Introduction

Over millions of years of evolution, predators have successfully coevolved with their prey to maintain an ecological balance ¹. In marine habitats, interactions between algae and marine herbivores dominate marine ecosystems. Most algae are consumed by marine herbivores ². They produce secondary metabolites as a chemical defense to protect themselves against predators. In brown algae Eisenia bicyclis, laminarin is a major storage carbohydrate, constituting 20%–30% of algae dry weight. The sea hare Aplysia kurodai, a marine gastropod, preferentially feeds on the E. bicyclis with its 110 and 210 kDa β-glucosidases (akuBGLs), hydrolyzing the laminarin and releasing large amounts of glucose. Interestingly, such a feeding strategy has attracted attention for producing glucose as a renewable biofuel source ³. However, to protect themselves against predators, brown algae produce phlorotannin, a secondary metabolite, thereby reducing the digestion by A. kurodai by inhibiting the hydrolytic activity of akuBGLs. As the 110 kDa akuBGL is more sensitive to phlorotannin than the 210 kDa BGL ⁴, we focused on the 110 kDa akuBGL in this study (hereafter, akuBGL refers to 110 kDa akuBGL).

To counteract the antipredator adaptations of algae, herbivores use diverse approaches, such as detoxification, neutralization, defense suppression, and physiological adaptations ⁵. A. kurodai inhibits the phlorotannin-defense of brown algae through Eisenia hydrolysis-enhancing protein (EHEP), a protein from their digestive system that protects akuBGL activity from phlorotannin inhibition ⁴. Previous studies have shown that incubating E. bicyclis with akuBGL in the presence of EHEP results in increased glucose production because EHEP binds to phlorotannin and forms an insoluble complex ⁴.

The akuBGL–phlorotannin/laminarin–EHEP system exemplifies the digestion process of A. kurodai as well as the defense and antidefense strategies between E. bicyclis and A. kurodai. Although the defense/antidefense strategy has been established, the detailed molecular mechanism of this interplay remains unknown. Further, phlorotannin inhibition hinders the potential application of brown algae as feedstocks for enzymatically producing biofuel from laminarin. Thus, understanding the underlying molecular mechanisms will be beneficial for the application of this system in the biofuel industry.

Despite the potential use of laminarin hydrolytic enzymes in the biofuel industry, only a few BGLs of glycoside hydrolases belonging to the GH3 and GH1 family are known to hydrolyze laminarin (e.g., Talaromyces amestolkiae BGL⁶, Ustilago esculenta BGL ⁷, and Vibrio campbellii BGL ⁸ from the GH3 family and Saccharophagus degradans 2-40^T BGL ⁹ from the GH1 family). GH3 is a multidomain enzyme family characterized by N-terminal (β/α)₈ (NTD) and C-terminal (β/α)₆ (CTD) domains, with or without auxiliary domains ¹⁰; the nucleophile aspartate and the acid/base glutamate residues exist in the NTD and CTD, respectively. In contrast, the members of the GH1 family generally share a single (β/α)₈-fold domain (hereafter referred to as GH1 domain [GH1D]), and the two glutamic acid catalytic residues are located in the carboxyl termini of β-strands 4 and 7. Therefore, the two families may use different substrate-recognition and catalytic mechanisms for laminarin. Intriguingly, although akuBGL possesses laminarin hydrolytic activity and belongs to the GH1 family, its molecular weight is considerably greater than that of other GH1 members. Sequence analysis has indicated that akuBGL consists of ≥2 GH1Ds. Because no structural information of BGL active on polysaccharides is available, the catalytic mechanism toward laminarin is unclear.

There is limited information on EHEP, a novel cysteine-rich protein (8.2% of the amino acid content), because no structural or functional homologous protein exists in other organisms. EHEP was predicted to consist of three peritrophin-A domains (PADs) with a cysteine-spacing pattern of CX₁₅CX₅CX₉CX₁₂CX_5-9C. The PADs consist of peritrophic matrix proteins, which have been proposed to play an important role in detoxifying ingested xenobiotics ¹¹. For instance, Aedes aegypti intestinal mucin 1 (AeIMUC1) consists of a signal peptide followed by three PADs with an intervening mucin-like domain; its expression is induced by blood feeding. AeIMUC1-mediated blood detoxification during digestion is completed by binding to toxic heme molecules¹². Despite the similar domain organization of EHEP and AeIMUC1, their function and binding partner are completely different, implying their different characteristics. However, the characteristics of the EHEP-phlorotannin insoluble complex remain unknown; moreover, it remains unclear why and how EHEP protects akuBGL from phlorotannin inhibition.

In this study, we determined the structures of apo and tannic acid (TNA, phlorotannin-analog) bound-form of EHEP (EHEP, EHEP–TNA), as well as akuBGL; all isolated from A. kurodai. The structure of EHEP consists of three PADs arranged in a triangle shape, with TNA bound at the surface of the triangle center. A structural comparison of EHEP and EHEP–TNA revealed no significant changes in conformation upon TNA binding, implying that EHEP maintains its structure when precipitated with TNA. Then, we found the conditions to resolubilize EHEP–TNA precipitate for EHEP recycling. The obtained akuBGL structure suggests that only one GH1D (GH1D2) possesses laminarin hydrolytic activity; subsequently, ligand-docking experiments demonstrated that TNA/phlorotannin has a higher docking score than laminarin. Our results revealed the mechanisms by which EHEP protects akuBGL from phlorotannin inhibition and phlorotannin inhibits the hydrolytic activity of akuBGL, providing structural support for the potential application of brown algae for biofuel production.

