Architecture of crossed-lamellar bivalve shells: the southern giant clam (Tridacna derasa, Röding, 1798)

Tridacna derasa shells show a crossed lamellar microstructure consisting of three hierarchical lamellar structural orders. The mineral part is intimately intergrown with 0.9 wt% organics, namely polysaccharides, glycosylated and unglycosylated proteins and lipids, identified by Fourier transform infrared spectrometry. Transmission electron microscopy shows nanometre-sized grains with irregular grain boundaries and abundant voids. Twinning is observed across all spatial scales and results in a spread of the crystal orientation angles. Electron backscatter diffraction analysis shows a strong fibre texture with the [001] axes of aragonite aligned radially to the shell surface. The aragonitic [100] and [010] axes are oriented randomly around [001]. The random orientation of anisotropic crystallographic directions in this plane reduces anisotropy of the Young's modulus and adds to the optimization of mechanical properties of bivalve shells.


Introduction
Bivalve shells are complex biocomposites consisting of calcium carbonate intimately intergrown at the nanoscale with organic  macromolecules [1,2]. This composite nature creates enhanced material properties, for example high mechanical strength [3] and fracture toughness [4,5], which optimize shell stability and protective function for the organism [6]. Much recent research has focused on the nacreous shell structure in molluscs, while other shell structures in this phylum, such as the most widespread crossed lamellar structure [7], are yet to receive comparable attention. Here, we present one of the first in-depth characterizations of both the inorganic and the organic parts in Tridacna derasa (southern giant clam) shells. Furthermore, we combine here electron backscatter diffraction (EBSD) with transmission electron microscopy (TEM) analysis to identify some of the multi-scale strategies for the optimization of mechanical properties across all structural hierarchies in the shell.

Structure and micro-texture of Tridacna derasa shells
Tridacna derasa shells are entirely aragonitic and consist of an approximately 10 mm thick massive outer layer and a slightly translucent inner layer with visible growth increments (figure 1a,c). The shell comprises crossed lamellar shell structure, which is the most common structure of mollusc shells and has been described in detail by a number of authors (e.g. [7][8][9][10][11][12][13]). With very few exceptions, shells with this structure are aragonitic rather than calcitic [14]. Crossed lamellar shells from different species vary in structural arrangement but bear a basic architectural similarity [15]: aragonite grains are arranged in hierarchically organized lamellae (figure 1b,d) with alternating orientations at an angle of approximately 70-90°depending on the species [7,12]. Three or four hierarchical orders can be identified and growth twinning of the aragonite crystals is very common (e.g. [12,16] The first structural hierarchy in T. derasa shells comprises a series of undulating aragonite bands 20-50 µm thick (first-order lamellae, figure 1b) consisting of nanometre-thin stacked second-order lamellae. These sheets of stacked lamellae are oriented at an angle of approximately 70°to each other and are perpendicular to the shell surface. Each sheet of second-order lamellae, in turn, consists of nanometresized parallel aragonite laths, which form the third order of hierarchy (figure 1b).
Tridacna species are known to form daily growth lines [17,18] with growth increments around 15 µm in width. Growth lines in molluscs usually have increased organic content compared with the increments between a set of lines and can be visualized using histochemical staining methods (figure 1e).

Material and methods
The southern giant clam, T. derasa (Röding 1798) (Mollusca: Bivalvia), is the second largest species in the family Tridacnidae, reaching shell lengths of up to 520 mm [19]. Tridacnidae occur naturally in the tropical and subtropical waters of the Indo-Pacific and host photosymbiotic algae in their tissues [20]. Shells of T. derasa, cultured on Ishigaki Island, Okinawa, Japan (for environmental details see [18]), were used for structural analyses of the inorganic part of the shells, while organic matrix analysis was performed using a recently alive shell from One Tree Island, Queensland, Australia.

