Functional analyses of chitinolytic enzymes in the formation of calcite prisms in

Background: The mollusk shells present distinctive microstructures that are formed by small amounts of organic matrices controlling the crystal growth of calcium carbonate. These microstructures show superior mechanical properties such as strength or flexibility. The shell of Pinctada fucata has the prismatic layer consisting of prisms of single calcite crystals. These crystals contain small-angle grain boundaries caused by a dense intracrystalline organic matrix network to improve mechanical strength. Previously, we identified chitin and chitinolytic enzymes as components of this intracrystalline organic matrix. In this study, we analyzed the function of those organic matrices in calcium carbonate crystallization by in vitro and in vivo experiments. Results: We analyzed calcites synthesized in chitin gel with or without chitinolytic enzymes by using transmission electron microscope (TEM) and atom probe tomography (APT). TEM observations showed that grain boundary was more induced as concentration of chitinolytic enzymes increased and thus, chitin became thinner. In an optimal concentration of chitinolytic enzymes, small-angle grain boundaries were observed. APT analysis showed that ion clusters derived from chitin were detected. In order to clarify the importance of chitinolytic enzymes on the formation of the prismatic layer in vivo , we performed the experiment in which chitinase inhibitor was injected into a living Pinctada fucata and then analyzed the change of mechanical properties of the prismatic layer. The hardness and elastic modulus increased after injection of chitinase inhibitor. Electron back scattered diffraction (EBSD) mapping data showed that the spread of crystal orientations in whole single crystal also increased by the effect of inhibitor injections. Conclusion: Our results suggested that chitinolytic enzymes may function cooperatively with chitin to regulate the crystal growth and mechanical properties of the prismatic layer, and chitinolytic enzymes are essential for the formation of the normal prismatic layer of P. fucata . ion beam (FIB) (FB 2100, Hitachi). The samples were locally coated with tungsten and trimmed using Ga ion beam and thinned down to a final thickness of 200 nm. These electron-transparent specimens were observed with a JEM-2010UHR TEM (JEOL) operated at 200 kV. The intracrystalline organic matrices were imaged by Fresnel contrast, not focusing on the organic matrices.


Results:
We analyzed calcites synthesized in chitin gel with or without chitinolytic enzymes by using transmission electron microscope (TEM) and atom probe tomography (APT). TEM observations showed that grain boundary was more induced as concentration of chitinolytic enzymes increased and thus, chitin became thinner. In an optimal concentration of chitinolytic enzymes, small-angle grain boundaries were observed. APT analysis showed that ion clusters derived from chitin were detected. In order to clarify the importance of chitinolytic enzymes on the formation of the prismatic layer in vivo, we performed the experiment in which chitinase inhibitor was injected into a living Pinctada fucata and then analyzed the change of mechanical properties of the prismatic layer. The hardness and elastic modulus increased after injection of chitinase inhibitor. Electron back scattered diffraction (EBSD) mapping data showed that the spread of crystal orientations in whole single crystal also increased by the effect of inhibitor injections.

Conclusion:
Our results suggested that chitinolytic enzymes may function cooperatively with chitin to regulate the crystal growth and mechanical properties of the prismatic layer, and chitinolytic enzymes are essential for the formation of the normal prismatic layer of P. fucata.

