Kunitz‐type trypsin inhibitor from durian (Durio zibethinus) employs a distinct loop for trypsin inhibition

Abstract

Kunitz‐type trypsin inhibitors are ubiquitous in plants. They have been proposed to be a part of a defense mechanism against herbivores. Trypsin inhibitors also have potential applications in the biotechnology industry, such as in mammalian cell culture. We discovered that durian (Durio zibethinus) seed contains Kunitz‐type trypsin inhibitors as identified by N‐terminal sequencing and mass spectrometry. Eleven new trypsin inhibitors were cloned. The D. zibethinus trypsin inhibitors (DzTIs) that are likely expressed in the seed were produced as recombinant proteins and tested for trypsin inhibitory activity. Their inhibitory activity and crystal structures are similar to the soybean trypsin inhibitor. Surprisingly, a crystal structure of the complex between DzTI‐4, the DzTI with the lowest inhibitory constant, and bovine trypsin revealed that DzTI‐4 utilized a novel tryptophan‐containing β1‐β2 loop to bind trypsin. Site‐direct mutagenesis confirmed the inhibitory role of this loop. DzTI‐4 was not toxic to the HEK293 cells and could be used in place of the soybean trypsin inhibitor for culturing the cells under serum‐free conditions. DzTI‐4 was not toxic to mealworms. However, a mixture of DzTIs extracted from durian seed prevented weight gain in mealworms, suggesting that multiple trypsin inhibitors are required to achieve the antinutritional effect. This study highlights the biochemical diversity of the inhibitory mechanism of Kunitz‐type trypsin inhibitors and provides clues for further application of these inhibitors.

1. INTRODUCTION

Proteinaceous protease inhibitors have various biological roles. In plants, protease inhibitors could regulate nutrient utilization during seed germination (Savelkoul et al. 1992), antagonize phytophagous insects (Haq et al. 2004; Murdock and Shade ²⁰⁰²), and defense against invasive microorganisms (Kim et al. 2009). Plant protease inhibitors are classified into families based on sequence homology, such as Kunitz‐type trypsin inhibitor (KTI), Bowman‐Birk type inhibitor, cystatin, potato‐type I inhibitor, potato‐type II inhibitor, and Mustard‐type. Most plant protease inhibitors are found in three plant families: Solanaceae, Gramineae, and Fabaceae (Clemente et al. 2019; Jamal et al. ²⁰¹³; Srikanth and Chen ²⁰¹⁶). Numerous types of plant protease inhibitor have been investigated both for basic biochemical research as well as applications in the industry.

One of the most widely studied trypsin inhibitors is the Kunitz‐type trypsin inhibitor (KTI). The founding member of the KTI is the soybean trypsin inhibitor (SBTI) that is also a standard model for KTI research. SBTI is described as a canonical inhibitor which is cleaved by trypsin but remains bound to the protease (Blow et al. 1974). KTI and related Kunitz‐type trypsin inhibitors are found in many members of the Fabaceae family as well as other plant families (Cid‐Gallegos et al. ²⁰²²). There are usually multiple trypsin inhibitor genes within each plant species which result from gene duplication and polyploidization (whole‐genome duplication) (Major and Constabel 2008; Wang et al. ²⁰¹²; Zhu et al. ²⁰¹⁹). KTI pseudogenes were also identified (Habu et al. 1997). Despite the low sequence identities among KTIs (below 50%) (Guerra et al. 2023), there are some common features. They have a β‐trefoil fold structure with two or three conserved disulfide bonds. The monomeric size is approximately 21 kDa (18–24 kDa). Oligomerization is rarely found.

Even though several KTIs have been studied in molecular details, it is difficult to predict inhibition specificity from the amino acid sequence. KTIs usually inhibit trypsin or chymotrypsin but other family members often have diverse targets. BbCI and BbKI from Bauhinia bauhinioides have high structural similarity as measured by circular dichroism (85%) and high sequence identity (83%) but these two proteins have different targets. BbCI inhibits cathepsin L, human neutrophil elastase, and cruzain, while BbKI inhibits trypsin and kallikrein (Araújo et al. 2005). Some KTIs can inhibit proteases other than serine proteases. For example, E3Ad (Guerra et al. 2016) and AtKSTI (Arnaiz et al. 2018) can inhibit aspartic and cysteine proteases, respectively. Some KTIs also inhibit a non‐protease. For example, BASI from barley could inhibit barley α‐amylase in addition to subtilisin (Mundy et al. 1984). Some KTIs could even have other activities. For example, miraculin from Richadella dulcifica can convert sour taste to sweet taste but cannot inhibit any protease activity, while miraculin‐like proteins (30–55% identity) can do vice versa (Ito et al. 2007; Ohkura et al. ²⁰¹⁸; Selvakumar et al. ²⁰¹¹). A KTI‐like protein from Brassica oleracea is a chlorophyll‐binding protein induced by drought and heat stresses (Satoh et al. 2001). Therefore, bioinformatic predictions alone is often not sufficient to annotate the function and elucidate the target‐bound structures of KTIs. Thus, investigation of new KTIs may reveal unexpected functionality.

Durians (Durio zibethinus, Malvaceae Family) is an essential economic fruit in many Southeast Asian countries. The durian pulp is highly priced while durian seeds are often discarded in household consumption as well as in durian processing industries. Studies of important components in durian seeds may enhance the value of the by‐products. We discovered that durian seed contains Kunitz‐type trypsin inhibitors (Durio zibethinus trypsin inhibitors or DzTIs). There are multiple previously reported DzTI sequences as a part of genomic and transcriptomic analyses: XP_022737266, XP_022738218, XP_022734232, XP_022733898, XP_022734070, XP_022734258, XP_022734216, XP_022734369, XP_022733892, XP_022743513, and XP_022743286 (Teh et al. 2017). However, they had not been biochemically characterized and it was not known that DzTIs were expressed in the seed. Thus, this study aims to characterize the function and structure of recombinant DzTIs. In addition, several previous reports suggested that some of the KTIs could have antinutritional effects and act as insecticides (Clemente et al. 2019; Cotabarren et al. ²⁰²⁰; Haq et al. ²⁰⁰⁴; Singh et al. ²⁰²⁰). Therefore, this study also explored the insecticidal activity of a DzTI. Moreover, a DzTI was compared with SBTI for its application in mammalian cell culture in serum‐free media. The data presented not only deepens the understanding of the basic biochemistry of KTI but will also provide a foundation for further application of DzTIs in various fields.

2. RESULTS

2.1. Protein identification, molecular cloning, and phylogenetic analysis

Phosphate‐buffered saline (PBS)‐soluble fraction of the ground durian (Durio zebethinus cultivar Chanee) seed contained a 21‐kDa protein (Figure 1a). Edman degradation revealed the first 10 amino acids to be KNEPVLD(T/I)DG (Table S1). BLASTP search suggested that the protein was a Kunitz‐type trypsin inhibitor (DzTI). The ambiguity in the 8th position in the amino acid sequence suggested that there were multiple types of DzTIs in durian seed. We next attempted to clone the DzTI genes so that they could be further studied.

FIGURE 1.

(a) SDS‐PAGE of the PBS‐soluble proteins from Chanee durian seed. The dominant KTI protein band is indicated with an arrowhead. (b) Sequence alignment of the KTI genes found in the Musang King durian that have the translated amino acid sequences in agreement with the N‐terminal sequencing results of the KTIs from Chanee durian. Primers used for subsequent cloning were indicated with arrows. The GenBank accession code was shown at the beginning of each respective sequence.

Focusing on the known DzTIs that had the amino acid sequences in agreement with the Edman degradation results (XP_022734216, XP_022734070, and XP_022733898, XP_022734258), our data revealed that the first amino acid (K) corresponds to the 27th amino acid in the translated open reading frame (Figure 1b). The nucleotide sequences of the 5′‐untranslated region (5′‐UTR), the signal peptide, and the 3′‐UTR were quite conserved, allowing a primer design that could be used to clone multiple DzTIs. Two forward primers (F1 or F2) and a reverse primer (R) were used for PCR amplification of the DzTI genes (Figure 1b). RNA extraction from the starch‐rich seed was proven difficult. Thus, we decided to clone the DzTI genes from the genomic DNA, as the genes of the previously reported DzTIs did not contain introns, and then later confirmed the DzTIs protein expression by mass spectrometry.