Results

Effects of TNA on akuBGL activity with or without EHEP

Phlorotannin, a type of tannin, is a chemical defense metabolite of brown algae. It is difficult to isolate a compound from phlorotannins because they are a group of polyphenolic compounds with different sizes and varying numbers of phloroglucinol units ¹³, such as eckol, dieckol, and so on ¹⁴. Previous studies have reported that the phlorotannin-analog TNA has a comparable inhibition effect on akuBGL to that of phlorotannin ⁴. Hence, we used TNA instead of a phlorotannin to explore phlorotannin binding with EHEP and akuBGL. This activity assay system involves multiple equilibration processes: akuBGL⇋substrate, akuBGL⇋TNA, and EHEP ⇋TNA. First, we confirmed TNA inhibition of akuBGL activity and clarified the protective effects of EHEP from TNA inhibition. The inhibition experiments showed that the galactoside hydrolytic activity of akuBGL decreased with increasing TNA concentration, indicating that TNA inhibits akuBGL activity in a dose-dependent manner (Fig. 1A, Fig. S1A). Approximately 70% akuBGL activity was inhibited at a TNA concentration of 40 μM. Moreover, protection ability analysis revealed that EHEP protects akuBGL activity from TNA inhibition in a dose-dependent manner, as indicated by the recovery of the inhibited akuBGL activity with increasing EHEP concentration (Fig. 1B, Fig. S1B). Further, approximately 80% of akuBGL activity was recovered at an EHEP concentration of 3.36 μM.

Figure 1.

Overall structure of EHEP

Considering the lack of known homologous proteins of EHEP, we determined the structure of EHEP using the native-SAD method at a resolution of 1.15 Å, with a R_work and R_free of 0.18 and 0.19, respectively (Table 1). The residues A21–V227 in purified EHEP (1–20 aa were cleaved during maturation) were built, whereas two C-terminal residues were disordered. The structure of EHEP consists of three PADs: PAD1 (N24– C79), PAD2 (I92–C146), and PAD3 (F164–C221), which are linked by two long loops, LL1 (Q80–N91) and LL2 (R147– G163), and arranged in a triangle shape (Fig. 2A). These three PADs share a similar structure, with a root-mean-squared difference (RMSD) of 1.065 Å over 46 Cα atoms and only ∼20.3% sequence identity (Figs. 2B and 2C). The three PADs share a canonical CBM14 fold consisting of two β-sheets containing three N-terminal and two C-terminal antiparallel β-strands. Additionally, two small α-helices were appended to the N- and C-terminus in PAD1 and PAD3 but not in PAD2 (Fig. 2B).

Figure 2.

Table 1

Although the Dali server ¹⁵ did not provide similar structures using the overall structure of EHEP as the search model, six structures showed similarities with a single PAD of EHEP. These structures were the members of the PAD family, including the chitin-binding domain of chitotriosidase (PDB ID 5HBF) ¹⁶, avirulence protein 4 from Pseudocercospora fuligena [PfAvr4 (PDB ID 4Z4A)] ¹⁷ and Cladosporium fulvum [CfAvr4 (PDBID 6BN0)] ¹⁸, allergen Der p 23 (PDB ID 4ZCE) ¹⁹, tachytitin (PDB ID 1DQC) ²⁰, and allergen Blot 12 (PDB ID 2MFK), with Z-scores of 4.7–8.4, RMSD values of 1.2–2.8 Å, and sequence identity of 19%–37%. The highest sequence disparity was detected in PAD2, whereas the greatest structural differences were noted in PAD3. The C^No1X₁₅C^No2X₅C^No3X₉C^No4X₁₂C^No5X_5-9C^No6 motif (superscripts and subscripts indicate the cysteine number and number of residues between adjacent cysteines, respectively) in each PAD of EHEP formed three disulfide bonds between the following pairs: C^No1–C^No3, C^No2–C^No6, and C^No4–C^No5 (Fig. 2B). Such rich disulfide bonds may play a folding role in the structural formation of EHEP, with >70% of the backbone in a loop conformation. A similar motif with disulfide bonds was observed in tachycitin ²⁰, PfAvr4 ¹⁷, CfAvr4 ¹⁸, and the chitin-binding domain of chitinase ¹⁶. Although these proteins share a highly conserved core structure, they have different biochemical characteristics. For example, the chitin-binding domain of human chitotriosidase, Avr4 and tachycitin possess chitin-binding activity, but the critical residues for chitin-binding are not conserved ^{16, 18, 21}, indicating that they employ different binding mechanisms. In contrast, EHEP and allergen Der p 23 do not possess chitin-binding activity ^{4, 19}. Thus, the PAD family may participate in several biochemical functions.

Modification of EHEP

Consistent with the molecular weight results obtained using MALDI–TOF MS ²², the apo structure2 (1.4 Å resolution) clearly showed that the cleaved N-terminus of Ala21 underwent acetylation (Fig. S2A), demonstrating that EHEP is acetylated in A. kurodai digestive fluid. N-terminal acetylation is a common modification in eukaryotic proteins. Such acetylation is associated with various biological functions, such as protein half-life regulation, protein secretion, protein–protein interaction, protein–lipid interaction ²³, metabolism, and apoptosis ²⁴. Further, N-terminal acetylation may stabilize proteins ²⁵. To explore whether acetylation affects the protective effects of EHEP on akuBGL, we used the E.coli expression system to obtain the unmodified recomEHEP (A21–K229). We measured the TNA-precipitating assay of recomEHEP. The results revealed that recomEHEP precipitated after incubation with TNA, at a comparable level to that of natural EHEP (Fig. S2B), indicating that acetylation is not indispensable for the phlorotannin binding activity and stabilization of EHEP. Future studies are warranted to verify the exact role of N-terminal acetylation of EHEP in A.kurodai.