Scanning electron microscopy and electron backscattered diffraction
Broken pieces of shell were imaged with a Leo Gemini 1530 field-emission secondary electron microscopy (SEM) instrument (Carl Zeiss, Germany) at the Max-Planck Institute for Chemistry, Germany. Samples were mounted on aluminium stubs using a conductive carbon tape. All samples were studied uncoated with an accelerating voltage of 5 kV, a sample current of 2 pA and a vacuum pressure of 5 × 10 −6 mbar.
Crystallographic preferred orientations (CPO) in the shell were determined by automated indexation of EBSD in a scanning electron microscope [21,22] using an EDAX TSL Digiview 3 EBSD camera and an OIM DC 5.0 detector. The sample was polished using standard methods with diamond pastes of different grain sizes down to 0.25 µm and a final step of chemical-physical polishing using a neoprene polishing cloth and an alkaline solution of colloidal silica for 1 h. The EBSD measurements were conducted in an area of the shell along the axis of maximum growth (figure 1a). Analyses were carried out under low vacuum (10 Pa of H 2 O) using the following parameters for SEM: 15 kV accelerating voltage; 8 nA beam current; 12 mm working distance; step size of 1 µm and 70°sample tilt. At these conditions, the electron beam size is 4 nm and approximately 90% of all diffraction patterns could be indexed.
Post-acquisition treatment included the standardization of the confidence index (CI) of different points and CI correlation between neighbouring points. 'Grain' dilation was carried out in three steps considering the grain tolerance angle of 10°and a minimum grain size of 10 pixels. Grain sizes as observed by TEM are usually in the nanometre range in biominerals (e.g. figures 4-6), hence the chosen 'grain' size cut-off for EBSD defines domains of several crystallographically well-aligned nanograins, rather than individual aragonite grains. These domains have misorientation angles less than 10°and are here termed 'first-order domains'.
The pseudohexagonal symmetry effect on aragonite caused by a rotation of 60°around [001] was also corrected. Data with CI > 0.1 are plotted in pole figures (figure 2d and figure 9b), which are stereograms with axes defined by an external reference frame using the shell length growth direction (GD), the direction of the growth lines (GL) and the axis normal to these features. Accumulation of points around a specific direction in the pole figures (pole maxima) shows a degree of texture in the polycrystalline material, quantified according to the colour scales in the figures. The rotations of the crystallographic dataset and plots of pole figures were carried out using the MTEX toolbox for Matlab [23,24].

Transmission electron microscopy
TEM foils approximately 10 by 15 µm in length and 0.15-0.20 µm thickness were cut from the polished section of the sample using an FEI FIB200 focused ion beam device (FIB) following procedures given in Wirth [25]. Six foils were cut from the inner and outer layers of the shell, either parallel or perpendicular to the growth lines. Samples were placed on a carbon-coated Cu grid without further carbon coating (ex situ lift out method). The FIB milling method involves sputtering the material surrounding the platinumprotected target area with gallium ions. This process can heat the target area, and drive amorphization through Ga implantation in the surface of the material [26]. Sample heating is proportional to the beam current, and the extent of amorphization is proportional to the beam energy; both depend on the angle of  beam incidence during milling [27]. To avoid heating of the sample, we used 30 keV with a beam current of 11 pA and an angle of incidence of 1.2°. At these conditions beam heating during FIB milling is less than 10 K [28] and sample amorphization is minimal. As the foils are thicker than 100 nm, the major part of the foil is thus not affected by ion implantation. If amorphization were a significant problem in the foils, Debye-Sherrer diffraction rings would be present in all collected diffraction patterns, but these features were not observed.
TEM imaging and analysis were undertaken with a FEI Tecnai™ G2 F20 X-Twin transmission electron microscope with a field emission gun source, operating at 200 kV acceleration voltage. A Gatan Tridiem™ filter allowed energy-filtered imaging, applying a 20 eV window to the zero-loss peak for all bright-field images in this study. Images were taken either in scanning TEM (STEM) mode or in high-angle annular dark field (HAADF) mode with a 330 mm camera length. At these conditions imaging is possible with z-contrast, diffraction, thickness and density contrast.
Great care was taken to minimize radiation damage to the material during TEM analysis. This involved low-dose analysis and a visual monitoring protocol developed for biominerals [29,30]. Foils were analysed in STEM mode, rapidly scanning using a small spot size and assigning the beam to areas outside the sample to avoid electron irradiation damage. At the start of the analytical session for each FIB foil, a rapid overview picture was taken using a defocused beam and this was repeated after STEM scanning and high-resolution electron microscopy analysis and at the end of each analytical session to scan for beam damage. All high-resolution TEM analyses were carried out at the end of the analytical session for each foil, using exposure times of 0.2 s and a spot size of 5. Using this protocol, irradiation damage was only observed on a few occasions, manifested either as holes from the electron beam or as localized small amorphized areas where a STEM scan had been carried out. Analyses and images from these areas were discarded from the dataset.