PCR amplification of DzTI genes from durian genomic DNA was successful and 14 DzTIs were identified (Table S2). Three out of these 14 DzTIs had identical amino acid sequences to previously known DzTIs (XP_022734258, XP_022734369, and XP_022733892). Not all of these cloned DzTIs have lysine as the predicted first amino acid residue. The amino acid sequences of previously known DzTIs and the new DzTIs cloned in this work are shown in Table S3. In total, there are 22 unique DzTIs. Phylogenetic analysis of these DzTIs is shown in Figure 2. The multiple sequence alignment and percent identity matrix are shown in Figure S1 and Table S4, respectively. For ease of name usage, we renamed these DzTIs from DzTI‐1 to DzTI‐20 according to the distance away from SBTI, with the exception of XP_022743513 and XP_022743286 which were distinct from the rest of the DzTI and were not identified either by cloning or mass spectrometry in this work. DzTIs covered a wide range of sequence identity (56%–99.5%) among each other. Comparing DzTIs to SBTI, the sequence identity was in the range of 22.8%–31.8%, which was at the borderline value generally considered to have a similar enough structure for successful crystallographic phasing by molecular replacement (Scapin 2013). Sequence identity of DzTIs was higher when compared with miraculin, the sour‐to‐sweet taste protein, with the percent sequence identity in the range of 42.1%–48.4%. A similar sequence identity range (45.5%–47.7%) was observed between DzTIs and VvMLP, a KTI with a known structure and confirmed protease inhibitory activity that had one of the highest sequence identities to DzTIs (Ohkura et al. 2018).

FIGURE 2.

Phylogenetic tree of DzTIs. The DzTIs were numbered by respective distance based on SBTI. XP022743513 and XP022743286 are found in the Musang King durian genome but could not be detected in our study of the Chanee cultivar. Miraculin and Vitis vinifera miraculin‐like protein (VvMLP), which are closely related to DzTI, were also included. The analysis was achieved by the maximum likelihood approach. Bootstrap values are indicated at each node on the tree.

To determine which DzTIs were expressed in durian seed, we mapped peptide fragments obtained from mass spectrometry of the DzTI band excised from the SDS‐PAGE gel onto the amino acid sequences of DzTIs (Table S3). Other than lysine as the first amino acid, alanine (DzTI‐2 and DzTI‐3) and threonine (DzTI‐14 to DzTI‐20) were identified as well. However, as alanine and threonine were not obvious as the first residue in the Edman degradation results, we reasoned that these DzTIs were not a major component in our durian seed samples. Combining the Edman degradation and mass spectrometry results, we selected DzTI‐4 to DzTI‐10 and DzTI‐12 for further characterization. These DzTIs had lysine as the first amino acid. Unique peptides were identified for DzTI‐4 and DzTI‐5. DzTI‐6 and DzTI‐7 pairs were not distinguishable from each other, but the peptide fragments common to only these two DzTIs were identified. The same situation applied to DzTI‐8 and DzTI‐9. DzTI‐12 was chosen because of the high sequence coverage. Among DzTI‐10, DzTI‐11, and DzTI‐13 that had similar sequence coverage did not have a unique peptide, DzTI‐10 was chosen arbitrarily for further characterization. These 8 DzTIs were then cloned into an expression vector for recombinant protein expression in Escherichia coli.

2.2. Inhibitory activity of the eight DzTIs

Eight recombinant DzTIs (DzTI‐4, DzTI‐5, DzTI‐6, DzTI‐7, DzTI‐8, DzTI‐9, DzTI‐10, and DzTI‐12) were successfully purified and they had comparable molecular weights to the DzTI purified from durian seed (Figure S2). The inhibitory activity of these DzTIs was evaluated against bovine trypsin using Nα‐Benzoyl‐DL‐arginine 4‐nitroanilide (BApNA) as the substrate. SBTI was used as a positive control. The results suggested that all eight DzTIs were competitive inhibitors of bovine trypsin (Figure S3). Because, DzTIs appeared to be tight‐binding inhibitors, like SBTI, kinetics data were fitted to the Morrison equation (Morrison 1969; Williams and Morrison ¹⁹⁷⁹) to obtain the inhibitory constant (K _i ) values (Figure 3). DzTI‐4 and DzTI‐6 appeared to have the lowest K _i values of 5.2 and 5.4 nM, respectively. Other DzTIs had K _i values in the range of 8.9–56.6 nM. SBTI could inhibit the bovine trypsin activity with the K _i value of 1.7 nM which was more efficient than DzTI‐4 and DzTI‐6. The difference in the K _i values could simply be due to the difference in the amino acids in the inhibitory loop. SBTI and DzTIs may also have different inhibitory loop conformations or other different inhibitory mechanisms towards bovine trypsin inhibition, which could be further explored by structural analyses. DzTI‐4 was chosen for further biochemical characterization because it had the highest affinity towards bovine trypsin and the best recombinant protein production yield of around 20 mg/L of Escherichia coli culture.

FIGURE 3.

Inhibition curves of bovine trypsin by DzTIs. The solid lines represent fits to the Morrison equation for tight‐binding inhibitors. The experiments were performed in triplicates.

2.3. Enzyme specificity of DzTI‐4

As mentioned earlier, KTI may inhibit enzymes other than trypsin. We tested the inhibitory activity of DzTI‐4 towards various proteases by determining the IC₅₀ values using azocasein as a common substrate (Figure 4). The types of proteases explored were serine proteases (trypsin, chymotrypsin, elastase, and subtilisin), cysteine proteases (papain and bromelain), an aspartic protease (pepsin), and a metalloprotease (carboxypeptidase A). As expected, DzTI‐4 could inhibit trypsin at the IC₅₀ value of 3.5 μM. DzTI‐4 also inhibited chymotrypsin with a similar IC₅₀ value of 4.8 μM. Carboxypeptidase A could be inhibited, but not completely even at 100 μM of DzTI‐4. The IC₅₀ value of DzTI‐4 towards carboxypeptidase A was estimated to be 67.1 μM, which was over an order of magnitude higher than trypsin and chymotrypsin. Given the incomplete inhibition curve, the estimated IC₅₀ value for carboxypeptidase A was likely inaccurate. Other proteases were not inhibited by DzTI‐4. Because a certain KTI, such as BASI, had been shown to inhibit amylases (Vallée et al. 1998), inhibition of α‐ and β‐amylases were also examined using potato starch as a substrate and detection of the reducing sugar released using the DNS assay. DzTI‐4 did not inhibit these amylases.

FIGURE 4.

DzTI‐4 specificity towards proteases and amylases. The solid lines represent fits to the IC₅₀ equation. The experiments were performed in triplicates.

2.4. Crystal structure of DzTIs

To explore the structure and function of DzTIs, we first determined the structures of individual DzTIs. All DzTIs selected for inhibition studies could be crystallized. The crystallization conditions, cryoprotection buffer, and phasing strategies are summarized in Table S5. Data collection and refinement statistics are shown in Table S6. All DzTIs contained β‐trefoil fold with connecting loops and three disulfide bonds, as expected for KTIs (Figure 5a). The β‐strand numbering is indicated on the structure of DzTI‐4 (Figure 5b). For most DzTIs, the residue 207 onwards did not exhibit a secondary structure or were not visible in the electron density map. However, a C‐terminal helix was observed for DzTI‐4 and DzTI‐5. Although there were minor variations in the connecting loops of the β‐trefoil core structure, the three‐dimensional structures of all these DzTIs were highly conserved (Figure 5c) with the RMSD values in the range of 0.115–0.406 Å (Table S7).

FIGURE 5.

Structures of DzTIs. (a) Crystal structures of the eight DzTIs determined in this study: DzTI‐4, DzTI‐5, DzTI‐6, DzTI‐7, DzTI‐8, DzTI‐9, DzTI‐10, and DzTI‐12. The DzTIs were shown in the same orientation. (b) The structure of DzTI‐4 as a representative DzTI structure showing the β‐trefoil strand names. (c) Superimposition of the eight DzTIs.

