs with a determined full length amino acid sequence. To investigate the inhibitory activity (K i ) of the nine HOSPIs against serine proteinase, we used mammalian trypsin, chymotrypsin, and elastase and bacteria subtilisin as representative proteinase model enzymes. None of the isolated HOSPIs was active as inhibitor of subtilisin, while all HOSPIs inhibited the proteolytic activity of trypsin. However, the inhibitory constants differed as K i values between 0.22 and 167 nM (Table 1). HOSPI-C1 showed the strongest inhibitory activity (K i = 0.22 nM), whereas HOSPI-B1 and B4, which contain the cleavage site on the first reactive loop and, exhibited the lowest trypsin inhibitory activity (K i = 167 and 110 nM, respectively). HOSPIs isolated from DEAE-fraction A showed stronger inhibition than HOSPIs from DEAE-fraction B (Table 1, Figure 1A). All HOSPIs, except HOSPI-B1 showed inhibition of chymotrypsin, but the inhibitory potencies were about 10-fold weaker as compared to trypsin inhibition. The weakest activity was observed in HOSPI-B4 and -C1 (K i = 1.2 mM), and HOSPI-B1 did not inhibit chymotrypsin. HOSPI-A6 and -A7 also inhibited elastase strongly, the K i value in case of HOSPI-A6 was 4.8 nM, while HOSPI-A7 could not determine it due to a low amount of native protein despite the activity similar trend with HOSPI-A6 (not shown in the table, we investigated details in a mutant analysis).
Interestingly, HOSPI-A6 and -A7 which has Leu-Met at the putative reactive site exhibited the highest inhibitory activity against trypsin, α-chymotrypsin and elastase. This is a unique sequence motif Leu-Met at the P1-P1' position in trypsin inhibitors (more details are described in the discussion section). The inhibitory activity profile of HOSPIs revealed that they have varying potencies against the different serine proteases.

Cloning of cDNA precursor proteins encoding HOSPIs
We isolated 15 novel serine proteinase inhibitors from hyacinth bulbs with different inhibitory activities. As an experimental approach to investigate the biosynthesis of these inhibitors in the plant, clones except HOSPI-x3 contained two HOSPI-domains (HOSPI-ω 1 and -ω 2, Figure 3B) and an additional sequence not encoding for the mature inhibitor, which we referred to as linker region. In addition, these precursors contained a C-terminal elongation, which we referred to as C-tail region.
The generated HOSPI-like cDNA clones were classified into two groups based on similarity of the sequence of mature domain ω1. 'type X' contained (i) a Leu-Met motif in the putative reactive site, (ii) 10 cysteine residues and (iii) a WPPVE motif at the N-terminus of the inhibitor coding sequence as observed in HOSPI-1-1, HOSPI-2-6, HOSPI-x3 ( Figure 3A). Comparison of the linker regions between HOSPI-1-1 and HOSPI-2-6, with 21 or 24 residues, respectively, suggested a deletion of three amino acids, Asp-His-Ser ( Figure S3). The C-tail region consists of 19 residues with high similarity (84% identity) among the three genes ( Figure 3A).

Recombinant expression of HOSPI precursors in Escherichia coli
The HOSPI-ω1 gene products encoded in the C-terminal region of the two types (X or Y) of cDNA clones were investigated for their proteinase inhibitory activity. Specifically, it was of interest to investigate proteinase selectivity, and possible effects of the linker sequence. Therefore two representative clones, HOSPI-1-1 and HOSPI-1-8 ( Figure 3A), were inserted into the expression vector pET32a and the fusion protein was produced using an E. coli expression system, (termed HOSPI-x1 and HOSPI-y1, Figure 4A). The sequences of the mature inhibitor HOSPI-X1 and Y1 encoded on LOSPI-x1 and LOSPI-y1 were compared with LOSPI-A7 of native inhibitor isolated from hyacinth bulb. As described above, HOSPI-A7 had a unique sequence motif, Leu-Met, at the P1-P1' position in the putative reactive site. The difference between HOSPI-A7 and HOSPI-X1 was Leu 59 or Gln 59 ( Figure 4B). On the other hand, HOSPI-Y1 was confirmed on the level of nucleic acid, but the native inhibitor has not yet been identified. Remarkably, the putative reactive site of HOSPI-Y1 is Trp 20 -Val 21 , which would be unique in BBI. The HOSPI-X1 and -Y1 were expressed as a fusion with thioredoxin-tag, His-tag, and S-tag derived of pET32a vector. Hereinafter referred to as, a protein expressed with linker sequence (underlined in Figure 4A) is HOSPI-X1 (or Y1)-lk, while a protein without the linker sequence is HOSPI-X1-mat (or Y1-mat). The recombinant inhibitors were obtained in soluble form, and their size of approximately 32 kDa and 30 kDa was confirmed by SDS-PAGE analysis, respectively. The purified fusion proteins were obtained by affinity chromatography using a Ni 2+ -sepharose column and subsequent reversed-phase HPLC ( Figure S4).