Thermogravimetry
The total amount of organic shell matrix was determined by thermogravimetric analysis (TGA) with a TGA 2050 thermogravimetric analyser (TA Instruments, USA) at the Department of Chemistry and Biomolecular Sciences, Macquarie University. Approximately 30 mg of powdered and sieved sample from the inner part of the shell (250 µm mesh size) was measured in a temperature interval from 30°C up to 1000°C at 10°C min −1 steps under a nitrogen atmosphere.

Extraction methods
To characterize the organic matrix, the shell was decalcified in 6 N HCl after cutting and removing its outermost part, followed by a cleaning step that involved immersing the shell in 30% H 2 O 2 (Merck KGaA, Darmstadt, Germany) and rinsing with Milli-Q water. The solution was stored at 4°C after decalcification. The supernatant was extracted twice, first with dichloromethane and then with butanol (BuOH), both fractions were combined and reduced to dryness before storage at −35°C.
The lipid-lipoprotein fraction was extracted with methanol/dichloromethane (2/1) at room temperature for 40 h, ultrasonicated and dried in a nitrogen atmosphere.
An aliquot of the total organic matrix extract was taken up in dimethylacetamide containing 5% lithium chloride [31], centrifuged and filtered. The specific optical activity of the filtrate was analysed with a JASCO P-1010 polarimeter (JASCO, Tsukuba, Japan) at the Department of Chemistry and Biomolecular Sciences, Macquarie University. A cell line of sodium D at 589 nm was used as the filter at room temperature of 21 ± 1°C.

Infrared spectroscopy
Fourier transform infrared (FTIR) spectra of different extracted and dried organic matrix fractions were measured with a Thermo Nicolet iS10 FTIR spectrometer (Nicolet, MA, USA) equipped with attenuated total reflection along with a smart performer assessor at the Department of Chemistry and Biomolecular Sciences, Macquarie University. Spectra were acquired between 4000 and 500 cm −1 with a resolution of 8 cm −1 and 64 accumulations. Backgrounds were recorded at the beginning of the analytical session and approximately every half hour.

Electron backscattered diffraction analysis
EBSD was carried out on a polished area of the inner layer of the shell situated between, but not overlapping with, two organic-rich annual growth lines (approximate location outlined in figure 1a). shell and amounts to approximately 26% of the 'grain' (i.e. first-order domain) boundaries (figure 2c). This value is derived from analysis of the boundaries between first-order domains with a misorientation angle of 60°in the CPO map (figure 2a), which is close to the angle of twinning on {110} in aragonite. It should be noted that the estimated value of 26% is a lower limit for the total amount of twin boundaries, because its precision is determined by the measurement conditions and the spatial resolution of the EBSD analysis. Nevertheless, even this rough estimate shows that aragonite twinning is very common in the shell across all structural hierarchies.
The CPO of the shell shows a strong alignment of [001] axes parallel to the growth direction of the shell (GD) as seen in the pole figures (figure 2d), with a maximum of about 12× the uniform distribution. The [100] and [010] axes are distributed at random along continuous and broad single girdles normal to GD, with a weak tendency for [100] to be parallel with the GL direction. Hence, the aragonite crystals are strongly co-aligned in the [001] axial direction but random in radial direction (i.e. concerning their [100] and [010] axes). These are the typical characteristics of a fibre texture. Figure 2a shows that first-order domains forming several second-order lamellae within a given firstorder lamella are co-oriented, forming areas of uniform or similar colour (arrow in figure 2a). These larger areas with small internal misorientation angles are generally around 50 µm in size and are separated