Background
Molluscan shells consist of more than 90% calcium carbonate and small amounts of organic matrices, which interact with calcium carbonate and regulate its crystallization.
The organic matrices may be classified into three groups: insoluble organic matrices (IOMs) such as chitin [1][2][3][4], soluble organic matrices (SOMs) binding IOMs with a domain that can interact with IOMs [5][6][7][8], and SOMs that exist freely [9][10][11]. IOMs generally provide the scaffold for crystallization, and SOMs contain acidic domains such as Asp-rich regions, which control the crystal polymorph, morphology, and growth. Such organic-inorganic interactions can generate a fine microstructure. However, most of the detailed structures and functions of these organic molecules are still unknown.
Pinctada fucata, a bivalve used for pearl cultivation, has a microstructure composed of a prismatic layer in its shell. The prismatic layer exists on the outer layer of shells, and is comprised of calcite prisms surrounded by an organic framework, which appears as a honeycomb structure [12,13]. The radius and length of one calcite prism are approximately 50 µm and 150 µm, respectively (Fig. 1a, b). Previous studies have reported that the calcite prism of P. fucata exhibits unique characteristics. Transmission electron microscopy (TEM) has revealed that the calcite prism of P. fucata is not a complete single crystal, but contains small angle grain boundaries, which are slight crystal misorientations [14,15]. TEM has also shown that organic matrices are localized in a network, and that the locations of the organic networks correspond to small angle grain boundaries, indicating that the organic networks might contribute to the formation of small angle grain boundaries [16], which may improve the strength and toughness of the prismatic layer. A calcite prism of P. fucata is harder than a calcite prism of other species such as Atrina pectinata, which does not contain small angle grain boundaries as in abiogenic calcite crystals. Small angle grain boundaries could prevent the penetration of cracks when a force is applied to the crystals [17]. Basically, calcite easily fractures along a {104} cleavage plane because of the high density of atoms in this cleavage plane.
Although the high density of atoms within one plane stabilizes it by retaining the atoms within one plane, it also reduces the attachment energy which holds parallel planes together [18]. Thus, small angle grain boundaries, which are distortions of atomic arrangements, may change energy balances and prevent penetration of cleavage, leading to improvement of mechanical properties of calcite prisms. Surprisingly, although it is difficult to artificially mimic the superior microstructure of calcite prisms, P. fucata easily produces these fine structures under normal temperature and pressure using organic matrices.
Recently, we characterized the mechanism of formation of small angle grain boundaries by focusing on the components of the organic networks. From organic networks, chitin was identified as the primary component [19]. In addition, the chitinolytic enzymes that degrade chitin were also identified when the proteins that formed a complex with chitin were analyzed. In vitro experiments revealed that when calcium carbonate was crystallized in chitin gel treated with chitinolytic enzymes, the lattice distortion of the synthesized crystal increased, depending on the concentration of enzymes. This finding indicated that chitin and chitinolytic enzymes might work cooperatively to induce small angle grain boundaries. In this study, we analyzed the synthesized calcium carbonate crystals in detail to reveal how the chitin condition was changed by degradation with chitinolytic enzymes. Furthermore, to demonstrate the importance of chitinolytic enzymes in the formation of the prismatic layer, we injected an inhibitor of chitinolytic enzymes into living P. fucata and examined whether the mechanical properties of calcite prisms were changing by inhibiting the activity of chitinolytic enzymes. These results provide new insights into the field of biomineralization.

Identification of components in an intracrystalline organic network
Previous reports have shown that small angle grain boundaries correspond to the position of an intracrystalline organic network, suggesting that the intracrystalline organic network was related to the formation of small angle grain boundaries [16]. It was therefore important to characterize the structure of components in an intracrystalline organic network. The extraction scheme is described in Figure 1c. To extract organic matrices from the intracrystalline organic network, the calcite prism was decalcified with acetic acid and the resulting organic matrices were classified as acid-insoluble or acid-soluble.
The organic network was mainly comprised of acid-insoluble organic matrices. Generally, the acid-insoluble organic matrices from the molluscan shell are comprised of complexes of proteins and sugars such as chitin. The protein components can be extracted by detergents such as sodium dodecyl sulfate (SDS). Detergent-soluble organic matrices (DSOMs) were resolved by SDS-PAGE and detergent-insoluble organic matrices (DIOMs) were analyzed by Fourier-transform infrared spectroscopy, which revealed that DIOMs were composed of chitin, with polysaccharides comprised of N-acetyl-Dglucosamine (NAG) (Fig. 1d) [19]. Two types of chitin polymorphs have been reported.
The chains of β-chitin bind each other in an antiparallel arrangement [20]. However, βchitin is involved in the parallel chain containing water between chains [21]. Chitin identified from molluscan species to date is β-chitin. When the band patterns of acidsoluble organic matrices were compared to that of DSOMs, a specific band with high molecular weight components was detected in DSOMs (Fig. 1e). The band was removed and digested with trypsin to determine the amino acid sequence by liquid chromatography-mass spectrometry/mass spectrometry. The identified protein in this band was a complex of chitinolytic enzymes (gene ID: pfu_aug1.0_10761.1_31980 and pfu_aug1.0_6745.1_67528). The chitinolytic enzymes can degrade the chitin chain. A BLAST search identified these proteins as chitinase and β-N-acetylhexosaminidase; the enzymes were named PfCN and PfCB, respectively. PfCN is composed of 466 amino acids with a molecular mass of 53.9 kDa, and belongs to the glycoside hydrolase (GH) family 18. The sequence of PfCN includes a signal peptide in the N-terminus and GH18 domain (Fig. 1f). PfCB is composed of 662 amino acids with a molecular mass of 72 kDa and belongs to the GH20 family. The sequence of PfCB includes the GH20 domain (Fig.   1f). Chitinolytic enzymes can be classified into two groups: exo or endo. β-Nacetylhexosaminidases, which belong to the GH20 family, are exoenzymes that degrade chitin from the nonreducing end to produce NAG. Chitinases of animals mainly belong to the GH18 family (some chitinases from plants and bacteria belong to the GH19 family) and degrade chitin randomly to produce (NAG)2 [22,23].
To investigate the true activity of chitinolytic enzymes, DSOMs were applied to native PAGE containing glycol chitin in the gel. These previous results suggested that the complex of chitinolytic enzymes kept enzymatic activity in the shell [19]. Although chitin has been identified in various kinds of molluscan shells, which provide scaffolds for crystallization [1,4,24], only a few chitinolytic enzymes have been identified from molluscan shells, and their functions in shells are unknown [25][26][27]. Because chitinolytic enzymes binding chitin were identified from calcite prisms and possessed the ability to degrade chitin in the shell, we identified the cooperative functions of chitin and chitinolytic enzymes involved in the formation of small angle grain boundaries in calcite prisms.