HOSPI-X1-lk and HOSPI-X1-mat showed strong trypsin inhibition within the same magnitude, K i values 7.0, or 7.7 nM, respectively ( Figure 5A). HOSPI-X1 was also a strong inhibitor of chymotrypsin whether with or without linker sequence ( Figure 5B), as shown by the K i values of 13 to 14 nM (Table 2). HOSPI-X1-lk and HOSPI-X1-mat showed a strong inhibition against elastase, as shown by the K i values of 6.3 or 3.1 nM, respectively ( Figure 5C, Table 2). However, they showed no inhibition for subtilisin a known serine proteinase derived from Bacillus sp. (data not shown).
These results showed similar extent of inhibitory activity to HOSPI-A7 (Table 1). In contrast, HOSPI-Y1-lk and -mat did not inhibit chymotrypsin or trypsin (Table 2). Thus, the linker sequences encoded on HOSPI-x1 and HOSPI-y1 were irrespective of activity against representative serine proteinases ( Figure 4A).

Sequence-activity studies of designed HOSPIs
We designed mutant inhibitors based on HOSPI-X1-lk by varying amino acid residues around the putative reactive site. First, the unusual Leu 20 -Met 21 at the P1-P1' position of HOSPI was interesting to mutate based on amino acids preference for trypsin, chymotrypsin or elastase inhibitions on the canonical BBI motif. Thus, the putative reactive site Leu 20 -Met 21 was changed to Ala-Met, Leu-Ala, Ala-Ala, and Arg-Ser by mutation experiments, as indicated with HOSPI-X1-AM, HOSPI-X1-LA, HOSPI-X1-RS in Table 2. Second, the residues of Arg 51 -Ser 52 located in the second canonical reactive site observed in dicot BBIs, Arg 51 was changed to Ala, as named mutant HOSPI-X1-R51A. Finally, in an alanine scan of the residues next to the reactive site, Thr 19 , Pro 23 and Pro 24 , were point mutated, as indicated with the probes HOSPI-T19A, HOSPI-X1-P23A, and HOSPI-X1-P24A in Table 2, respectively.
Regarding chymotrypsin inhibition, these changes of the reactive site and the vicinity residues These results revealed that the residues on P1, P2, P3' sites are pivotal interaction sites to maintain inhibition of elastase. All these mutants showed no inhibition for the subtilisin (data not shown).
We expected to recover part of the inhibitory activity of HOSPI-Y1 by designing an improved inhibitor using the putative reactive site sequence based on the Leu-Met motif in HOSPI-X1 and native HOSPI. The four mutants were prepared as follows; the mutant HOSPI-Y1-LM contained the reactive site of HOSPI-X1, the mutant HOSPI-Y1-RV had the Trp residue on the P1 site replaced to Arg, the mutant HOSPI-Y1-S19T had a Ser 19 changed to Thr based on result of mutant analysis of LHTI-X1, and the mutant HOSPI-Y1-Y22A had the Tyr 22 changed to Ala. However, they did not show inhibitory activity against trypsin and chymotrypsin, except HOSPI-Y1-S19T. Interestingly, the mutant HOSPI-Y1-S19T yielded inhibition against chymotrypsin with a K i value of 33 nM ( Figure   5D). These results suggested that the linker sequence does not affect the inhibitory activity and a Thr residue at the P2 site as well as a Pro at the P3' site are crucial for elastase inhibition.