Transmission electron microscopy
We analysed six TEM foils cut from the same polished section for which EBSD measurements were performed (figure 1a). Prominent characteristics observed in the TEM analyses are the typical particulate nature, well described for natural biominerals (e.g. [29,32,33]) associated with multi-scale porosity throughout the shell (figures 4-6). The outer shell layer (figure 4) displays micrometre-sized cavities between the third-order laths ( figure 4a,b), while the inner shell is less porous at the micrometre scale ( figure 5a,b). By contrast, the inner shell shows submicrometre porosity within the third-order aragonite laths (figure 5c,d), which we found less frequently in the foils cut from the outer part of the shell. We observed numerous and variably sized, irregular voids (approx. 30-130 nm in size) rimming the third-order lath boundaries, with smaller voids (approx. 5-15 nm in size) within these aragonite laths ( figure 5c,d, arrows). They appear to be focused along the outer areas of the laths in the inner shell layer, resulting in irregular 50-100 nm wide concentration zones of voids. While the foil cut perpendicular to the long axis of the aragonite laths (figure 5c) shows these voids apparently distributed at random, the  rims of the laths cut parallel to their longest axes in figure 5d are not completely enclosed by a focus zone of voids. It is possible that this apparent preferential focus in figure 5d is a sectioning effect from foil preparation. Crystal shapes in both layers of the shell are highly irregular (figure 6a-f ) and aragonites commonly display variable and strongly 'speckled' diffraction contrasts in TEM bright and dark field imaging, where different areas within an individual grain display sharply different diffraction contrasts (e.g. figure 6b, arrow). Polycyclic/polysynthetic twinning is common at the nanoscale in both shell layers (figures 5c and 6e,f ).
All areas studied by high resolution TEM were found to be crystalline, even the rim areas of the grains (figure 6d), which contrasts with nacroprismatic shells, where amorphous areas are observed [30,34], amounting to 10 at% in some shells [35]. Nevertheless, results from solid-state nuclear magnetic resonance spectroscopy on the same T. derasa shell sample indicate an overall amount of 3-7 at% of amorphous calcium carbonate [35], showing that while this phase is present in the shell, it was apparently not sampled by any of the TEM foils in this study.

Thermogravimetric analysis
Thermogravimetry was used to quantify water and organic contents in a powdered sample of the shell. The total organic matrix amounts to 0.94 wt% as calculated from the integrated mass loss in the temperature range 150-500°C (figure 7). Peaks at around 300°C are attributed to the decomposition of organic macromolecules. Aragonite converts to calcite at approximately 500°C, as verified by FTIR spectrometry, and decomposition to CaO and CO 2 is completed at around 800°C (figure 7).    Table 1. Band assignments for the main bands in the FTIR spectra (cm −1 ) of the shell organic matrix components for T. derasa. Band assignments carried out using data from [36][37][38].   [42,43], and 952-980 cm −1 [39] are indicative of sugars.