2.5. Structure of the DzTI‐4 in complex with trypsin

To understand the mechanism of inhibition, we crystallized and determined the structure of the most potent inhibitor DzTI‐4 in complex with bovine trypsin to 2.00 Å resolution (Tables S5 and S6). There were two identical DzTI‐4‐trypsin pairs in the asymmetric unit. From the sequence alignment between DzTI‐4 and SBTI (Figure S1), the residue in DzTI‐4 that was homologous with the inhibitory arginine of SBTI was K94 in the β4‐β5 loop (Song and Suh 1998). Since the substitution of arginine with lysine is generally considered conserved, we initially expected the K94‐containing β4‐β5 loop of DzTI‐4 to embed into the trypsin active site with the K94 residue inserted into the trypsin specificity (S1) pocket. Surprisingly, our crystal structure revealed that a tryptophan (W52)‐containing β1‐β2 loop, on the opposite side of the β4‐β5 loop on DzTI‐4, interacted with the trypsin active site instead (Figure 6a). This interaction was observed in both DzTI‐4‐trypsin pairs in the asymmetric unit. The electron density of the β1‐β2 loop did not appear broken (Figure 6b). The Cα of the G53 residue was the closest atom (3.0 Å) of DzTI‐4 to the catalytic S195 of trypsin. Even by visual inspection, the adjacent peptide bonds were not in the plausible trajectories for nucleophilic attack by the S195. The carbonyl groups of the W52 and G53 residue were 5.1 and 4.0 Å away from the catalytic serine. The (θ _x , θ _y ) values for the carbonyl groups of W52 and G53 were (22.7°, 99.6°) and (45.7°, 120.7°), respectively. These values deviated significantly from the (90°, 90°) reported for most inhibitors that undergo cleavage (Radisky et al. 2006; Radisky and Koshland Jr. ²⁰⁰²). Therefore, trypsin was unlikely to cleave the β1‐β2 containing loop of DzTI‐4. W52 was embedded into the S1 pocket but did not reach the D189 that normally interacted with the lysine or arginine residues of the trypsin substrate or SBTI. The backbone NH of W52 also formed a hydrogen bond with the backbone carbonyl group of G216 of trypsin.

FIGURE 6.

Crystal structure of the complex between DzTI‐4 (green) and bovine trypsin (pink). (a) Overall structure of the complex. (b) The β1‐β2 inhibitory loop of DzTI‐4 with the W52 residue burying into the S1 pocket of trypsin. The mesh represents the 2mFo‐DFc electron density map at 1.5σ. (c) The interactions between DzTI‐4 and trypsin. Dash lines indicate potential interactions. Numbers indicate distances in angstrom (Å). Residues colored in marine blue are the residues that form the oxyanion hole. (d) Superposition at the KTI structures between the DzTI‐4‐trypsin and the API‐A‐trypsin complexes.

In addition to the W52 residue, there were extensive interactions between DzTI‐4 and bovine trypsin (Figure 6c). The backbone carbonyl group of A50 in DzTI‐4 interacted with the side chain of S217 of trypsin. The backbone NH group of A54 in DzTI‐4 also formed a hydrogen bond with the backbone carbonyl group of S214 in trypsin. Residues in regions outside of the β1‐β2 loop also interacted with bovine trypsin. The backbone carbonyl group of N161 of DzTI‐4 formed a hydrogen bond with K224 of trypsin. R185 interacted with the backbone carbonyl group of A54 in the β1‐β2 loop and also formed a hydrogen bond with the backbone carbonyl group of S96 in trypsin. The residue D79 of DzTI‐4 not only formed a hydrogen bond with the backbone NH group of G53 in the β1‐β2 loop of DzTI‐4 but also interacted with the backbone NH group of G193 in trypsin that was a part of the loop that form the oxyanion hole. The side chain of D79 of DzTI‐4 was 3.9 Å away from the catalytic S195 of trypsin, thus there might be a weak interaction.

We next compared our DzTI‐4‐trypsin complex structure to the structure of API‐A (Bao et al. 2009), a double‐headed arrowhead protease inhibitor, in complex with trypsin (Figure 6d). The RMSD value between DzTI‐4 and API‐A was 2.050 Å with good alignment of the β‐trefoil core structures. API‐A bound two trypsin molecules with the β5‐β6 and β9‐β10 loops, whereas our DzTI‐4 can bind trypsin using the β1‐β2 loop as shown earlier. Superposition between our DzTI‐4‐trypsin complex structure and the API‐A‐trypsin complex structure revealed that the DzTI‐4‐bound trypsin was in a distinct position that did not overlap with the two API‐A‐bound trypsin molecules.

We then examine the structural differences of trypsin and DzTI‐4 upon complex formation. Superposition of the structure of trypsin alone (PDB ID 1S0Q) and the trypsin in complex with DzTI‐4 gave the RMSD value of 0.279 Å with no noticeable differences in the structure. Superimposition between the structures of DzTI‐4 and DzTI‐4 in complex with bovine trypsin did not reveal any appreciable differences (RMSD = 0.149 Å). The conformation of the peptide backbone of the β1‐β2 loop did not change significantly (Figure 7a). There was a slight rotation of the W52 side chain upon binding to trypsin. These observations suggested that the β1‐β2 loop was not flexible. There were numerous interactions between other parts of DzTI‐4 and the β1‐β2 loop that could stabilize the loop conformation in addition to D79 mentioned above (Figure 7b). The side chain of Q75 formed a hydrogen bond with the backbone carbonyl of G55. The side chain of Q75 also formed a hydrogen bond with the side chain of R185, which in turn stabilized the β1‐β2 loop via an interaction with the backbone carbonyl of A54. The indole NH of W201 formed a hydrogen bond with the side chain of S49. The carbonyl group of P202 interacted with the backbone carbonyl of A50. There were interactions within the β1‐β2 loop as well. There were peptide bond interactions between G57 and S49, G53 and G56, and W52 and G55. Therefore, the extensive intramolecular interactions of DzTI‐4 stabilized the conformation of the β1‐β2 loop.

FIGURE 7.

Comparison of inhibitory loops among KTIs. (a) Superposition of the β1‐β2 loop of DzTI‐4 between the unbound structure (light blue) and in complex with bovine trypsin (pale green). (b) The structure of the β1‐β2 loop showing intramolecular interactions within DzTI‐4. The pale green residues are the β1‐β2 loop. The rest of DzTI‐4 is shown in gray. The numbers indicate distances in Ångstrom. (c) The inhibitory loops of KTIs: β4‐β5 loops of SBTI (PDB ID 1AVW), BbKI (PDB ID 6DWH), EcTI (PDB ID 4J2Y), and TKI (PDB ID 4AN7), and the β9‐β10 loop of API‐A (PDB ID 3E8L). The dash lines indicate interactions between the respective spacer residues with the inhibitory loops. (d) The backbone of the β4‐β5 loop and the residue A41 of DzTI‐4 (green) superimpose onto the backbone of the inhibitory loops of the KTIs in (c) (gray). (e) Comparison of the Φ and Ψ torsion angles from residues P4 to P4′ of the β1‐β2 and β4‐β5 loops of DzTI‐4 to the inhibitory loops of other KTIs.

Because the β4‐β5 loop of DzTI‐4 did not appear to be involved in trypsin inhibition, we compared this loop to inhibitory loops of other KTIs. The loops chosen included the β4‐β5 loops of SBTI (Song and Suh 1998), BbKI (Li et al. 2019), EcTI (Zhou et al. 2013), and TKI (Patil et al. 2012), and the β9‐β10 of API‐A (Bao et al. 2009) (Figure 7c). The inhibitory loop of these KTIs contained a conserved asparagine as a spacer residue that prevent conformational change after acylation and promote re‐ligation of the scissile bond (Dasgupta et al. 2006). For API‐A, the spacer residue was an arginine residue. However, the β4‐β5 loop of DzTI‐4 did not have an amino acid side chain to interact with the peptide backbone at the position P1′ and P2 because the spacer position is an alanine residue (A41) instead (Figure 7d). The conformation of the β4‐β5 loop of DzTI‐4 was also distinct from other KTIs. Torsion angle analysis revealed that the structure of the β1‐β2 loop of DzTI‐4 did not change significantly after binding to trypsin (Figure 7e and Table S8). The torsion angle pattern of the β1‐β2 loop of DzTI‐4 was not similar to the reactive loop of other KTIs, even when compared to the inhibitory loops (β9‐β10 and β5‐β6) of API‐A. The β4‐β5 loop of DzTI‐4 had a similar Φ angle pattern with other KTIs, but a distinct Ψ angle pattern in the P3‐P1 region, which is an important region for the canonical inhibition mechanism (Guerra et al. 2023).

