Transformation mechanism of amorphous calcium carbonate into calcite in the sea urchin larval spicule

Abstract

Sea urchin larval spicules transform amorphous calcium carbonate (ACC) into calcite single crystals. The mechanism of transformation is enigmatic: the transforming spicule displays both amorphous and crystalline properties, with no defined crystallization front. Here, we use X-ray photoelectron emission spectromicroscopy with probing size of 40–200 nm. We resolve 3 distinct mineral phases: An initial short-lived, presumably hydrated ACC phase, followed by an intermediate transient form of ACC, and finally the biogenic crystalline calcite phase. The amorphous and crystalline phases are juxtaposed, often appearing in adjacent sites at a scale of tens of nanometers. We propose that the amorphous-crystal transformation propagates in a tortuous path through preexisting 40- to 100-nm amorphous units, via a secondary nucleation mechanism.

A widespread strategy in biomineralization is the initial formation of transient amorphous precursor phases that subsequently transform into one of the more stable crystalline phases (1). This process was first observed in the teeth of chitons where a disordered ferrihydrite precursor transforms into magnetite (2). It has also been observed in different invertebrate phyla (3–8). Amorphous calcium phosphate was recently identified in the newly deposited fin bones of zebrafish (9). The mechanistic details of these transformations are, however, still poorly understood. Here, we address this fundamental issue by studying the transformation of amorphous calcium carbonate (ACC) to crystalline calcite in the sea urchin larval spicule. Sea urchin larval spicules have long served as a model system for the study of CaCO₃ biomineralization processes, and the transient ACC precursor phase was first identified in this system (3). The mature larval spicule is composed of a single crystal of magnesium-bearing calcite (10, 11). Small amounts of organic macromolecules (0.1 wt%) are incorporated within the mineral and are known to play a role in the transient stabilization of the amorphous phase (12).

The spicules are formed inside a syncytium produced by specialized cells (13). The first deposit is a single rhombohedral-shaped calcite crystal. Further growth of the spicule radii follows crystallographic orientations dictated by the initial crystal (10, 14), even though the mineral deposited is mainly in the form of ACC. The rays elongate rapidly for ≈3 days, while the existing rays thicken. ACC is most probably introduced into the mineralization site by the cells in vesicles that fuse with the syncytial membrane (15). The spicule is tightly surrounded by this membrane, with no interstitial water solution detectable at any stage (16). In the polarized light microscope almost the entire spicule behaves as a homogeneously bright birefringent domain, despite being composed mainly of an amorphous phase. The exceptions are the growing tips of the spicule that show no birefringence, suggesting that they are completely amorphous (16). Partial demineralization (etching) of the spicule shows that the spicule is composed of densely packed mineral spherules 40–100 nm in diameter (3, 17). No crystallization front can be detected at the micrometer scale. Extended X-ray absorption fine structure (EXAFS) spectroscopy at the Ca K-edge showed that even at early stages, when the mineral is still predominantly amorphous, it already has a nascent short-range order around the calcium ions similar to that in calcite (18). In contrast to stable biogenic ACC, which contains 1 water molecule per CaCO₃, the amorphous phase in the spicules is mostly anhydrous when the spicules are extracted at an advanced developmental stage (12, 19). Macroscopically, therefore, the spicule displays both amorphous and crystalline qualities.

Results and Discussion

Unraveling the mechanistic complexities of the spatial and temporal interplay between the transforming amorphous and crystalline phases requires the use of high-resolution techniques. Here, we use X-ray photoelectron emission spectromicroscopy (X-PEEM) to study the transformation at high spatial resolution (20, 21). We analyze X-ray absorption near-edge structure (XANES) (22) spectra at the Ca L-edge along the length of spicules at two developmental stages. Ca spectra were acquired by recording 170 images, 0.1 eV apart, and arranging them in stacks in which the energy-dependent intensity of each pixel holds the full spectral information across the Ca L-edge. The pixel size depends on the magnification and is 40–200 nm in this study, while the probing depth is ≈3 nm at the Ca L-edge energy range (23). This technique offers the unique opportunity of characterizing the atomic order of the mineral phase (24) along a single larval spicule with sub-micrometer spatial resolution, providing time and space-resolved snapshots of the crystallization pathway through 2 distinct amorphous phases.