TNA binding to EHEP

To understand the mechanism by which TNA binds to EHEP, we determined the structure of EHEP complexed with TNA (EHEP–TNA) using the soaking method. In the obtained structure, both 2F_o–F_c and F_o–F_c maps showed the electron density of 1,2,3,4,6-pentagalloylglucose, a core part of TNA missing the five external gallic acids (Fig. 3A, Fig. S3A). Previous studies have shown that acid catalytic hydrolysis of TNA requires a high temperature of 130°C ²⁶; even with a polystyrene-hollow sphere catalyst, a temperature of 80°C is required ²⁷. Therefore, the five gallic acids could not be visualized in the EHEP–TNA structure most likely due to the structural flexibility of TNA.

Figure 3.

The apo EHEP and EHEP–TNA structures were extremely similar, with an RMSD value of 0.283 Å for 207 Cα atoms (Fig. S3B). However, the superposition of the two structures showed a decrease of the loop part in EHEP–TNA. TNA binding caused a slight increase in the α-helix and β-sheet contents of PAD2 and PAD3 (Fig. S3B). In the EHEP–TNA structure, the residues C93–Y96 of PAD2 folded into an α-helix and each β-sheet of the first β-strand in PAD3 elongated by incorporating one residue in the first (G¹⁷⁶) and second β-sheets (S¹⁸⁶) and three residues in the third β-sheet (H¹⁹⁷MP¹⁹⁹). The EHEP–TNA structure revealed that TNA binds to the center of the triangle formed by the three PADs, a positively charged surface (Fig. 3A and Fig. S3C). The binding pocket on EHEP surface was formed by the C-terminal α-helix of PAD1, the N-terminal α-helix of PAD2, and the middle part (loop) of PAD3 assisted by two long linker loops (LL1 and LL2). TNA was primarily bound to EHEP via hydrogen bonds and hydrophobic interactions (Fig. 3B). Gallic acid1, 4, and 6 interacted with EHEP via hydrogen bonds and additional hydrophobic contacts, whereas gallic acid2 and 3 only interacted hydrophobically with EHEP. The 3-hydroxyl groups of gallic acid1 and 6 individually formed a hydrogen bond with the main chain of G74 and the side chain of N75 in PAD1. The backbone carbonyl of Y96 and P199 in PAD2 and PAD3, respectively, formed a hydrogen bond with the 3,5-hydroxyl groups of gallic acid4. Additionally, some hydrogen bonds were formed between TNA and water molecules. TNA binding was also stabilized by hydrophobic interactions between the benzene rings of gallic acid and EHEP. For instance, gallic acid4 and 6 showed alkyl–π interaction with P201 and P77, respectively; moreover, gallic acid3 and 4 formed amide–π stackings with P199.

The EHEP–TNA structure clearly showed that TNA binds to EHEP without covalent bonds and the binding does not induce significant structural changes; thus, we attempted to recover EHEP from EHEP–TNA precipitates by adjusting the pH. As hypothesized, re-solubilization of the EHEP–phlorotannin precipitate is pH-dependent (Fig. 3C). The EHEP–TNA precipitate did not resolubilize at pH 7.0; however, after incubating for >1 h at pH 7.5, the precipitate started resolubilizing. Most of the precipitate rapidly resolubilized at an alkaline pH (≥8.0) after incubation for 15 min. Further, the resolubilized EHEP had the same elution profile as that of the natural EHEP (Fig. 3D) in SEC, suggesting that resolubilized EHEP maintained the native structure and its phlorotannin-precipitate activity (Fig. S3D).

Two domains of akuBGL

To reveal the structural basis of akuBGL recognition of laminarin and inhibition by TNA, we attempted to determine its structure. We soaked crystals in TNA as well as various substrate solutions but finally obtained the optimal resolution using crystal soaking in TNA. There was no electron density of TNA or something similar in the 2Fo–Fc and Fo–Fc map of the obtained structure; thus, we considered this structure as the apo form of akuBGL.

Two akuBGL molecules were observed in an asymmetric unit (MolA and MolB). These molecules lacked the N-terminal 25 residues (M1–D25), as confirmed using N-terminal sequencing analysis of purified natural akuBGL. This N-terminal fragment was predicted to be a signal peptide using the web server SignalP-5.0. The residues L26–P978 were constructed in both MolA and MolB with glycosylation, whereas the remaining C-terminal residues (A979–M994) could not be visualized as they were disordered. The electron density map of Fo–Fc revealed N-glycosylation at three residues, i.e., N113, N212, and N645 (Figs. S4A, B). N-glycosylation of GH enzymes prevents proteolysis and increases thermal stability ^{28, 29}. Additionally, a study on β-glucosidase Aspergillus terreus BGL demonstrated that N-glycosylation of N224 affected the folding stability, even when it is located close to a catalytic residue ³⁰. In akuBGL, all N-glycosylation sites were present on the surface, far away from the catalytic site. Therefore, we speculate that akuBGL glycosylation does not affect its activity. Except for the difference in visualized glycans resulting from glycosylation, MolA and MolB were similar, with a RMSD value of 0.182 for 899 Ca atoms; therefore, we used MolA for further descriptions and calculations.