Because the inhibition of trypsin by the β1‐β2 loop of DzTI‐4 was unexpected, we sought further evidence to confirm the validity of the structure. Therefore, we performed site‐directed mutagenesis on key residues of DzTI‐4 and examined their effects on trypsin inhibition. The W52A mutation was performed because the W52 residue was buried into the active site of bovine trypsin. The D79A mutation was made because of its interaction with the β1‐β2 loop and the G193 that formed the oxyanion hole in trypsin. The K94A mutation was created because it is a homologous residue of the inhibitory arginine in SBTI. All these mutated DzTI‐4 variants were expressed and purified then tested for inhibition of bovine trypsin along with the wild‐type DzTI as a control using BApNA as the substrate (Figure 8). The K94A mutation did not drastically affect the inhibitory activity compared to the wild‐type DzTI‐4 with the IC₅₀ value of 81.5 and 70.0 nM, respectively. However, the W52A and D79A mutations caused a significant increase in the IC₅₀ values to 331.0 and 274.3 nM, respectively. These results agreed with our DzTI‐4‐trypsin complex crystal structure that K94 was not involved in trypsin inhibition. The W52‐containing β1‐β2 loop and surrounding residues indeed bound and inhibited trypsin.

FIGURE 8.

The effect of DzTI‐4 site‐directed mutagenesis on trypsin inhibition. The IC₅₀ values are indicated behind the name of each variant. The experiments were performed in triplicates.

2.6. Stability of DzTI‐4

For future utilization of DzTI‐4, we explored the effect of various physical or chemical factors on the function of DzTI‐4. DzTI‐4 was treated with the various factors before adding the treated DzTI‐4 to the reaction mixture. Considering the metal ions, reduction of DzTI‐4 inhibitory activity was observed with Fe³⁺ and Zn²⁺ (Figure S4a). DzTI‐4 was stable in the pH 2–9 range (Figure S4b). Reducing agents did not noticeably affect DzTI‐4 inhibitory activity, except for the slight reduction of the inhibitory activity at 10 mM DTT (Figure S4c). DzTI‐4 was able to tolerate various organic solvents, including ethanol, acetone, acetonitrile, DMSO, and DMF, even up to 50% (Figure S4d). With regards to thermal stability, DzTI‐4 could tolerate the temperature of 70°C for at least 1 h (Figure S4e). At 85°C, DzTI‐4 was irreversibly damaged in 10 min or less.

2.7. Inhibition of trypsin in mammalian cell culture by DzTI‐4

Because SBTI is employed in animal serum‐free cell culture to inhibit trypsin used in cell dissociation and dislodgement, we examined whether DzTI‐4 could be employed in such an application as well. The HEK293 cells were employed as a model of an adherent cell. After trypsinization, the HEK293 was cultured on the serum‐free medium CDM4HEK293 with SBTI, DzTI‐4, or without trypsin inhibitor. Initially, the cells were rounded and did not yet attach to the culture vessel surface. After 3 hours, cells that were cultured in the medium with SBTI or DzTI‐4 began to attach to the surface. After overnight incubation, dose‐dependent cell attachment was observed visually (Figure 9a), while the cells that were not treated with trypsin inhibitors were not attached. MTT assay was then performed to measure the amount of viable attached cells (Figure 9b). Exponential plateau analysis revealed that DzTI‐4 could inhibit trypsin in the mammalian cell culture system similarly to SBTI with the estimated maximum absorbance value from the MTT assay at 0.5559 and 0.5774, respectively.

FIGURE 9.

The effect of DzTI‐4 and SBTI on HEK293 attachment and viability after cell trypsinization. (a) HEK293 morphology after culturing overnight with various concentrations (% w/v) of trypsin inhibitors. The black bar at the bottom‐right of each figure indicates the 200 μm length. (b) The MTT assay for viable cells after trypsinization at various concentrations of DzTI‐4 and SBTI.

2.8. Antinutritional effect of DzTI‐4 against mealworms

Several investigators have hypothesized that plant protease inhibitors could inhibit food digestion in insects, thus suppressing growth or killing invasive insects (Bolter and Jongsma 1997; Bonturi et al. ²⁰²²; Koiwa et al. ¹⁹⁹⁷; Singh et al. ²⁰²⁰). Because DzTI‐4 could inhibit trypsin and chymotrypsin, which are proteases found in the gastrointestinal tract of animals, we wonder if DzTI‐4 might have antinutritional effect in insects. Therefore, we used mealworms as an insect larvae model for feeding experiments with SBTI, DzTI‐4, as well as the mixture of DzTIs extracted from durian seed. Ten mealworms were fed with 0.3 g of an artificial diet supplemented with protease inhibitors. The diet was replaced every 5 days for 1 month. In terms of the survival rate (Figure 10a), the lethal concentration 50 (LC₅₀) value could not be calculated because the mealworms had high survival rates (>80%) during the month‐long experiment, even when treated by the highest concentrations of DzTI‐4, SBTI, or seed extracted DzTI. Using the highest amount of inhibitors (750 μg inhibitor per 0.3 g diet), we next examined the weight of the larvae and compared the linear rate, represented as ln(mean weight at each time point) divided by the ln(mean weight at the start of the experiment) for 30 days (Figure 10b). Similar weight accumulation rates were observed for the no inhibitor control (0.0012/day), DzTI‐4 (0.0010/day), and SBTI (0.0007/day) groups. However, the larvae treated with the seed‐derived DzTIs showed a very low weight accumulation rate (0.0001/day).

FIGURE 10.

Biological activity of SBTI, DzTI‐4, and seed‐derived DzTIs against mealworm larvae. (a) The survival rate of mealworms when fed with an artificial diet supplemented with different trypsin inhibitors (mg inhibitor per 0.3 g artificial diet). (b) The change in mealworm weight during 1 month when treated with different trypsin inhibitors. The increasing weight is displayed as the ln(weight) at different time points relative to the ln(weight) at the start.

3. DISCUSSION

KTIs are commonly found in plants. Our work is the first to report and characterize KTIs in durian (DzTIs) which is the most abundant soluble protein in the seeds. Fourteen new DzTI nucleotide sequences were cloned. Eleven of which coded for previously undescribed KTIs. Because we did not sequence the entire genome of the Chanee durian, it is possible that there might be many more DzTIs to be discovered.

The amino acid sequences of the DzTIs are more closely related to miraculin than SBTI. However, the eight DzTIs tested could still inhibit the trypsin activity. The Vitis vinifera miraculin‐like protein (VvMLP), which has about the same amino acid sequence identity as DzTIs compared to miraculin, is also an active protease inhibitor (Ohkura et al. 2018). Thus, prediction of the biological activity of KTIs and related proteins may not currently be able to rely on the examination of the amino acid sequence alone. Because there are at least 20 DzTIs, it is not clear whether each of them has unique properties or they may all have redundant functions. Regardless, the DzTIs will need to be individually characterized experimentally. Previous reports in other plants suggested that the ubiquitous trypsin inhibitors may play protective roles against insects (Cotabarren et al. 2020; Richardson ¹⁹⁷⁷). It is also possible that these KTIs could protect the seeds from proteases when they pass through the digestive tract of seed dispersers (Cochrane 2003; Traveset ¹⁹⁹⁸; Traveset et al. ²⁰⁰⁸). Therefore, the abundance and diversity of DzTIs might serve various biological functions.