Discussion
Among the proteinase inhibitors derived from plants, BBI constitute a large and diverse family [26].
BBIs are serine proteinase inhibitors abundantly expressed in dicotyledonous plants. BBIs from dicotyledonous plant have two reactive sites (double-headed) and a molecular weight of about 8 kDa [27]. On the other hand, BBIs from monocotyledonous plants can be classified into two groups: one that has a single reactive site and a molecular weight of < 8 kDa (they lack the second reactive site), and the other group that has two reactive sites with a molecular weight of < 16 kDa (i.e. two 8 kDa repeats) [28]. The reactive site is located in a unique loop connected by two β-sheets and a disulfide bridge [29]. This loop structure is a common feature in BBI family proteins and contributes to stabilization of the enzyme inhibition motif ( Figure S1).
In this study, 15 HOSPIs from hyacinth bulbs with a molecular weight between 6.5 -8.0 kDa were isolated. By homology to known inhibitors they have been classified as BBIs with conserved cysteine residues. In HOSPIs, the second reactive site conserved in other dicot plants, including soybean BBI, was not present due to a missing disulfide bond in the inhibitory loop ( Figure 2).
Remarkably, HOSPI-B1 and HOSPI-B4 have a six-residue reactive loop with cleavage site in the first reactive site loop similar to "modified inhibitors" [30]. This is the smallest inhibitor motif in BBIs that have been reported so far, and to the best of our knowledge the first of its kind observation in monocot BBIs.
Previously, we isolated a BBI from onion. The inhibitory activity of the onion BBI against chymotrypsin was 100-times weaker than the inhibition of trypsin. We demonstrated that Arg 17 in the onion BBI is an important residue for inhibitory activity against trypsin, and predicted that Leu 46 might be the responsible P1 site residue for chymotrypsin inhibition even though the onion BBI lacks a disulfide bond as observed in double-headed BBIs [16]. In this study, seven of nine LOSPIs for which we determined the complete amino acid sequence have an Arg in the P1 site of the first reactive loop, and they exhibit varying degrees of inhibitory activity, i.e. the K i values for trypsin inhibition were 4 to 5,000 -times higher than for chymotrypsin inhibition ( In the case of BBIs, the importance of the P3' residue is known; Pro at the P3' site of BBIs isolated from legumes is conserved. This Pro residue adopts a cis peptide bond and as a result the reactive site loop forms a type VIb β-turn [29,33]. The analysis of HOSPI-X1 further revealed the broad specificity against trypsin, elastase, and chymotrypsin. The inhibitory activity for trypsin and elastase were significantly decreased by mutations around the reactive site, especially the P2 site of HOSPI-X1. Additionally, the mutant P23A (Pro residue at the P3' site was replaced by Ala) showed low activity. These results suggested that the surrounding residues of the reactive site might contribute to the recognition and/or interaction with the enzymes. For instance, the type VIb β-turn supports inhibition activity [29,33]. Thus, the proteinase inhibitory activity of HOSPI is not only dictated by the few amino acids of the reactive site, but also by a larger segment usually referred to as binding face in BBI, which are surrounding amino acids of the reactive site or neighboring loops. processing enzymes having different substrate specificity will be necessary to process the precursor peptide into the mature HOSPI. Since the boundary of the HOSPI-X1 and the tail sequence is almost the same amino acid sequence as the N-terminal on the linker sequence of HOSPI-X1, the tail sequence might be separated by the same processing enzyme. We have not yet isolated the enzyme that processes the HOSPI precursor, however, our preliminary data (not shown) suggested that this enzyme belongs to the aspartic acid proteinases.
In summary, we isolated 15 novel BBIs from the bulb of H. orientalis, which potently inhibited mammalian serine proteinases. Notably, one of them contains the smallest inhibitory motif in BBIs discovered to date, and the others contain an unusual reactive site at the P1-P1' position resulting in distinct substrate preference. Studying recombinant inhibitor mutants led to the conclusion that the key residues for the observed the activity profile were Pro residues at the P2' and P3', respectively, and a Thr residue at the P2 site. Our data suggest that HOSPIs might interact with nearby residues of the catalytic site to build a shield in the enzyme's pockets. An ongoing three-dimensional structure analysis of a HOSPI in complex with a mammalian serine proteinase will clarify more molecular details of the inhibitor/enzyme recognition and inhibition mechanism. This might advance the design of selective inhibitors in the future.
