Multi-scale crystallographic ordering in the cold-water coral Lophelia pertusa

Lophelia pertusa is a widespread colonial cold-water coral which can form large three-dimensional habitats for benthic communities. Although it is known to construct an aragonite skeleton with optically opaque and translucent bands, details of its biomineralized structure are unclear. New crystallographic data obtained from Lophelia pertusa using electron backscatter diffraction (EBSD) reveal a remarkably high degree of multiscale self-ordering and provide unprecedented detail on crystallographic orientations within the coral skeleton. The EBSD data unequivocally demonstrate a self-regulated architecture across a range of spatial scales, resulting in a specific structure which contributes to the physical robustness of its skeleton and an evolutionary advantage in such habitats.

revised explanation of the origin of opaque and translucent bands is given based on textural variation of the aragonite crystals. We critically examine the degree of ordering at a variety of spatial scales and propose a new model for the self-regulation of observed structures. Understanding details of the skeletal architecture of L. pertusa provides an insight into the structural resilience of this cold-water coral, and how its possible demise may affect both biodiversity and carbon sequestration in the oceans. Our findings also have a bearing on the interpretation of geochemical proxy data from the coral skeleton used to examine signals of environmental change. Finally, we suggest that EBSD may be used to provide a similar level of insight into biogenic controls on crystallography in other important habitat-forming organisms.

Results
Crystal architecture of the skeleton wall. The skeleton wall of L. pertusa shows different microstructures on each side of the main RAD (Fig. 2). The external part of the corallite wall (corresponding to the TD area) consists of a three-dimensional array of fan-like sclerodermites crystallized from roughly equidimensional RADs.    Fig. 1d. The map shows the distribution of aragonite crystallographic axes with respect to the radial growth direction R of the corallite. The dominant bluish colour in the external part of the corallite wall (i.e. the TD area located in the lower half of the map) indicates that aragonite [001] axes are preferentially oriented parallel to R (within a tolerance of ~±45°), i.e. normal to the corallite margin. The crystallites coloured in reddish and greenish are those for which the axes close to R are [010] and [100], respectively, i.e. those intersecting the acquisition surface at high angles. A distinctive feature of the internal part of the corallite wall (upper half of the map) is the RAD lamella from which aragonite needles are radiating along their [001] axes. RADs appear black because they did not produce any diffraction pattern, possibly because of poor crystallinity and/or small crystal size. Optically opaque bands identified in Fig. 1d have been superimposed in light grey and delimited by dashed lines. R and L as in Fig. 1.
In contrast, the internal part of the corallite wall contains flat RAD lamellae, from which aragonite needles grow without forming fan-like microstructures.
Crystallographic anatomy of sclerodermites. A sclerodermite is typically 100-250 µm long and ~100 µm wide, with a symmetry axis oriented parallel to the radial growth direction (R) of the corallite. It is made of acicular aragonite needles, ~10 µm in diameter and ~50 µm in length, radiating from the RAD along their [001] axes. As a consequence of radiating growth, the angle formed by [001] axes and the sclerodermite symmetry axis increases towards the sclerodermite periphery up to several tens of degrees (Fig. 3a). Each needle contains smaller subunits called crystallites, <5 µm in size, which are slightly elongated along [001]. The distribution of misorientation angles between adjacent crystallites is predominantly non-random, with peaks forming at angles of 11.4°, 52.4°, and 63.8° (Fig. 3b). EBSD data show that crystallites belonging to the same needle are roughly dominated by three main crystallographic orientations distributed around [001] (Fig. 3c). This result can be further refined using the same data represented in pole figures (Fig. 3d), which show a strong clustering of aragonite [001] towards the radial growth direction R of the corallite, and clusters of [100] and [010] axes in a plane oriented normal to R. Misorientation angles of 63.8° are observed between each cluster, whereas secondary maxima of the same cluster have misorientation angles of 11.4° (the subtraction of the two values being equal to 52.4°). The predominance of these misorientation angles along with the observed distribution of crystallographic axes is attributed to pseudo-hexagonal polycyclic twinning of aragonite on {110}, which is typically characterized by rotations of 11.4°, 52.4°, and 63.8° around [001]. As a consequence of polycyclic twinning, most grain boundaries between aragonite crystallites are special twin boundaries (Fig. 3e,f). Non-twin grain boundaries are a minority but, when present, they generally correspond to bent sections of the needles. Bending does not exceed a few degrees and rarely occurs more than once or twice within a needle. This feature is significant enough, however, to account for the slight misorientation observed between the needle elongation axis and the orientation of [001] axes (Fig. 3d).