We limited our investigation to eight DzTIs that contain lysine as the N‐terminal residues, which agree with the N‐terminal sequencing results, and can be detected by mass spectrometry. Other types of DzTIs contain threonine or alanine as the N‐terminal residues that could also be detected by cloning and mass spectrometry. The eight chosen DzTIs, including DzTI‐4, DzTI‐5, DzTI‐6, DzTI‐7, DzTI‐8, DzTI‐9, DzTI‐10, and DzTI‐12, are evaluated for their inhibitory activity against bovine trypsin. DzTIs could be grouped into two groups according to their K _i values. The low K _i group (DzTI‐4, DzTI‐5, DzTI‐6, and DzTI‐7) has the K _i values in the range of 5.2–13.5 nM. The high K _i group (DzTI‐8, DzTI‐9, DzTI‐10, and DzTI‐12) has the K _i values in the 43.3–56.6 nM range. The amino acid sequences of the β1‐β2 loops of these DzTIs are identical (Figure S1) except that the low K _i group has glycine or alanine at the 54th position while the high K _i group has proline at this position. Because the NH of the peptide bond between the 53^rd and 54^th residue forms a hydrogen bond with the trypsin (Figure 6), having a proline at this position will result in a loss of a hydrogen bond, thus lowering the binding affinity for the high K _i group. Proline might also add a steric hindrance as well. As we do not yet have the structures of all these DzTIs in complex with trypsin, there might be other differences in the binding interface that contributed to the variation in K _i values. Overall, these DzTIs appeared to be worse inhibitors than SBTI (K _i  = 1.7 nM). We attribute the affinity difference, in part, to the utilization of a different binding loop, at least for DzTI‐4.

DzTI‐4 utilizes the tryptophan‐containing β1‐β2 loop to inhibit trypsin in contrast to the SBTI that uses the arginine‐containing β4‐β5 loop (Blow et al. 1974; Song and Suh ¹⁹⁹⁸). Other KTIs, such as API‐A, are known to use loops other than β4‐β5 to inhibit trypsin. API‐A could use both the β5‐β6 and β9‐β10 loops for trypsin inhibition (Bao et al. 2009). Therefore, inhibition of trypsin by the β1‐β2 loop of DzTI‐4 is a novel mechanism. Our DzTI‐4‐trypsin complex structure suggests no cleavage of the peptide bond in the β1‐β2 loop of DzTI‐4 due to the suboptimal trajectory for nucleophilic attack by the catalytic serine of trypsin. No significant conformational change was observed for both DzTI‐4 and trypsin, thus they bound in the lock and key fashion. The binding mechanism that utilizes the unexpected W52‐containing β1‐β2 loop also highlights the importance of experimental structure determination in investigating the inhibition mechanism of KTIs. Amino acid sequence alignment could not have predicted such an interaction site.

The β1‐β2 loop is stabilized by several interactions within the loop and with other residues within DzTI‐4, like in other protease inhibitors (Radisky et al. 2005). However, there is no spacer residue and nearby disulfide bond like in other reactive inhibitory loops of other KTIs. β4‐β5 loop does not appear to be a trypsin inhibitory loop. The conformation at the base of the loop is comparable to other reactive inhibitory loops of other KTIs, but the torsion angle differences in the P3‐P1 region make the β4‐β5 loop not inhibitory. The lack of the side chain of the spacer residue also renders the loop inactive, consistent with previous mutational analysis (Dasgupta et al. 2006). In addition to the conserved asparagine in the β4‐β5 loop of the canonical KTIs, glutamine and arginine could serve as a spacer residue as well (Bao et al. 2009; Dasgupta et al. ²⁰⁰⁶; Radisky et al. ²⁰⁰⁵). It would be intriguing to determine whether the replacement of the A41 residue of DzTI‐4 with asparagine, glutamine, or arginine would restore the inhibitory activity of the β4‐β5 loop. The 41st residues of all known DzTIs are either alanine or threonine (Figure S1) that have short side chains. Therefore, they may not function well as spacer residues. In addition, the structures of all the DzTIs reported here (Figure 5) are virtually identical. Thus, it is possible that the β4‐β5 loops of all these DzTIs are not inhibitory.

Our results provided more evidence of the diversity of loops that KTIs can use for enzyme inhibition. Therefore, there are many loops on KTIs that could be engineered to create new protein interaction sites. Moreover, superposition of our DzTI‐4‐trypsin and API‐A‐trypsin complex structures revealed no steric clashes among the three trypsin molecules. Thus, our results suggested that it is possible to engineer a three‐headed inhibitor. Intriguingly, there is no reason why three‐headed inhibitors could not exist in nature.

Mutation of the W52 significantly reduces the inhibitory activity DzTI‐4 towards trypsin, but does not completely render DzTI‐4 inactive. It is possible that the W52A mutation does not alter the interaction interface enough to result in a complete loss of binding. However, we cannot eliminate the possibility that DzTI‐4 might be able to use other loops to inhibit trypsin as well, like the case of API‐A. Because DzTI‐4 could also inhibit chymotrypsin that prefers aromatic residues in the P1 site of the substrate (Ma et al. 2005), it is possible that DzTI‐4 also uses the W52‐containing β1‐β2 loop for chymotrypsin inhibition.

DzTI‐4 could be produced at the yield of around 20 mg per liter of E. coli culture, which is sufficient for most small‐scale applications. DzTI‐4 could tolerate different metal ions. DzTI‐4 retained activity after exposure to a wide range of pH, like other KTI, which suggested that it passes through the animal digestive tract while retaining its inhibitory activity. DzTI‐4 is also resistant to reducing agents and organic solvents, which could be attributed to the multiple disulfide bonds (Betz 1993; Pallavi and Rajender ²⁰²¹) in the structure of DzTI‐4. DzTI‐4 could tolerate exposure to the temperature of 70°C for at least 1 h. The stability information will be useful for extraction and purification of DzTIs from natural sources (Batista et al. 1996; Bhattacharyya et al. ²⁰⁰⁶; Song et al. ²⁰¹⁶; Xu et al. ²⁰¹⁹) as well as further biotechnological application of DzTIs.

Like SBTI, DzTI‐4 could be utilized in animal serum‐free cell culture for cell recovery after trypsinization or protection of recombinant proteins from proteolytic degradation. DzTI‐4 is not toxic to the HEK293 cell, which is a common host for recombinant protein production. The application of DzTI‐4 as an insecticide is less promising. DzTI‐4 neither has insecticidal activity nor resulted in weight loss in mealworm larvae. However, the mixture of DzTIs extracted from the durian seeds could inhibit weight gain in the larvae. Therefore, multiple types of DzTIs, which may have various inhibition targets, might be required to produce biological effects (Harsulkar et al. 1999; Lopes et al. ²⁰⁰⁴). Consequently, an application of DzTIs as insecticides might require a cocktail of DzTIs.

In conclusion, durian seed has multiple types of the Kunitz‐type protease inhibitor DzTIs. Of the eight selected DzTIs for characterization, DzTI‐4 has the best inhibitory activity against trypsin. DzTI‐4 inhibits trypsin using the β1‐β2 loop which has not been observed before in KTIs. This demonstrates the versatility of the β‐trefoil scaffold both in protein evolution and engineering. DzTI‐4 could inhibit chymotrypsin as well. DzTI‐4 is stable after exposure to various physical and chemical factors, making it suitable for biotechnological applications. DzTI‐4 is not toxic to the HEK293 cells and could be used in place of SBTI to inhibit trypsin in cell culture. DzTI‐4 alone did not have pronounced effect on mealworm larvae mortality or weight gain. However, the seed‐derived DzTI mixture could prevent weight gain in mealworm larvae, suggesting that multiple DzTIs are required for biological activity. Overall, DzTI is a rich source to explore enzyme target specificity and inhibition mechanism, as well as potential biotechnological applications.

4. MATERIALS AND METHODS

4.1. N‐terminal sequencing and cloning of DzTI genes

Chanee durian seeds were pounded into a paste with a mortar and pestle. The paste was then dissolved in PBS to 5% w/v and clarified by centrifugation. The supernatant was then analyzed by SDS‐PAGE. For N‐terminal sequencing, protein in SDS‐PAGE gel was electro‐blotted onto a PVDF membrane and stained with Coomassie Blue. After destaining and briefly rinsing with water, the PVDF membrane was air dried overnight. The 21‐kDa band was excised and sent for N‐terminal sequencing (The Tufts University Core Facility, Tufts Medical School, Boston, MA). The sequence obtained from N‐terminal sequencing (KNEPVLD) was searched against the Musang King durian genome using BLASTP. The sequences surrounding the candidate genes were aligned and primers targeting the conserved region were designed.