The structure of akuBGL consisted of two independent GH1 domains, GH1D1 (L26–T494) and GH1D2 (D513–P978), linked by a long loop (D495–Y512) (Fig. 4A). There was little interaction between GH1D1 and GH1D2, only in a buried surface area comprising 2% of the total surface (708.9 Ų) (Fig. S4C). GH1D1 and GH1D2 have a sequence identity of 40.47% and high structural similarity with an RMSD value of 0.59 Å for 371 Cα atoms (Fig. S5A upper panel).

Figure 4.

Glucosidases of the GH1 family utilize the retaining mechanism with two glutamic acids for catalyzing glucoside hydrolysis. In general, the distance between the two catalytic oxygen atoms of the side chain of two glutamic acids is approximately 5 Å ³¹. Sequence and structure alignment of GH1D1 and GH1D2 of akuBGL with other members of the GH1 family revealed that the second glutamate is conserved (E404), but the first glutamate is replaced by D192 in GH1D1. The oxygen atoms of the side chains between D192 and E404 of GHD1 were 8.4 Å apart. In contrast, GH1D2 conserved two glutamic acids (E675 and E885) at the carboxyl termini of β-strands 4 and 7; the distance between oxygen atoms of E675 and E885 side chains was 5.1 Å (Fig. S5A bottom panel), similar to that of Neotermes koshunensis BGL (NkBGL, PDB ID 3VIH) ³², Nannochloropsis oceanica BGL (NoBGL, PDB ID 5YJ7) ³³, and Spodoptera frugiperda BGL (PDB ID 5CG0) (3.9–4.9Å) ³⁴. Furthermore, regarding the two other conserved essential regions for β-glucosidase activity, namely, glycone-binding site (GBS) and catalysis-related residues (CR), GH1D1 conserved neither GBS nor CR, whereas GH1D2 conserved both (Fig. 4B). Altogether, we suggest that GH1D1 does not possess catalytic activity. We expressed and purified the recombinant GH1D1, which did not show any hydrolytic activity toward O-PNG (Figs. S4D, E), although we could not rule out the effect of N-glycosylation.

Structural comparison of GH1D2 with other BGLs, including NkBGL (PDB ID 3VIH) ³², rice (Oryza sativa L.) BGL (OsBGL, PDB ID 2RGL) ³⁵, and microalgae NoBGL (PDB ID 5YJ7) ³³, revealed the characteristics of each active pocket (Fig. S5B). OsBGL and NoBGL have a deep, narrow, and straight pocket, whereas GH1D2 and NkBGL have a broad and crooked pocket. Such active pocket shapes reflect the substrate preferences of OsBGL and NoBGL; they hydrolyze laminaribiose with no detectable activity toward laminaritetraose ^{33, 36}. Furthermore, the difference in the features of large active pockets between NkBGL and GH1D2, wherein GH1D2 often possesses an auxiliary site with several aromatic residues bound to the carbohydrate via CH–π interactions ³⁷, may explain their substrate specificity. NkBGL efficiently hydrolyzes laminaribiose and cellobiose but has weak hydrolytic activity toward laminarin ³⁸. In contrast, the GH1D2 of akuBGL has similar activity levels toward cellobiose and laminarin ³⁹. Therefore, the GH1D2 of akuBGL may recognize larger substrates than that of other BGLs. Laminarin typically has a curved conformation; accordingly, narrow- and straight-shaped pockets are incompatible for binding. Furthermore, we docked GH1D2 with laminaritetraose, wherein the four glucose units formed extensive contacts with GH1D2. Hydrogen bonds involved the catalytic residues E675. In addition, several aromatic residues, such as W631, F677, W681, F689, Y819, Y846, W857, and W935, formed CH–π stacking (Fig. S6). Some interacting residues belonged to GBS and CR sites, such as E675, W631, Y819, and W935. Additionally, the docking structure revealed that the +3 and +4 glucose of laminaritetraose is located at the auxiliary binding site and that atom O1 of the +4 glucose is positioned outside the pocket (Fig. S6), implying that the auxiliary binding site with several aromatic residues (F677 and W681) of GH1D2 facilitates laminarin binding.

Inhibitor binding of akuBGL

As we could not obtain the complex structure of akuBGL with TNA, we performed docking calculations of akuBGL GH1D2 with TNA to explore the inhibition mechanism. The docking model of akuBGL–TNA showed that seven gallic acid rings of TNA formed an extensive hydrogen bond network with akuBGL in the binding pocket (Fig. 5). The hydroxyl groups of TNA formed hydrogen bonds with the residues N552, E675, D735, K739, K759, Q840, T844, D852, and K859 of GH1D2. Moreover, benzene rings showed hydrophobic interactions with several hydrophobic residues. In particular, stable π–π stacking was observed between TNA and residues F547, W631, F689, Y846, W857, and W935. Among these residues, the conserved E675 was the catalytic residue, and W631, W935, and E934 contributed to GBS and CR sites.

Figure 5.

In addition, we performed a docking calculation of GH1D2 with the characteristic inhibitors eckol and phloroglucinol ⁴⁰. The binding mechanisms of eckol and phloroglucinol were similar to those of TNA but with different contact residues (Fig. S7). For eckol, the six hydroxyl groups formed hydrogen bonds with residues E675, D735, E737, K759, E885, and E934. Additionally, residues W631, F677, F689, Y819, W857, W935, F943, and W927 formed π–π stacking with eckol. For phloroglucinol, the three hydroxyl groups formed hydrogen bonds with E675, E885, and E934, whereas residues W631, F689, Y819, W857, W935, and W927 formed π–π stacking with the benzene ring.