Synthesis of calcium carbonate crystals in chitin gel treated with chitinolytic enzymes in vitro
During formation of the prismatic layer, the organic frameworks are constructed before calcification. The space surrounded by the organic frameworks is filled with organic gel solution comprised mainly of chitin and matrix proteins, then calcium carbonate is crystallized with organic gel solution in the space [13,28]. We identified the chitinolytic enzymes as matrix proteins in the organic network. The organic gel solution containing chitin and chitinolytic enzymes can develop into organic networks during formation of the prismatic layer and affect calcite crystallization. To investigate how chitin and chitinolytic enzymes affect the formation of calcite crystals, we performed an in vitro calcium carbonate crystallization experiment using chitin gel and chitinolytic enzymes (Fig. 2a).
The chitin gel was prepared by dissolving chitin powder in methanol saturated with calcium chloride dehydrate. The chitin solution was then added to a large amount of water for gelatinization. Calcium carbonate was crystallized in the chitin gel with or without yatalase, a commercially available chitinolytic enzyme containing the activities of both GH18 and GH20 produced by a strain of Streptomyces. Because we identified both GH18 and GH20 family chitinolytic enzymes in the prismatic layer, yatalase was an appropriate mimicking agent of the chitinolytic enzymes in this layer. The treated chitin gel was washed with calcium chloride solution and incubated in the desiccator with ammonium carbonate powder to supply CO2 gas, which dissolved in the chitin gel and became a carbonate ion. Electrostatic coupling of the carbonate ion and calcium ion in the chitin gel induced crystallization of calcium carbonate. Following crystallization, calcium carbonate crystals were collected by dissolving only the chitin gel with sodium hypochlorite; the remaining crystals were observed using scanning electron microscopy (SEM). The normal shape of calcium carbonate crystals (a typical rhombohedral calcite) was formed in the chitin gel without yatalase treatment (Fig. 2b). Conversely, the shape of calcite crystals became polygonal in the chitin gel treated with 0.12 mg/mL chitinolytic enzymes (Fig. 2c). In the chitin gel treated with 1.2 mg/mL chitinolytic enzymes, the shape of the calcite crystal was completely changed and appeared to be round (Fig. 2d).
Chitin likely aggregated between polysaccharide chains via hydrogen bonding.  [29]. When calcium carbonate crystals synthesized in the chitin nanofiber solution were observed using SEM, the shape of the calcite crystals was polygonal, similar to crystals in the presence of 0.12 mg/mL enzymes (Fig. 2e). This change in crystal morphology was likely because of the potential change of chitin fibers attaching to the growth surface of crystals. The thick chitin fiber (0 mg/mL) may not have as much potential because it could not attach to the growth surface, resulting in a few steps involving substances absorbed on the growth surface. However, as the chitin fibers became thinner, the number of crystal steps increased. In addition, each step was not aligned, and slightly rotated. Thus, the crystals gradually formed a round shape because the chitin fibers become thinner.
These results suggested that chitin directly interacted with and affected the morphology of calcium carbonate. Chitin is a polymer composed of NAG, which possesses an acetyl amino group at C6, whereas cellulose is composed of glucose molecules with hydroxyl groups. To investigate the specific function of chitin in calcium carbonate crystallization, we used a cellulose gel to compare the results. When calcium carbonate was crystallized in a cellulose gel with or without treatment of cellulase at concentrations of 0, 0.12, and 1.2 mg/mL, rhombohedral crystals were observed at all concentrations, indicating that cellulose had no effect on crystallization (Fig. S1).
Because an acetyl amino group has both negative and positive charges, the acetyl amino group may interact with calcium and carbonate ions in calcium carbonate crystals during crystal growth.