Experiments in this work were approved by the Institutional Biosafety Committee of Chulalongkorn University (Approval No. SC CU‐IBC‐008/2022). A sample of the genomic DNA from the leaf of Chanee durian (a gift from Pinnapat Pinsorn and Supaart Sirikantaramas) (Pinsorn et al. 2024) was amplified with the primers F1 (5′‐AAAACAACACATATTGATCAACGTATTAACC‐3′) and R (5′‐TTTTGGCAGTACAGTAGTTCATATAAAAACTACTGTACTTA‐3′) as well as primers F2 (5′‐CATATTTCTTTGGGGTAGCCAACGCTAAA‐3′) and R. The PCR product was then phosphorylated and cloned into pNEB193 (New England Biolabs) that had been cut with SmaI and dephosphorylated. Standard blue‐white screening was used to screen for recombinant plasmids. The plasmids were then sequenced by Sanger sequencing with M13 primers.

4.2. Phylogenetic tree

The sequences obtained from the cloning mentioned above were analyzed by multiple sequence alignment and phylogenetic tree using the maximum likelihood approach by the MEGA‐X software (Kumar et al. 2018). Soybean trypsin inhibitor (SBTI; NCBI accession number: NP_001238611), miraculin (NCBI accession number: P13087), and Vitis vinifera miraculin‐like protein (PDB: 5YH4) were also included as the most related protein. Signal peptides of all sequences were excluded.

4.3. Mass spectrometry

For enzymatic “in gel” digestion, Coomassie stained gel slices were completely de‐stained in MeOH/H₂O/NH₄HCO₃ (50%/50%/100 mM), dehydrated for 5 min in ACN/H₂O/NH₄HCO₃ (50%/50%/25 mM) then for 30 s in 100% ACN. After drying in a Speed‐Vac for 1 min, the gel slice was reduced in a 25 mM DTT solution (Dithiotreitol in 25 mM NH₄HCO₃) for 15 min at 56°C, alkylated with 55 mM CAA (Chloroacetamide in 25 mM NH₄HCO₃) at room temperature in darkness for 15 min, washed once in H₂O, dehydrated for 2 min in ACN/H₂O/NH₄HCO₃ (50%:50%:25 mM) then in 100% ACN for 30 s. The gel slice was then dried and rehydrated with 20 μL of trypsin solution containing 0.01% ProteaseMAX™ surfactant (10 ng/μL Trypsin from Promega Corp. in 25 mM NH₄HCO₃/0.01% w/v of ProteaseMAX™ from Promega Corp.). The gel slice was left for 2 min at room temperature then, to keep gel pieces immersed throughout the digestion, an additional 30 μL of the digestion solution (25 mM NH₄HCO₃/0.01% w/v of ProteaseMAX™) was added. The digestion was conducted at 42°C for 3 h. Peptides resulted from the digestion were transferred into a new tube and acidified with 2.5% TFA (trifluoroacetic Acid) to the final concentration of 0.3%. Gel slices were extracted further with ACN:H₂O:TFA (70%:29.25%:0.75%) for 10 min with vortexing. The solutions were then combined and completely dried in a Speed‐Vac. The dried peptides were solubilized in 30 μL of 0.05% TFA. ProteaseMAX™ was removed by centrifugation at maximum speed for 10 min. The peptides were solid phase extracted with the ZipTip® C18 pipette tips (MilliporeSigma) according to manufacturer protocol. Peptides were eluted off the C18 SPE column with 5 μL of acetonitrile/H₂O/TFA (60%:40%:0.1%) dried to completion then resolubilized in 30 μL total volume with 0.1% formic acid and 2 μL was loaded on the instrument.

Peptides were analyzed by nanoLC‐MS/MS with the Agilent 1100 nanoflow system (Agilent) that is connected to a hybrid linear ion trap‐orbitrap mass spectrometer (LTQ‐Orbitrap Elite™, Thermo Fisher Scientific) equipped with an EASY‐Spray™ electrospray source at constant 35°C. Chromatographic separation of peptides prior to mass spectral analysis was performed using a capillary emitter column (PepMap® C18, 3 μM, 100 Å, 150 × 0.075 mm, Thermo Fisher Scientific) onto which 2 μL of the extracted peptides was applied. NanoHPLC system utilized solvent A (0.1% (v/v) formic acid) and solvent B (99.9% (v/v) acetonitrile, 0.1% (v/v) formic acid) at 0.50 μL/min to load the peptides over a 30‐min period and 0.3 μL/min to elute peptides directly into the nano‐electrospray using a gradual gradient from 0% (v/v) B to 30% (v/v) B over 80 min then finished with a 5‐min fast gradient from 30% (v/v) B to 50% (v/v) B at which time a 4‐min flash‐out from 50% to 95% (v/v) B was performed. An overall run time of 150 min covered column conditioning at 95% B for 1 min and equilibration at 100% A for 30 min. As the peptides came out from the HPLC‐column/electrospray source, survey mass spectrometry scans in the Orbitrap were obtained with a resolution of 120,000. CID‐type MS/MS was then performed with 2.0 AMU isolation and 10 ms activation time with normalized collision energy fragmentation (35%) of 30 most intense peptides detected in the MS1 scan from 350 to 1800 m/z. Redundancy was limited by dynamic exclusion. Monoisotopic precursor selection and charge state screening were enabled, +1 and undefined charge states were rejected.

Raw MS/MS data were converted to the mgf file format using MSConvert (ProteoWizard: Open Source Software for Rapid Proteomics Tools Development) for subsequent analysis. Resulting mgf files were used to search against Durio zibethinus (Durian) Uniprot reference proteome database (UP000515121; 53026 total entries, 11/05/2021) containing user defined cloned sequences along with a list of common contaminants (172 total entries) using in‐house Mascot search engine 2.7.0 (Matrix Science) with fixed cysteine carbamidomethylation and variable methionine oxidation and asparagine‐glutamine deamidation. Peptide mass tolerance and fragment mass were set at 10 ppm and 0.6 Da, respectively. The Scaffold software (version 4.11.0, Proteome Software Inc., Portland, OR) was used for protein annotations, significance of identification, and spectral based quantification. Peptide identifications were accepted if they could be established at higher than 99.0% probability for an FDR less than 1.0% by the Scaffold Local FDR algorithm. Protein identifications were accepted if they could be established at higher than 23.0% probability for an FDR less than 1.0% with at least 1 identified peptide. The Protein Prophet algorithm (Nesvizhskii et al. 2003) was used to assign the protein probabilities. Proteins that contained similar peptides and could not be distinguished based on the MS/MS analysis were grouped together to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were clustered together.

4.4. Protein expression and purification

The DzTI genes were PCR amplified from their respective pNEB193 plasmids. For DzTI‐6, the primers A (5′‐CTGGGTACCGAAAACCTGTATTTTCAGGGTAGCAAAAATGAGCCTGTGCTTGACACTG‐3′) and B (5′‐CGGCGGCCGCTCATTAGTGCTTCGCATTAACAACTTGTTTG‐3′) were used. For DzTI‐7, the primers A and C (5′‐CGGCGGCCGCTCATTAGTTCTTCGCATTAACAACTTGTTTG‐3′) were used. For DzTI‐9, the primers D (5′‐CTGGGTACCGAAAACCTGTATTTTCAGGGTAGCAAAAATGAGCCTGTGCTTGACATTG‐3′), and E (5′‐CGGCGGCCGCTCATTAGTTCTTCGCATAAACAACTTGTTTG‐3′) were used. For DzTI‐10, the primers D and B were used. For DzTI‐12, the primers D and C were used. The PCR products were cloned into the KpnI and NotI sites of pET‐32b (Novagen). The resulting expression constructs contained a thioredoxin tag, a hexahistidine tag, and a tobacco etch virus (TEV) protease cleavage site at the N‐terminus of the DzTI coding sequences. For the genes that clones were not obtained from Chanee durian but their presence was evident by mass spectrometry (DzTI‐4, DzTI‐5, and DzTI‐8), the expression constructs in pET‐32b were obtained by gene synthesis (Biomatik, Canada).