The docking scores of the inhibitors TNA, eckol, and phloroglucinol were −8.8, −7.3, and −5.7 kcal/mol, respectively, whereas the substrate laminaritetraose had a docking score of −6.6 kcal/mol. Those docking scores corroborated well with the inhibition activity toward akuBGL, that TNA had a more robust inhibition activity than phloroglucinol ⁴, indicating that the docking results are reasonable. In summary, the three inhibitors interacted with akuBGL through similar binding mechanisms to occupy the substrate-binding site, suggesting a reversible competitive inhibition mechanism.

Discussion

In marine habitats, the ecological interactions between brown algae and herbivores dominate marine ecosystems ⁴¹. The akuBGL–phlorotannin/laminarin–EHEP system represents the feeding defense–offense associations between A. kurodai and brown algae. We focused on this system to understand the molecular mechanism at the atomic level. In contrast to most GH1 BGLs containing one catalytic GH1 domain, akuBGL consists of noncatalytic GH1D1 and catalytic GH1D2. The noncatalytic GH1D1 may act as a chaperone of GH1D2, as we successfully overexpressed GH1D1 but failed to do the same for GH1D2. Such multi-GH1D assembly and a similar function has been suggested in β-glucosidase CjCEL1A of Corbicula japonica ⁴² and glycosidase LpMDGH1 of the shipworm Lyrodus pedicellatus⁴³. CjCEL1A has two tandem GH1Ds with a sequence identity of 43.41% ⁴². Two catalytic glutamic acids and the residues related to substrate binding are conserved in the second GH1D, whereas the first GH1 domain lacks these conserved residues and may play a role in folding the catalytic domain. LpMDGH1 consists of six GH1Ds, among which GH1D2, 4, 5, and 6 contain the conserved residues for activity, whereas others do not contain these residues and might be involved in protein folding or substrate interactions ⁴³. Future assay of GH1D2 and inactive mutants is the complement to validate the molecular mechanism of akuBGL.

BGLs have different substrate preferences in the degree of polymerization and type of glycosidic bond. In general, BGLs prefer to react with mono-oligo sugars over polysaccharides. For instance, OsBGL, NoBGL, and NkBGL hydrolyze disaccharides (cellobiose and laminaribiose) but display no or weak activity toward polysaccharides (cellulose and laminarin) ^{33, 36, 38}. The structure of GH1D2 explained the substrate preference for the polysaccharide laminarin. GH1D2 contains an additional auxiliary site composed of aromatic residues (Fig. S5B) in the substrate entrance pocket, which putatively enables it to accommodate a long substrate, contributing to akuBGL activity toward laminarin, as supported by docking calculations (Fig. S6). In addition, docking analysis of akuBGL GH1D2 with inhibitors (TNA, phlorotannin, eckol, and phloroglucinol) revealed that these inhibitors bind to the substrate-binding site via hydrogen bonds and hydrophobic interactions similar to laminarin. Such binding mechanisms suggest the presence of competitive inhibition to occupy the binding site, consistent with previous research ⁴. Future kinetic experiments are required to validate quantitatively the competitive inhibition of phlorotannin against akuBGL.

EHEP, expressed in the midgut of A. kurodai, was identified as an antidefense protein, protecting the hydrolysis activity of akuBGL from phlorotannin inhibition ⁴. Such an ecological balance also exists between plants and their predatory mammals and insects. Similar to brown algae, plants use the toxic secondary metabolite tannins as their defense mechanism against predators, which constitute 5%–10% of dry weight of leaves. In vertebrate herbivores, tannins reduce protein digestion ⁴⁴. In phytophagous insects, tannins may be oxidized at the alkaline pH of the insect midgut and cause damage to cells ⁴⁵. The evolution of plant–herbivore survival competition has led to the development of remarkably unique adaptation strategies. Mammals feeding on plants that contain tannin may overcome this defense by producing tannin-binding proteins, proline-rich proteins, and histatins ^{46 47}. Proline constitutes at least 20% of the total amino acid content in proline-rich proteins; for some species, the proportion of proline reaches 40%. Histidine constitutes 25% of total amino acid content in histatins. Both proline-rich proteins and histatins are unfolding proteins with randomly coils in solution. In caterpillars, the oxidation damage of tannin is reduced by the low oxygen level. Some insects use the peritrophic membrane to transport tannins into the hemolymph, where they are excreted ⁴⁸. Additionally, the peritrophic envelop protects insects from tannins by forming an impermeable barrier to tannins ⁴⁴. A. kurodai uses a similar strategy to mammals by secreting the tannin-binding protein EHEP. Although EHEP has a completely different amino acid composition with proline-rich proteins and histatins, EHEP also binds to phlorotannin. Therefore, EHEP may be a specific counteradaptation that allows A.kurodai to feed on brown algae, as there are no homologous proteins in other organisms.

The three PADs of EHEP are arranged in a triangle shape, forming a large cavity on the surface at the triangle center to provide a ligand-binding site. EHEP has a positively charged surface at a pH of <6.0, whereas the surface becomes negatively charged at a pH of >7.0 (Fig. S3C). Meanwhile, TNA has a pKa of 4.9–8 ^49–51, showing minor negative charges at an acidic pH and the highest negative charge at a pH of >7.0 ⁵². Therefore, TNA binds to EHEP at a pH of <6.0 (pH of crystallization = 4.5), but it shows charge repulsion with EHEP at a pH of >8.0. Altogether, TNA is protonated and behaves as a hydrogen bond donor when the pH is below its pKa, whereas when the pH is above its pKa, TNA is deprotonated and the hydrogen bonding cannot be maintained. As losing hydrogen bonds and increasing repulsive forces at a pH >8.0, the precipitated EHEP–TNA could dissolve in the buffer of pH >8.0. This pH-induced reversible interaction also occurred in other proteins, such as BSA, pepsin, and cytochrome C ⁵³. The phlorotannin members share a similar structure with TNA; thus, we speculate that the EHEP–phlorotannin complex also exhibits a pH-induced reversible interaction. In vivo, the pH of the digestive fluid of A. kurodai is approximately 5.5⁴, which favors the binding of EHEP to phlorotannin. In the alkaline hindgut ⁵⁴, the EHEP–phlorotannin disassociates (Fig. 6), and the phlorotannin is subsequently excreted from the anus.