RAD distribution and transcurrent layering.
RADs are not distributed randomly within the TD area.
They are aligned within planes oriented normal to the corallite wall, preferably at the outer edge of opaque bands. This regular spatial distribution of the RADs results in a remarkable alignment of sclerodermite needles illustrated by a succession of continuous transcurrent layers (colour-coded in yellow and red in Fig. 4a) alternating longitudinally at a right angle to the corallite wall. Pairs of transcurrent layers, which result from the stacking of sclerodermites on top of each other, form the trabeculae described by several authors 17,26 . Boundary zones between adjacent trabeculae are RAD-free and correspond to areas in which needles impinge at high angles.
Pole figures of Fig. 4b indicate a non-uniform spread of aragonite [001] axes within transcurrent layers. This is illustrated by a strong clustering of axes at an angle of ~25° on either side of the sclerodermite symmetry axis. Interestingly, the clustering strength is significantly larger in the yellow colour-coded than in the red colour-coded transcurrent layers, with multiple of uniform distribution (mud) values of 9.73 for the former and 7.68 for the latter.

Specific crystallographic misorientations in opaque and translucent bands.
Opaque and translucent bands show subtle differences in crystallographic preferred orientations and microstructures. When considering misorientations <5°, crystallites located within opaque bands show higher average misorientation angles than those in translucent bands (Fig. 5a). This difference cannot be convincingly attributed to morphological factors, as crystallites from both opaque and translucent bands have similar grain area (Fig. 5b) and aspect ratio ( Fig. 5c) distributions. Pole figures generated from either band types are also quite similar (Fig. 5d), with aragonite [001] axes oriented preferentially towards the corallite radial direction R. Careful observation, however, indicates a more dispersed distribution of [001] axes within the RL plane for translucent bands than for opaque ones, with two maxima on each side of the R direction for the former, and only one maximum close to R for the latter (Fig. 5d). This small discrepancy in the distribution of [001] axes can be explained by the more widespread occurrence of well-developed needles in the translucent bands than in the opaque ones (s. Figs. 2 and 4). When looking at the distribution of misorientation angles (Fig. 5e), values of 63.8° are more common in the translucent bands, whereas values of ~5° are predominant in the opaque bands (the distribution of 11.4° and 52.4° being roughly equal). This feature is in agreement with the notion that the sclerodermites are slightly better developed in the translucent bands than in the opaque ones.
Aragonite growth normal to calcification interface. In the internal part of the crystallite wall RADs take the form of flat lamellae (Fig. 2) from which acicular aragonite needles grow. The pole figures presented in Fig. 6 show a strong clustering of aragonite [001] axes roughly parallel to the needle elongation axis, i.e. at right angle to the calcification interface. This feature is particularly striking at the tips of the lamella, in which [001] axes are systematically distributed normal to the calcification interface despite its abrupt change in orientation.