For protein expression, the expression plasmids were individually transformed into SHuffle® T7 competent Escherichia coli (New England Biolabs) using the heat shock method. Inoculum of the E. coli was seeded into terrific broth at a concentration of 2% and grown at 30°C. Isopropyl β‐d‐1‐thiogalactopyranoside (Gold Biotechnology) (IPTG, 0.05 mM) was added when the OD600 reached 0.6 to induce protein expression. Protein expression was continued at 16°C for 16–18 h. The cells were then pelleted and subsequently lysed in PBS containing 50 mM imidazole by sonication. The lysate was clarified by centrifugation at 40,000g for 30 min. The recombinant protein in the supernatant was then purified by Ni‐NTA resin using the imidazole gradient from 50 to 250 mM. Fractions containing pure protein, as examined by SDS‐PAGE, were pooled and TEV protease was added (Kapust et al. 2001). The mixture was dialyzed against PBS that contained 3 mM reduced glutathione and 0.3 mM oxidized glutathione. The cleaved protein was applied to a Ni‐NTA column (Gold Biotechnology). The flow‐through was subsequently purified by Q Sepharose Fast Flow (GE Healthcare) anion exchange chromatography. The purified protein was dialyzed in 20 mM Tris pH 8.0. The protein was concentrated to at least 10 mg/mL using the Microsep Advance Centrifugal Device (3 K MWCO Omega membrane, Pall Life Sciences). The concentrated protein solutions were snap frozen with liquid nitrogen and kept at −80°C until further use.

The native DzTI from durian seed (Chanee cultivar) was obtained by crushing the durian seeds with seed coat removed. The paste was stirred in cold PBS at 5% w/v overnight. The mixture was clarified by centrifugation at 6,000g and the cleared lysate was further sonicated to reduce its viscosity. Solid ammonium sulfate was added to the lysate to achieve 50% saturation. The mixture was centrifuged at 40,000g to collect the precipitated protein. The protein pellet was redissolved in PBS containing 0.5 M ammonium sulfate and purified by Phenyl‐Sepharose hydrophobic interaction chromatography. The collected fractions were pooled before dialysis against 20 mM Bis‐Tris pH 6.5. The dialyzed protein was then purified by Q‐Spharose anion exchange chromatography. The adsorbed protein was gradually eluted by linear increasing gradient of 0–300 mM NaCl. The seed‐derived DzTIs were dialyzed, concentrated, stored at −80°C in the same manner as the recombinant DzTIs until further use.

4.5. Trypsin inhibition assay

The inhibitory constant (K _i ) of each recombinant DzTI against bovine pancreatic trypsin was determined using Nα‐Benzoyl‐DL‐arginine 4‐nitroanilide (BApNA) as a chromogenic substrate. For the Lineweaver‐Burk plots, trypsin (T8003, Sigma‐Aldrich), at 200 nM, was incubated at room temperature with individual DzTIs for 10 min in 50 mM Tris pH 8.0, 1 mM CaCl₂. Subsequently, BApNA in 100% DMSO was added into the solution at the final concentrations of 0.25–2.00 mM at a fixed concentration of individual DzTIs. Initial velocity was measured by monitoring A₄₀₅ every 20 s for 10 min at 25°C. K _i values were determined by fitting the data to the competitive inhibition model using Prism 8 (GraphPad Software). The K _i value of soybean trypsin inhibitor (Gold Biotechnology T‐166‐100) was also determined for comparison. Lineweaver‐Burk plots were performed for demonstration purposes only. The original untransformed data were used for fitting. For K _i determination using the Morrison equation, the assay was performed similarly and the data were fitted to the Morrison equation using Prism 8 (GraphPad Software). The final concentration of BApNA was 1 mM. The K _M of BApNA used was 2.5 mM.

4.6. DzTI specificity against proteases and amylases

DzTI‐4, the DzTI with the lowest K _i value, was screened for its inhibition specificity against four major groups of proteases including serine protease, cysteine protease, aspartic protease, and metalloprotease. Bovine trypsin, porcine chymotrypsin (C4129, Sigma‐Aldrich), elastase (E1250, Sigma‐Aldrich), and subtilisin (P5380, Sigma‐Aldrich), which represented serine protease group, at 20 μM in 50 mM Tris–HCl, pH 8.0, 1 mM CaCl₂ (serine protease buffer) at 50 μL, were added into 50 μL of DzTI‐4 at varying on concentration in the same buffer. The mixture was incubated for 10 min at room temperature before starting the reaction by adding 40 μL of 1.0% azocasein in the serine protease buffer. The reaction was incubated at 37°C for 30 min and stopped by adding 40 μL of cold 20% trichloroacetic acid (TCA). The mixture was centrifuged at 10,000g for 10 min. A 40 μL solution of 2 M NaOH was added into an 80 μL aliquot of the supernatant. A₄₅₅ of the solution was measured to evaluate the remaining activity of the enzyme. Papain (76220, Sigma‐Aldrich) and bromelain (B4882, Sigma‐Aldrich), which represent cysteine proteases, were tested for inhibition by DzTI‐4 in the same way as the serine proteases, but using 100 mM sodium phosphate, pH 7.0, 3 mM reduced glutathione, 0.3 mM oxidized glutathione, and 1 mM EDTA as the reaction buffer. Pepsin (P7000, Sigma‐Aldrich), representing an aspartic protease, was prepared at 20 μM in 50 mM glycine, pH 2.0. The pepsin solution (50 μL) was added into solutions (50 μL) with various DzTI concentrations. The mixture was incubated at room temperature for 10 min prior to adding 40 μL of 5.8% human hemoglobin (H7379, Sigma‐Aldrich) in pepsin buffer to start the reaction. The reaction was incubated at 37°C for 30 min. The catalytic product was separated by adding cold TCA and centrifugation as described above. The supernatant (40 μL) was mixed with 200 μL of 10% Folin–Ciocalteu reagent (E463562, Carlo Erba). The solution was mixed with 160 μL of 7.5% sodium carbonate and the A₇₂₀ was subsequently measured. Carboxypeptidase A (CPA) (C9268, Sigma‐Aldrich) as a metalloprotease, was examined for its activity using hippuryl‐L‐phenylalanine as the synthetic substrate. CPA (50 μL) at 20 μM in 50 mM Tris–HCl, pH 7.5, 500 mM NaCl, 1 mM ZnCl₂ (CPA buffer) was added to 50 μL of the DzTI‐4, at various concentrations, in CPA buffer for 10 min at room temperature. Additional CPA buffer was added to achieve 138.6 μL total volume. Hippuryl‐L‐phenylalanine (1.4 μL at 100 mM) in absolute ethanol was added into the previous mixture to initiate the reaction. The initial velocity of the enzyme was monitored at A₂₅₄ for 3 min.

DzTI‐4 also was examined for inhibitory activity to α‐amylase (A3176, Sigma‐Aldrich) and β‐amylase (A7130, Sigma‐Aldrich). Enzyme (1 mg/mL in 100 mM sodium phosphate, pH 7.0, 10 mM NaCl for α‐amylase and 20 mM sodium acetate, pH 4.8 for β‐amylase) at 50 μL was mixed with 50 μL of varying concentrations of DzTI‐4 the incubate for 10 min. Potato starch solution (1% solution, 100 μL) was then added and the mixture was incubated at 37°C for 30 min. The reaction was stopped by adding 200 μL dinitrosalicylic acid solution (DNS) and then boiled for 3 min. The boiled solution was diluted by 1:5 with its corresponding buffer before measuring A₅₄₀.

4.7. X‐ray crystallography

Crystallization was performed using the hanging‐drop vapor‐diffusion method. Crystallization conditions are summarized in Table S5. Two microliters of DzTI (10 mg/mL) were mixed with the crystallization buffer at 1:1 ratio. To produce the trypsin‐DzTI‐4 complex, purified DzTI‐4 was dialyzed against 100 mM NaCl overnight. Bovine trypsin (T1426, Sigma‐Aldrich) was then added to the 1:1 molar ratio and incubated for 10 min. The protein mixture was further purified using HiPrep™16/60 Sephacryl™ S‐100 HR (GE Healthcare) gel filtration column to separate the complex. The purified complex was concentrated to 60 mg/mL for crystallization using the same method described above.