Figure 6.

Based on the EHEP–TNA structure and docking models of akuBGL– inhibitor/substrate, we proposed a mechanism of phlorotannin inhibition on akuBGL activity and EHEP protection from phlorotannin inhibition (Fig. 6). Because laminarin lacks the benzene rings essentially to form CH–π stacking interactions with EHEP, the EHEP can be considered not binding with the laminarin. In the absence of EHEP, phlorotannin occupies the substrate-binding site of akuBGL, inhibiting the substrate from entering the active site and resulting in no glucose production. In the presence of EHEP, it competitively binds to phlorotannin, freeing the akuBGL pocket. Then, the substrate can enter the active pocket of akuBGL and glucose can be produced normally. The digestive fluid of A. kurodai contains EHEP at a high concentration (>4.4 µM) ⁴, which is slightly higher than the concentration of EHEP (3.36 µM) that protects akuBGL activity (Fig. 1B). The high concentration of EHEP allows A. kurodai feeding of phlorotannin-rich brown algae. The balance between phlorotannin inhibition and protection is controlled by the concentrations of phlorotannin and EHEP in vivo.

The akuBGL–phlorotannin/laminarin–EHEP system is the digestive–defensive– offensive associations between algae and herbivores. Our study presented the molecular mechanism of this system at the atomic level, providing a molecular explanation for how the sea hare A. kurodai utilizes EHEP to protect akuBGL activity from phlorotannin inhibition. Further, such a feeding strategy has attracted attention for producing glucose as a renewable biofuel source, so our studies provide a molecular basis for the biofuel industry applications of brown algae.

Materials and Methods

EHEP and akuBGL preparation

Natural EHEP (22.5 kDa) and akuBGL (110 kDa) were purified from A. kurodai digestive fluid as described previously ⁴. For crystallization, we further added one step for purification of EHEP using size exclusion chromatography (HiLoad 16/60 Superdex 75, GE Healthcare), for which the column was equilibrated with 20 mM MES-NaOH buffer (pH 6.5). Obtained EHEP was then concentrated to 15–25 mg/mL using Vivaspin-4 10K columns (Sartorius, Göttingen, Germany). About akuBGL, we exchanged buffer from 20 mM Tris-HCl pH 7.0 to 20 mM Bis-tris pH 6.0 and concentrated it to 11 mg/mL using Amicon with a cutoff of 50 kDa.

To verify whether the chemical modification which was indicated by previous study²² affects the function of EHEP, we prepared recombinant EHEP (recomEHEP) without the N-terminal signal peptide (1–20 aa) and chemical modifications ²². EHEP cDNA was obtained via reverse transcription–polymerase chain reaction (RT–PCR) using the total RNA of A. kurodai as a template. The reamplified fragment was digested and ligated to a plasmid derived from pET28a (Novagene, Darmstadt, Germany). We transformed the plasmid containing recomEHEP into E. coli B834(DE3) pARE2 and expressed it with a C-terminal hexahistidine-tag. The cells were cultured in lysogeny broth (LB) medium with the antibiotics kanamycin (25 mg/L) and chloramphenicol (34 mg/L) until the optical density at 600 nm (OD₆₀₀) reached 0.6. Subsequently, overexpression was induced by adding 0.5 mM isopropyl b-D-L-thiogalactopyranoside for 20 h at 20 ℃. After harvesting by centrifugation, the cells were resuspended in a buffer containing 50 mM Tris-HCl pH 7.4, 300 mM NaCl, DNase, and lysozyme and were disrupted via sonication. The insoluble part was removed by centrifugation for 30 min at 40000 ×g at 4 ℃. We loaded the supernatant onto a 5-mL Histrap HP column and the recomEHEP was eluted using increasing concentrations of imidazole (0–500 mM). The purified proteins were dialyzed against a solution containing 50 mM Tris-HCl pH 7.4 and 50 mM NaCl and subsequently loaded onto a Hitrap Q column and eluted by a linear gradient of a solution containing 50 mM Tris-HCl and 1 M NaCl. Fractions containing recomEHEP were concentrated and then purified using a gel filtration column (Hiload 16/60 superdex 75 pg) equilibrated with 20 mM sodium acetate pH 6.0 and 100 mM NaCl. We collected the fractions containing recomEHEP and concentrated them to 2.1 mg/mL using Amicon (Merck, American).