Analysis of synthesized calcium carbonate crystals
TEM bright field images of the cross-sections prepared by a focused ion beam revealed that the calcite crystal formed in the chitin gel without yatalase treatment contained many thick chitin fibers with a bright contrast, and typical interference fringes of equal inclination of single crystals were observed (Fig. 3a), indicating that chitin gel without chitinolytic enzyme treatment did not affect calcium carbonate. In addition, electron diffraction patterns of selected areas indicated that the calcium carbonate crystals synthesized in these conditions were single crystals (Fig. S2). By contrast, following 0.12 mg/mL chitinolytic enzyme treatment (Fig. 3b), slightly disordered lines of fringes were partially observed. If the sample was a perfect single crystal, equal inclination interference fringes would have shown clear lines. At a concentration of 0.12 mg/mL, fringe lines were interrupted by lattice distortions such as small angle grain boundaries.
Furthermore, the positions of partial interrupted fringes corresponded to that of chitin fibers. Thus, chitin might induce the small angle grain boundaries in calcium carbonate crystals. At 1.2 mg/mL (Fig. 3c), thin chitin fibers were embedded with calcium carbonate.
The selected area electron diffraction pattern of the crystal was vastly different between the boundaries of chitin fibers. Combined with the electron diffraction patterns, the calcite crystal was polycrystal. In the chitin nanofiber solution, the fringes were disordered throughout the entire area (Fig. 3d). Based on these results, TEM observation also supported the possibility that lattice distortion increased as the chitin fibers were degraded by chitinolytic enzymes, because the increasing surface areas of the chitin fibers strengthened the physical and/or chemical interactions between calcium carbonate and the chitin fiber. This strong interaction may have allowed chitin fibers to attach to the crystal growth front and inhibit growth of the crystal face in a random manner to induce a rounded shape.
To examine the interaction of chitin with calcium carbonate crystals in detail, these crystals were analyzed by atom probe tomography (APT). APT is a useful technique to determine the three-dimensional (3D) distribution of atoms or ions in different materials [30][31][32]. In APT analysis, the specimen shaped into a needle-like tip with a radius of 50-100 nm by a focused ion beam was ionized by field evaporation to detect atoms of ions with 2D ion distributions, followed by a 3D reconstruction. In Fig. 3e-h, the 3D reconstruction images show that ion clusters of COH + (yellow) and COH2 + (pink) derived from chitin are detected within calcium carbonate ions (red). Fragments of chitin were partially detected in calcium carbonate at a concentration of 0 mg (Fig. 3e). Most of signals were localized in the edge of sample. When the thick chitin fibers were embedded in the tip, the tip was broken, because thick chitin fiber is soft and decreased the strength of the tip. The localized signal of the edge may be derived from the neighbor chitin thick fiber outside the tip. On the other hand, chitin signals increased and were distributed in wide area within calcium carbonate at higher concentrations (Fig. 3f, g). Since both homogeneous and continuous signals were observed in all area, these signals were likely to be both chitin oligomers and chitin nanofibers degraded by chitinolytic enzymes. These APT results support the hypothesis that chitin without the treatment of enzyme has no effect on the crystal formation and the effect of the chitinolytic enzymes on the crystallization becomes higher as the concentration increases. In case of the artificial chitin nanofiber produced by chemical and physical procedure, high density signals derived from chitin were detected (Fig. 3h). This probably indicates that the artificial chitin nanofiber has high dispersibility and can be involved in the tip, because the nanofiber does not affect the strength of the tip.