Diffraction data were collected at the Life Sciences Collaborative Access Team beamline (Advanced Photon Source, Argonne National Laboratory). The data were processed with XDS (Kabsch 2010). Scaling was done by Aimless (Evans and Murshudov 2013). The phasing strategy is summarized in Table S4. Molecular replacement was performed with Phaser (McCoy et al. ²⁰⁰⁷) as a part of the Phenix software suite (Adams et al. 2010). Refinement and model improvement were performed using phenix.refine (Afonine et al. 2012) and Coot software (Emsley et al. 2010), respectively. Note that the numbering of the bovine trypsin amino acid residues used in this paper follows that of PDB ID 1AQ7 (Sandler et al. 1998) to facilitate direct comparison. However, in our deposited structure (PDB ID 8WK1), we used the consecutive numbering system, such as UniProt A0A4W2FD21.

4.8. Site‐directed mutagenesis

pET‐32b containing DzTI‐4 was used as a template for site‐directed mutagenesis. For the W52A mutation, the primers used were 5′‐AGTATTATGTTGTGAGCGCAATTGCGGGTGCCGGCGG‐3′ and 5′‐CCGCCGGCACCCGCAATTGCGCTCACAACATAATACT‐3′. For the D79A mutation. The primers used were 5′‐GGTTCAGCGTCGCAGTGCTCTGGATTATGGTACAC‐3′ and 5′‐GTGTACCATAATCCAGAGCACTGCGACGCTGAACC‐3′. For the K94A mutation, the primers used were 5′‐TGTGTTTTATAATCTGGATACCGCAGATGATATCGTTCGTCGCAGC‐3′ and 5′‐GCTGCGACGAACGATATCATCTGCGGTATCCAGATTATAAAACACA‐3′. The PCR reactions consisted of 5% DMSO, 1× HF Phusion buffer, 0.3 mM each deoxyribonucleotide (dNTP), 250 ng of forward and reverse primers, 10 ng of the plasmid template, and 0.4 U Phusion Hot Start II DNA Polymerase (Thermo Fisher Scientific). The PCR conditions consisted of initial denaturation for 30 s at 98°C, 15 cycles of 10 s at 98°C, 30 s at 57°C, and 7 min at 68°C, and a final extension for 5 min at 68°C. The template was then digested by DpnI for 30 min before purifying the PCR product via BioFact™ PCR Purification Kit. The products were transformed into E. coli DH10β using the heat shock method. The clones with the desired mutation were identified by DNA sequencing.

The mutated DzTI‐4 proteins were produced and were purified using the same method as the wild‐type protein. Trypsin inhibition was monitored using the BApNA assay as described above, using BApNA at 1 mM.

4.9. Chemical and physical stability of DzTI‐4

To examine the effects of metal ions on the inhibitory activity of DzTI‐4, DzTI‐4 in serine protease buffer (25 μM) was incubated with 10 mM of metal ion solutions (100 μL total reaction volume) for 10 min before diluting 1:100 into the serine protease buffer containing 200 nM bovine trypsin. The mixture was left at room temperature for 10 min. BApNA was added into the mixture at the final concentration of 1 mM to start the reaction. The rate of the reaction was measured at A₄₀₅ as described previously. To investigate the effect of pH on DzTI‐4, DzTI‐4 (25 μM) was preincubated with a buffer consisting of 20 mM glycine, 20 mM succinate, and 20 mM Tris at pH 2–9 for 10 min. The protein was examined for the remaining inhibitory activity by the BApNA assay as described previously. The effects of solvents and reducing agents were examined in the same way. To investigate the temperature tolerance of DzTI‐4, DzTI‐4 (25 μM in serine protease buffer) was incubated at 25–100°C using a heat block. The incubation time was also varied between 10 and 60 min before measuring the inhibitory activity with the same assay.

4.10. DzTI‐4 supplementation after HEK293 trypsinization

HEK293 cells were cultured in completed DMEM in a T75 flask to achieve >80% confluency. The cells were washed with DMEM without FBS and detached using 0.25% Trypsin–EDTA (Gibco catalog number: 25200056). Cells suspension (150 μL, 1.42 × 10⁶ cells/mL) were mixed with 150 μL of 0.1%, 0.5%, 1.25%, 2.5%, and 4.0% (w/v) of DzTI‐4 in 20 mM Tris pH 8.0 and then left for 3 min. The cell mixtures (40 μL) were added into 160 μL of CDM4HEK293 medium (Cytiva) in wells of a 96‐well plate, resulting in 28,400 cells/well. The treated cells were cultured overnight. The cells were then washed with PBS. CDM4HEK293 medium containing 0.5 mg/mL 3‐(4,5‐dimethylthiazol‐2‐yl)‐2,5‐diphenyltetrazolium bromide (MTT) was added into the washed cells and incubated for 4 h. The solution was then replaced by 100 μL of 100% DMSO to dissolve the formazan product from viable cells. The A₅₅₀ of the resulting solution was then measured. SBTI was used as a positive control.

4.11. Mealworm antinutritional assay

One month‐old mealworm (Tenebrio molitor) was employed as an insect model for investigating the DzTI insecticidal activity (Animal use protocol no. 2323017, Chulalongkorn University Animal Care and Use Committee). Ten mealworms were placed in a 7.5 cm × 7.5 cm × 3.5 cm plastic box and were fed with an artificial diet. The artificial diet was prepared by mixing 25 g corn, 25 g cabbage, and 25 g rice bran in 200 mL of 0.1% calcium carbonate solution. The mixture was then blended together using a blender then filtered through a cheesecloth. The resulting filtrate (50 mL) was mixed with 50 g of glutinous rice flour before steaming for 15 min. The diet lump was dusted with a small amount of potato starch to reduce its stickiness. DzTI‐4, SBTI, and seed‐derived DzTI, (0.047–0.750 mg in 100 μL deionized water) were added to the artificial diet (0.3 g) in each the box. Deionized water was used as a negative control. The mealworms were housed at a temperature between 26 and 30°C and received 12 h of white LED light. Water (100 μL) supplemented with 1% of a commercial vitamin solution (3 μg/mL vitamin A, 7.5 μg/mL vitamin B1, 7.4 μg/mL vitamin B2, 10 μg/mL vitamin B6, 1.2 μg/mL folic acid, 240 μg/mL vitamin C, 40 μg/mL vitamin E, 33 mg/mL dietary fiber, 53 mg/mL sucrose, and 0.16 mg/mL sodium) was added to the diet daily. The diet was replaced every 5 days. This 0.3 g of the diet contained approximately 360 μg of soluble protein content estimated by Bradford assay. Therefore, the highest concentration of the inhibitor, 750 μg in the diet, is equal to 67.6% w/w protein content. Weight, abnormal morphology, behavior, and number of dead worms were recorded. The treatments were performed in triplicates for a month.

AUTHOR CONTRIBUTIONS

Peerapon Deetanya: Conceptualization; methodology; formal analysis; investigation; writing – original draft; visualization. Kakanang Limsardsanakij: Investigation. Grzegorz Sabat: Methodology; formal analysis; investigation; writing – original draft. Sittiporn Pattaradilokrat: Investigation. Chatchawan Chaisuekul: Investigation; methodology. Kittikhun Wangkanont: Conceptualization; methodology; formal analysis; investigation; writing – original draft; writing – review and editing; visualization; supervision; project administration; funding acquisition.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Data S1. Supporting Information.

Deetanya P, Limsardsanakij K, Sabat G, Pattaradilokrat S, Chaisuekul C, Wangkanont K. Kunitz‐type trypsin inhibitor from durian (Durio zibethinus) employs a distinct loop for trypsin inhibition. Protein Science. 2024;33(12):e5230. 10.1002/pro.5230

DATA AVAILABILITY STATEMENT

Protein sequence data have been deposited to GenBank under the accession codes indicated in Table S2. The diffraction data and three‐dimensional structures have been deposited to the Protein Data Bank under the accession codes indicated in Table S6. Other data and materials are available upon reasonable request.