DNA encoding akuBGL without signal peptide was codon optimized (GENEWIZ, China) for overexpress in E. coli, and cDNA encoding GH1D1 domain was subcloned to a modified pET-32a vector (Invitrogen, USA) with an N-terminal TrxA tag and a 6×His tag followed by a TEV protease recognition site. The plasmid pET-32a-GH1D1 was electorally transformed into E.coli origami2 cells. We grew the cells in LB medium supplemented with 100 µg/mL of ampicillin at 37 °C until the OD₆₀₀ reached 0.6. Then the cultures were cooled and induced with 0.5 mM isopropyl-β-D-thiogalactopyranoside at 20 ℃ for 16 h. After resuspending the harvested cells in a buffer containing 50 mM Tris-HCl (pH 7.4), 250 mM NaCl, and 5% glycerol, we disrupted the cells by sonication. The cell lysate was centrifugated at 40,000 ×g for 30 min at 4 ℃. The supernatant was filtered by 0.45 μm membrane and then loaded onto a Histrap HP column (GE, USA). After washing with lysis buffer supplemented with 20 mM imidazole, the recomGH1D1 was eluted linearly from the column with lysis buffer supplemented with 500 mM imidazole. The fractions containing recomGH1D1 were added with TEV protease at a ratio of 1:10 and dialyzed against the buffer containing 50 mM Tris (pH 7.4), 50 mM NaCl, and 2 mM DTT at 4 ℃ overnight. Then, we purified the recomGH1D1 by a Histrap HP column again, and collected the flowthrough. Finally, we purified the recomGH1D1 by size exclusion chromatography on a Hiload 16/60 200 pg column (GE, USA) equilibrated with a buffer containing 20 mM Bis-Tris (pH 6.0). The recomGH1D1 was concentrated to 1.5 mg/mL by centrifugation using Amicon Ultra with 10 kDa. cutoff.

N-terminal sequencing of akuBGL

We performed a N-terminal sequencing of purified akuBGL using the Edman degradation method. akuBGL was separated by SDS–PAGE, followed by electrophoretic transfer onto a PVDF (polyvinylidene fluoride) membrane (GE, American). The membrane was subsequently stained with Ponceau S solution. The band corresponding to akuBGL was excised and analyzed using a PPSQ-53A Protein sequencer (Shimadzu, Japan) at the Instrumental Analysis Service of Hokkaido University.

Effects of TNA on akuBGL activity with or without EHEP

Due to the yellow color of TNA, which affects absorbance at 420 nm of the reaction product o-nitrophenol of akuBGL, we used HPLC to measure the akuBGL activity in the reaction system involving TNA. Ortho-nitrophenyl-β-galactoside (ONPG) was used as a substrate to measure akuBGL activity. The reaction system (100 μL) included 2.5 mM ONPG, 49 nM akuBGL, and different TNA concentrations in a reaction buffer (50 mM CH₃COONa pH 5.5, 100 mM NaCl, and 10 mM CaCl₂). After incubation for 10 min at 37 ℃, 100 μL of methanol was added to each sample to terminate the reaction. Then, the mixture was centrifuged for 10 min at 15000 ×g at 4 ℃ and the supernatant was used for analyzing akuBGL activity via HPLC. To measure the protective effect of EHEP on akuBGL, we added different amounts of EHEP to the reaction system.

RecomGH1D1 activity assay

We measured simply the activity of the recomGH1D1 using a spectrophotometer because the reaction product o-nitrophenol is yellow. The reaction system (100 μL) included 2.5 mM ONPG, 49 nM akuBGL or recomGH1D1 in a reaction buffer (50 mM CH₃COONa pH 5.5, 100 mM NaCl, and 10 mM CaCl₂). After incubation for 10 min at 37 ℃, the reaction was terminated by adding 100 μL of 500 mM Na₂CO₃. The absorbance at 420 nm was measured using a SpectraMax spectrophotometer (Molecular Devices, Japan).

Binding assay for recomEHEP with TNA

We measured the binding activity of recomEHEP using precipitation analysis in same method of natural EHEP, as described previously ⁴. Briefly, recomEHEP or EHEP was incubated with TNA at 25 ℃ for 90 min and centrifuged for 10 min at 12000 ×g at 4 ℃. Then, we washed the precipitates twice and resuspended them in an SDS–PAGE loading buffer for binding analysis.

Resolubilization of the EHEP–eckol precipitate

A mixture of 2 mg of EHEP and 0.4 mg of eckol was incubated at 37 °C for 1 h, followed by centrifugation at 12000 ×g for 10 min, and the supernatant was removed. The sediment was dissolved in 50 mM Tris-HCl buffer at different pH (7.0–9.0) and the absorbance at 560 nm was measured over time. After checking the elution peak by SDS–PAGE, the resolubilized EHEP was analyzed by a Sephacryl S-100 HR column (2.0 ×110 cm). Moreover, the eckol/dieckol-binding activity of resolubilized EHEP was assessed, as mentioned above in this section.

Crystallization and data collection

The crystallization, data collection, and initial phase determination of EHEP were described previously ²². As EHEP precipitates when bound to TNA, we could not cocrystallize EHEP with TNA. Therefore, we used the soaking method to obtain the EHEP–TNA complex. Owing to the poor reproducibility of EHEP crystallization, we used a co-cage-1 nucleant ⁵⁵ to prepare EHEP crystals for forming the complex with TNA. Finally, we obtained high-quality EHEP crystals under the reservoir solution containing 1.0 M sodium acetate, 0.1 M imidazole (pH 6.5) with co-cage-1 nucleant. Subsequently, we soaked the EHEP crystals in a reservoir solution containing 10 mM TNA at 37 ℃ for 2 days; then, they were maintained at 20 ℃ for 2 weeks. Next, we soaked the EHEP crystals in a reservoir solution containing 10 mM phloroglucinol. For data collection, the crystal was soaked in a cryoprotectant solution containing 20% (v/v) glycerol along with the reservoir solution. Diffraction data were collected under a cold nitrogen gas stream at 100 K using Photon Factory BL-17 (Tsukuba, Japan) or Spring 8 BL-41XU (Hyogo, Japan).