At a more general level, BBI proteins are considered for their applications in drug development and agriculture. Firstly, they are emerging drug candidates for development of novel therapeutics of cancer and inflammatory diseases [6]. Secondly, it is known that BBI gene expression is activated as a response of the plants' innate immune system following herbivore attack or microbial infection. Therefore, BBIs are important plant defense molecules that may also be engineered in transgenic plants [38]. In this study, it was possible to recombinantly express functional HOSPI inhibitors in E. coli that target several different proteinases. This is a first step to demonstrate feasibility for upscaling production of these novel BBIs. By modifying the non-active HOSPI-Y1 it was possible engineer a desired functionality. This can be considered the groundwork to design more valuable HOSPI-based protein inhibitors in the future to be used, for example, as insecticides or therapeutics. We also thank Mrs. Rie Maenaka and Mrs. Yuki Uemura for helpful sample preparations. We appreciate staffs of Radioisotope Centre of Fukuoka University for the analysis of protein sequences in this study.  Tables   Table 1. Inhibitory activity (K i ) of isolated native HOSPIs against trypsin and chymotrypsin.

Inhibitors
Residues in vicinity of the reactive site a K i (nM) P 4 P 3 P 2 P 1 P 1 'P 2 'P 3 'P 4 ' trypsin chymotrypsin HOSPI-A4 A C T R M W P P 12 460 HOSPI-A6 A   Cysteine residues are highlighted in dark, and the first and second reactive site loops (inhibitory loops) based on soybean BBI are indicated by a connecting line (see Figure S1). The reactive sites (P1-P1' sites) against serine proteinases are marked with asterisks. Sequence gaps are indicated by hyphens. Black triangular marks indicate the proposed cleavage site of the separate chains linked by disulfide bonds. The N-terminal region of HOSPI-A2, -A3, -A5 and -B3 wasn't determined completely, therefore "x" indicates a partial sequence. The N-terminal residue of HOSPI-A7 is pyroglutamic acid and has been indicated with a superscript of "py". These complete or a partial amino acid sequences of native HOSPIs were assembled using the peptide sequences determined by Edman sequencing of peptides obtained by digestion of these proteins by several endoproteinases (see also Figure S2).  Figure S1).    Table S1. The average theoretical molecular weight (MW, molar extinction coefficient (M -1 cm -1 at 280 nm), and isoelectric point (pI) of purified each HOSPI was calculated computationally based on the amino acid sequence as shown in Figure 2 by the ExPASy server (https://web.expasy.org/protparam/). In this list, assuming all pairs of cysteine residues form disulfide bonds.

HOSPIs
The  Table S2. PCR primers designated based on amino acid sequences of the purified native inhibitors.  Figure S2. Example of amino acid sequences determination for native HOSPIs. The complete or a partial amino acid sequences of HOSPIs were assembled using the peptide sequences determined by Edman degradation, which were obtained by digestion of pyridylethylated HOSPIs peptides by each of the following enzymes (trypsin, Asp-N and Arg-C endoproteinases) or chemical breakdown (cyanogen bromide), respectively. Full length sequences were obtained for LOSPI-A4, -A6, -A7, -B1, -B4 to -B7, and -C1. Full length sequences were obtained for HOSPI-A4, -A6, -A7, -B1, -B4 to -B7, and -C1. The pyridylethylated protein after proteolytic digestion were separated from the reaction mixture by RP-HPLC on a MC-Pack ODS C 18 column (5 μm, 300 Å, 4.6 x 150 mm) using a linear gradient from 0% acetonitrile in 0.1% TFA to acetonitrile in 0.1% TFA. Each N-terminal sequence from obtained fragments is indicated as follows; P, the intact pyridylethylated protein, R; Arg-C, T; trypsin, E; Glu-C (V8 proteinase), D; Asp-N, M; cyanogen bromide. Cysteine residues in fragments are detected as a pyridylethylated cysteine.