The influence of chitinase inhibitor injection on the formation of calcite prisms of P. fucata in vivo
As previously mentioned, we showed that chitinolytic enzymes may play important roles in formation of the P. fucata prismatic layer. To show the importance of chitinolytic enzymes on the formation of the prismatic layer, allosamidin, an inhibitor of GH18 chitinase [33], was injected into a living P. fucata every 3 days for a month. Chitinases hydrolyze β-1,4 linkages of the chitin polymer randomly to produce chitin oligomers, especially chitobiose, followed by β-N-acetylhexosaminidases, which degrade chitin oligomers into NAG because β-N-acetylhexosaminidases prefer short oligosaccharides [34,35]. It is therefore important to use allosamidin to evaluate the effect of chitinolytic enzymes. Following injection, P. fucata was grown in natural sea water. Measurement of chitinase activity from mantle extracts showed that allosamidin suppressed activity for 1 month, indicating that the shell was formed under conditions of low chitinase activity (Fig. S3). If our hypothesis that chitinolytic enzymes degrade chitin and affect crystal growth was correct, the prism should be affected by inhibition of activity. Fig. 4a and b shows microdiffraction mapping of the longitudinal cross-section of prismatic layers along the c axis and line profiles of one prism. In the normal prismatic layer without inhibitor, the color of one prism was found to be uniform from mapping the image and angle spread of one prism measured from the line profile within 3 degrees; there was minimal crystal disorientation within one prism. However, in the prismatic layer treated with inhibitor, the crystal orientation of one prism was found to be gradually changed, with the mapping image and angle spread of one prism approximately 10 degrees. The reason for the increase of angle spread was likely due to intracrystalline chitin thickening from inhibiting chitin degradation. We previously reported that thicker chitin fibers inside the calcite prism treated with allosamidin compared with those of normal calcite prism were observed using TEM. Chitin fibers that increased in thickness may have contributed to crystal disorientation.
A previous study reported that a biotic calcite crystal with slight crystal distortion was stronger than an abiotic calcite crystal without crystal distortion [36] because the abiotic calcite is easily cracked under high pressure. The crystal distortion blocked transmission of strength to inhibit propagation. Because allosamidin treatment caused crystal distortions in calcite prisms, we investigated how such crystal distortions influenced mechanical properties such as hardness and the elastic modulus of calcite prisms. To examine the mechanical properties of calcite prisms, nanoindentation analysis of the prism surface was conducted. A nanoindentation force-displacement curve was drawn from the applied load versus depth profile, and hardness and the elastic modulus were obtained from this curve. The force-displacement curves are shown in Fig. 4c and d. When a force was loaded up to 100 mN, the maximum displacement was approximately 1-3 µm in the non-injection treatment (measurements at eight points). However, with allosamidin treatment, the maximum displacement was approximately 1-1.5 µm (measurements at six points). In addition, the cracks caused by indentation following allosamidin treatment were blocked to a greater extent than the non-injection treatment (Fig. S4), and both hardness and the reduced elastic modulus of prisms following allosamidin treatment increased by approximately 80.6% and 105%, respectively, compared with those of the non-injection treatment (Fig. 4e). These results indicated that inhibition of chitinase activity strengthened prisms, likely because crystal defects Thus, reducing the dislocation motion results in enhanced mechanical strength. A grain boundary caused by a disorientation or enclosing substance is one factor that can reduce the dislocation motion. This theory could enable us to examine why a biotic calcite containing slight crystal distortions was stronger than abiotic calcite. In the present study, because micro-EBSD mapping revealed crystal disorientation in prisms treated with an inhibitor, this crystal defect may have resulted in harder prisms than normal.

Conclusion
This study showed a novel mechanism of calcite prisms during shell formation. We

Synthesis of calcium carbonate in the chitin hydrogel
Chitin hydrogel was prepared according to the previous method [39]. The prepared chitin hydrogel was treated with Yatalase (an enzyme complex containing chitinase and chitobiase activities from Corynebacterium sp. OZ-21; TaKaRa) as chitinolytic enzyme.
Yatalase was added to this chitin hydrogel in various concentrations and stirred for 24 hours. After reaction, chitin hydrogel was filtered and washed with 10 mM calcium chloride to remove Yatalase. The washed chitin hydrogel was spread on a plate (33.5 mm x 33.5 mm). This plate was put into desiccator filled with the gas of 5 g of ammonium carbonate (Kanto Chemical), followed by the calcium carbonate crystallization in the chitin hydrogel for 24 h. Chitin hydrogel was dissolved with 50% sodium hypochlorite to collect the calcium carbonate crystals.

Calcium carbonate crystallization using chitin nanofiber
1.1% (w/v) chitin nanofiber in 0.5% acetic acid solution was prepared [29]. The chitin nanofiber solution was neutralization with 1 M NaOH. 10 mM calcium chlorite solution was added to the chitin nanofiber solution. Calcium carbonate was then crystallized in the chitin nanofiber solution using the same method described above. The formed calcium carbonate crystals were observed with SEM and the cross-sections prepared by FIB were observed by TEM.

TEM observations of the calcium carbonate crystals
The calcium carbonate crystals synthesized in the chitin hydrogel were washed with distilled water to remove the sodium hypochlorite. The inside of these calcium carbonate crystals was also observed by TEM to investigate the inside detailed microstructure.
Electron-transparent thin specimens for TEM observation were prepared using a focused

Atom probe tomography
The method of ATP analysis was performed according to previous study [32]. Briefly, the tip specimens of the synthesized calcium carbonate crystals were prepared with a radius

Competing interests
The authors declare that they have no competing interests.

Availability of Data and Materials
All data in this study are included in this published article and its supplementary file.