For akuBGL crystallization, the initial crystallization screening was performed using the sitting-drop vapor-diffusion method with Screen classics and Classics II crystallization kits (Qiagen, Hilden, Germany) and PACT kits (Molecular Dimensions, Anatrace, Inc.) at 20 ℃. Crystallization drops were set up by mixing 0.5 μL of protein solution with an equal volume of the reservoir solution. The initial crystals were obtained under condition no. 41 (0.1 M sodium acetate pH 4.5 and 25% polyethylene glycol [PEG] 3350) of Classics II, no. 13 (0.1 M MIB buffer [25 mM sodium malonate dibasic monohydrate, 37.5 mM imidazole, and 37.5 mM boric acid], with pH 4.0 and 25% PEG 1500), and no. 37 (0.1 M MMT buffer [20 mM DL-malic acid, 40 mM MES monohydrate, and 40 mM Tris], with pH 4.0 and 25% PEG 1500) of PACT. After optimization by varying the buffer pH and precipitant concentration and adding co-cage1 nucleant, the optimal crystals were obtained using 0.1 M sodium acetate pH 4.5, and 20% PEG 3350 with a co-cage1 nucleant at a protein concentration of 5.4 mg/mL. Diffraction data were collected under a cold nitrogen gas stream at 100 K using Photon Factory BL-1A (Tsukuba, Japan) after cryoprotection by adding glycerol to a 20% final concentration into the reservoir solution. The optimal resolution of diffraction data was obtained by soaking a crystal with 5 mM TNA in the reservoir buffer at 37 ℃ for 4 h.

All datasets were indexed, integrated, scaled, and merged using XDS/XSCALE program ⁵⁶. Statistical data collection and process are summarized in Table 1.

Structure determination and refinement

For EHEP structure determination, after initial phasing via the native-SAD method^{22, 57}, the model was obtained and refined with auto-building using Phenix.autobuild of Phenix software suite ⁵⁸. The obtained native-SAD structure was used as a model for rigid body refinement using phenix.refine ⁵⁹ of Phenix software suite with a native data at high resolution of 1.15 Å. The structure of EHEP was automatically rebuilt using Phenix.autobuild of the Phenix software suite again ⁵⁸. Several rounds of refinement were performed using Phenix.refine of the Phenix software suite ⁵⁸, alternating with manual fitting and rebuilding using COOT program ⁶⁰. The final refinement statistics and geometry are shown in Table 1.

The structure of the EHEP–TNA complex was determined via the molecular replacement (MR) method using the EHEP structure as a search model with Phaser of Phenix software suite ⁶¹. The electron density block of TNA was clearly shown in both 2F_o–F_c and F_o–F_c maps. Subsequently, TNA structure was manually constructed, followed by several rounds of refinement using Phenix.refine ⁵⁸, with manual fitting and rebuilding using COOT ⁶⁰. We also determined the structure of phloroglucinol-soaked crystals at a resolution of 1.4 Å via the MR method using the refined EHEP structure as a search model with Phaser, but no electron density block of phloroglucinol was obtained. Therefore, we referred to this structure as the apo form (apo structure2). The final refinement statistics and geometry are shown in Table 1.

We determined the structure of akuBGL via the MR method using Phaser of Phenix software suite ⁶¹. We used one GH domain (86–505 aa) of β-klotho (PDB entry: 5VAN) ⁶² as the search model. This GH domain of β-klotho shares 30% sequence identity with akuBGL. Four GH domains of two molecules in an asymmetric unit were found and subsequently rebuilt with Phenix_autobuild of Phenix software suite ⁵⁸. Finally, refinement of akuBGL structure was performed as described for EHEP.

Docking studies of akuBGL with phlorotanins and laminarins

We used Schrodinger Maestro program for performing docking studies ⁶³. First, we superimposed the structure of OsBGL mutant complexed with cellotetraose (PDB ID 4QLK ⁶⁴) to that of akuBGL GH1D2 to define the ligand position in the ligand-binding cavity. Then, we modified the structure of the akuBGL GH1D2 using wizard module to remove water molecules and add hydrogen atoms for docking. The 2D structures of the inhibition ligands, including TNA, phloroglucinol, and eckol, were downloaded from Pubchem ⁶⁵ and further converted to 3D structures using the LigPrep module of Schrodinger Maestro program. The structure of the substate laminaritetraose was extracted from the Zobellia galactanivorans β-glucanase–laminaritetraose complex structure (PDB ID: 4BOW) ⁶⁶. Then, a receptor grid was constructed in the center of the ligand-binding cavity. We performed docking using the Glide standard precision mode without any constraints. The optimal binding pose was determined using the lowest Glide score, and docked structures were analyzed using PyMol.

Data Availability

The atomic coordinates were deposited in the PDB with the accession codes as follows: EHEP with 1.15 Å resolution (8IN3), EHEP with 1.4 Å resolution (8IN4), EHEP complexed with tannic acid (8IN6), akuBGL(8IN1).

Acknowledgements

This work was supported in part by Grant-in-Aid for Scientific Research (B) (Grant Number 21H01754 To M. Y) and Platform Project for Supporting Drug Discovery and Life Science Research (Basis for Supporting Innovative Drug Discovery and Life Science Research (BINDS)) from Japan Agency for Medical Research and Development (AMED) under Grant Number JP18am0101071 and JP19am0101083. We are grateful to the Photon Factor and SPring-8 (No. 2017B2545, 2017A2551, 2018B2538) for beam time and the beamline staff for their assistance for data collection.

Supporting information

Fig. S1.

Figure. S2.

Figure S3.

Figure S4.

Figure S5.

Figure S6.

Figure S7.