Characteristics of the multi-scale shell architecture and organic moiety
Compared with the prism-nacre microstructure of mollusc shells, the crossed-lamellar architecture of the T. derasa shell is highly mineralized with only 0.9 wt% total organic content compared with greater than 3 wt% in nacre [45]. Its structure is distinctly different and overall more complex than the nacro-prismatic structure [12], with three hierarchy orders of aragonite laths creating an interlocking fabric reminiscent of plywood (figure 1) [7,8]. Some authors have interpreted polycyclic aragonitic growth twins occurring in the third-order structure as a fourth hierarchical order (e.g. [15,46,47]).
In T. derasa shells, the first order lamellae run approximately perpendicular to the growth layers (GL in figure 2), and radially with respect to the shell surface. They comprise thin particulate lamellae with high aspect ratios (second-and third-order structures: figure 1b), showing voids along grain boundaries, and within grains at the nanoscale (figure 5c,d) that contain organic macromolecules [48]. In contrast to nacro-prismatic bivalve shells, whose architecture consists of well-defined inter-crystalline organic matrix (the so-called interlamellar membranes) interlayered with mineral incorporating intra-crystalline organic matrix molecules, the crossed-lamellar T. derasa shell does not show a similarly prominent organic inter-crystalline framework [49].
Within the first-order lamellae, the CPO map (figure 2a) shows areas of second-order lamellae that are highly co-oriented, meaning that crystal co-orientation of aragonites for these areas is coherent across the organic layers between the aragonite laths as well as the organic growth lines shown in figure 1e. Similar aragonite tablets highly co-aligned across their organic envelopes occur in nacre [50] and are thought to form by a combination of epitaxial crystallization via mineral bridges across the organic envelopes [51] as well as due to competition for space upon growth [52]. It is conceivable that a similar model could be proposed for the formation of the domainal organization in crossed-lamellar structures.
Voids are a characteristic feature of biominerals and are common in bivalve shells irrespective of their microstructure [3,48,53,54]. The voids shown by the TEM analysis (figures 4-6) have larger diameters along the grain rims, are smaller within the aragonites (figure 5c) and are more common and larger in the outer shell layer (figure 4) than in the innermost layer of the shell (figure 5). Voids in nacre platelets are smaller (2-40 nm) than found in this study [48,53,54]. They are focused towards the inner part of the nacre platelets and form an approximately 50 nm wide void-depleted zone along the outer platelet rim adjacent to the interlamellar organic sheets [48]. This contrasts with our observations, where small voids are more common along the outer areas of each grain ( figure 5c,d).  While mollusc shell microstructures differ distinctly at the micrometre and millimetre scale, a common structural motif exists at the nanoscale. It is an almost universal characteristic of biominerals that they are granular at the nanoscale [35]. The granular features have sizes in the range of tens of nanometres, are embedded in an intergranular organic medium (or cortex), and are easily observed by phase-contrast atomic force microscopy (e.g. [13,32,55,56]), and their presence is supported using TEM (e.g. [29,30]): irregular grain boundaries (figures 5c and 6b,e,f ) and small voids (figure 5c,d) outline the shapes of the nanogranules in the TEM images. This granular nanostructure in mollusc shells is a consequence of their colloid-mediated, non-classical mode of growth [57,58].
The organic shell matrix in T. derasa consists of a mixture of polysaccharides as well as glycosylated and unglycosylated proteins and lipids (figure 8a), which is a rather typical general organic assemblage in mollusc shells [1,59]. The polysaccharides play a major role providing organic 'scaffolding' for the mineralized part (e.g. [1,60]), but may also have an active function of lowering the energy barriers for mineral nucleation [61]. The presence of polysaccharides, not only in nacro-prismatic mollusc shells but also in shells with crossed-lamellar microstructure, reiterates the applicability of the general biomineralization models that promote polysaccharides as the major organic template in this process (e.g. [62]).

Aspects of mechanical properties of the crossed-lamellar microstructure
The strength and toughness of shells are specifically distinct and superior to non-biogenic aragonite [5], with the crossed-lamellar shell microstructure displaying the highest fracture toughness [3]. These emergent mechanical properties of shells are a consequence of a combination of parameters including (but not limited to) the complex, multi-order hierarchy [3], their nano-granular texture [63], the organicinorganic nanocomposite nature of the material [64,65], and high flaw tolerance at the nanoscale [66].
Crystallographically, the crossed-lamellar structure of the T. derasa shell belongs to a family of similar, but not identical structures. In a comprehensive study of 40 mollusc species including a closely related species, Tridacna gigas, Almagro et al. [12] distinguished five crystallographic groups of crossed-lamellar structures. Among these, Tridacna shells display a strong fibre texture with highly oriented crystallographic c-axes, and randomly oriented a and b axes (figure 2c). In fact, the texture of T. derasa (figure 2d for scale of textural index) ranges among the strongest mollusc shell textures (e.g. [12,67,68]).
The aragonite crystallographic c-axes ([001]) in the T. derasa shell are oriented perpendicular to GL (figure 2c), thus, are radially oriented with respect to the shell surface. Aragonite single crystals are elastically anisotropic [69], with the anisotropy of the Young's modulus reaching values around 50% between the weakest and stiffest axes (figure 9a). The stiffest axis is [100] with a Young's modulus value