Discussion
Crystallographic data reveal a remarkably high degree of multi-scale self-ordering in the skeletal microstructure of cold-water coral L. pertusa. Evidence for this architectural organization includes the occurrence of ubiquitous polycyclic twinning between needle-forming crystallites, the axially-symmetric arrangement of needles into bundle-shaped sclerodermites, and the formation of crystallographically distinct transcurrent layers normal to the corallite wall. The occurrence of continuous transcurrent layers requires a non-random distribution of RADs within the microstructure. The alignment of RADs within planes with only minimal off-axis shift from the radial direction, together with their preferential positioning on the outer edges of opaque bands and/or yellow colour-coded transcurrent bands with respect to both radial and longitudinal growth directions, lead to a surprisingly regular tridimensional array in which order is only marginally disrupted. In this configuration, aragonite [001] axes are always kept within the reference plane of transcurrent layers, even for the sclerodermite needles intersecting the acquisition surface at high angles (Fig. 7). The highly-ordered nature of the microstructure is further illustrated within opaque and translucent bands, where crystallites display specific abundances of reciprocal crystallographic misorientations. The presence of sclerodermites in both opaque and translucent bands without discrimination or marked stops refutes the accepted idea that optical opacity of bands is caused by strongly disorganized crystallographic orientations.
Crystallographic data also shed new light on the growth mechanisms at play during the skeleton formation of L. pertusa. Although sclerodermites form individually from single source regions (i.e. RADs), simultaneous growth of adjacent bundles means that crystal growth competition occurs, as previously suggested by Barnes 27 . The degree of freedom available for needles to grow can be inferred from the distribution of aragonite [001] axes in transcurrent layers. A stronger clustering of axes in yellow colour-coded bands compared to the red colour-coded bands suggests that needles of the former, located on the front edge of mineralization on the longitudinal axis, had more freedom to grow than the latter, which had to compete for space with the needles of the previously-formed adjacent sclerodermites. This observation reinforces the idea of simultaneous radial and longitudinal growth of the corallite wall of L. pertusa, and strongly supports coral biomineralization models 9,28 and the interpretation of growth increments in the skeleton 14,16 .
Aragonite needles always grow with [001] axes normal to the calcification interface, would it be from equidimensional or lamellar RAD areas. Such a preferred orientation clearly suggests the prevalence of an interface-controlled growth process in L. pertusa. Extra-cellular mucus substances produced by cold-water corals, including L. pertusa, have previously been reported to serve several roles such as protection from attacks of endolithic and boring organisms, trapping of food particles and initiation of biomineralization processes 29-31 . Our findings confirm the possible role of extra-cellular mucus as a pre-requisite for the formation of RADs in initiating new aragonite nucleation centres in cold-water corals.
Progressing acidification of the contemporary oceans is predicted to result in shoaling of the aragonite saturation horizon and a reduction in calcification rates 32 . Ocean acidification will also cause a decrease in breaking  Fig. 2 and shown in Fig. 4a. Only misorientations <5° were considered in this map. A pixel is coloured in red when its average misorientation with respect to the 11 × 11 surrounding pixels is close to 5°. A pixel is coloured in white when its average misorientation with respect to the 11 × 11 surrounding pixels is close to 0°. The map shows a higher abundance of red pixels in opaque bands (O) and of white pixels in translucent bands (T).   . Three-dimensional model of the sclerodermite organization. Black dots and lines represent RADs and needles, respectively. On the right panel, the left part of the front surface corresponds to a yellow colour-coded transcurrent layer (as defined in Fig. 4) while the right part corresponds to a red colour-coded transcurrent layer. Because sclerodermites form three-dimensional bundles expanding towards all free directions, needles growing from RADs lying slightly off-axis with respect to the radial growth direction intersect the acquisition surface at high angles and form the multi-coloured patches visible in Fig. 2. They do not, however, interrupt the continuity of transcurrent layers. R and L as in Fig. 2. Artwork by A. Lethiers, used with permission. strength of L. pertusa 8 , which we suggest is most likely to propagate along the structurally weakest zones of the skeleton, namely the RAD, although the effects of this acidification on the degree of self-ordering is currently unconstrained. The highly-ordered structures described for L. pertusa likely contribute to the physical robustness of its skeleton, making it resistant to strong hydrodynamic currents. Several studies have reported preferred orientations of cold-water coral colonies, including that of L. pertusa 33 , with a correlation between colony orientation and local hydrodynamic and sediment supply characteristics. The capacity of L. pertusa to control its orientation for optimum growth is enhanced by its ability for remarkable multi-scale self-ordering of its skeletal structure.

Materials and Methods
Sample preparation. The specimen of L. pertusa studied here originated as a coral rubble grab sampled at a water depth of 340 m in the Porcupine Seabight, offshore SW Ireland (51°25′12″N, 11°30′13″W), during the RV Celtic Explorer CE-13001 cruise of January 2013. Standard sample preparation involved cleaning at room temperature in an aqueous solution of hydrogen peroxide (H 2 O 2 ) 5%, and several rinsing steps in an ultrasonic bath filled with deionised water. The specimen surface was carefully examined before analysis to avoid potential areas of bioerosion or other alteration features.
Electron backscatter diffraction and scanning electron microscopy imaging. Electron backscatter diffraction (EBSD) is an SEM-based technique which allows for an in-situ determination of the lattice orientation within crystalline solids. For each point of analysis, a diffraction pattern (also called EBSD or Kikuchi pattern) is generated through the interaction of the electron beam with the crystal lattice. The pattern is indexed and compared to a structure file containing the crystallographic parameters of the material, from which the orientation of the crystal lattice can be determined. Areas of interest can be scanned and displayed as colour-coded orientation maps, such as misorientation or inverse pole figure (IPF) maps. Points of analysis are characterized by three Euler angles (ϕ 1 , Φ, ϕ 2 ) which can be compiled into pole figures to display the orientation of crystal axes (e.g. aragonite [001]) with respect to a sample reference frame (e.g. longitudinal and radial growth directions in corals). A review of the basic principles of EBSD can be found in Prior et al. 34 .
The portion of the specimen to be investigated using EBSD was sawed longitudinally (Fig. 1), hot-pressed in conductive resin, and polished stepwise down using diamond paste (3 µm for three hours, then 1 µm for three hours) and colloidal silica suspension. No carbon coating was applied to ensure high-quality diffraction patterns. EBSD analyses were carried out at the University of Lausanne, Switzerland, using a Tescan Mira LMU FE-SEM operated at an acceleration voltage of 20 kV, a probe current of 2.5 nA, a working distance of 23 mm, and a sample tilt of 70°. The instrument was equipped with a Nordlys S detector and the AZtec 2. EBSD maps were collected using a point collection time of 0.15 s and a step size of 1.5 µm. Points of analysis with mean angular deviation (MAD) >1° were considered unreliable and discarded. Post-processing of the EBSD maps involved noise-reduction using the standard wildspike correction method of the AZtec software and a six-neighbour zero solution extrapolation.