Fig. 1 shows spectra acquired from a 48-h embryo spicule with a pixel size of 200 nm. At this stage the spicule is at the triradiate stage of development and is composed of 70–90% ACC (18). The spectra were extracted from areas near the tip and along 1 of the spicule radii. The spectra near the tip are more heterogeneous than those from the rest of the spicule, revealing that newly-formed regions of the spicule are structurally diverse. Similar results on a different 48-h spicule are presented in supporting information (SI) Fig. S1. For comparison, spectra from synthetic calcite and synthetic ACC are also shown in Fig. 1 D and E, respectively. In calcite, the 2 main peaks (denoted 1 and 3, which are the Ca L₂ and L₃ peaks, respectively) are split, giving rise to 2 minor peaks [denoted as 2 and 4, the crystal field peaks (25)]. In synthetic ACC, peaks 2 and 4 are less intense and shifted closer to the main peaks, where they appear as shoulders.

Fig. 1.

Ca L-edge XANES spectra and an X-PEEM micrograph of a 48-h spicule. (A and B) Ca L-edge XANES spectra extracted from: near the tip (left yellow line in C) (A) and middle part of the spicule (right yellow line in C) (B). (C) X-PEEM micrograph of part of a fresh 48-h spicule. (D) Ca L-edge XANES spectra of synthetic calcite; the L₂ peak is split into peaks 1 and 2, and the L₃ is split into peaks 3 and 4. The main peaks are 1 and 3, and the crystal field peaks are 2 and 4. (E) Ca L-edge XANES spectra of synthetic ACC. These and all spectra hereafter were extracted from adjacent pixels along a line. The bold spectra at the top of A, B, D, and E are the averages of all spectra below. Blue in A and B highlights a spectrum similar to calcite. Green highlights a spectrum with intense peak 2 and small peak 4. Red, present in A but not in B, highlights a spectrum similar to synthetic hydrated ACC. Each spectrum in A, B, D, and E was extracted from a 200-nm pixel.

Analysis of numerous single pixel spectra revealed 3 independent calcium absorption line-shapes that correspond to 3 different mineral phases (Fig. 1). The first spectrum type (red in Fig. 1) is similar to synthetic ACC, where both peaks 2 and 4 are weak. This type 1 spectrum is found only near the tips of 48-h spicules. The second type (green in Fig. 1) has a pronounced peak 2 but a weak peak 4. The third type of spectrum (blue in Fig. 1) is similar to that of calcite, with both peaks 2 and 4 being intense. This third type of spectrum resembles calcite and is found at locations everywhere on the spicule surface, but with greater frequency and intensity in pixels at the center of the triradiate structure, where the initial rhombohedral calcite crystal was observed (10) (Fig. S1). All other spectra fall between these 3 types. The type 2 spectrum is distinct from those of synthetic ACC, and calcite spectra and cannot be the result of a linear combination of types 1 and 3, because any linear combination of these leads to a parallel change in both peaks 2 and 4 (Fig. S2). The suppression of the L₃ crystal field peak 4 in type 2 indicates that this spectrum represents a disordered form of calcium carbonate, although no similar spectrum has been recorded from standard materials. In both 48- and 72-h spicules, the most abundant phase is type 2.

Over time, the amorphous material present in fresh spicules is known to crystallize (3), and we sought to determine whether the ACC phases found here in fresh spicules, 48 h after fertilization, are transient phases. We repeated the measurements on fresh spicules grown for a longer period (72 h after fertilization) and then at precisely the same location on the same spicules after 10 months in storage (Fig. 2 and Fig. S3). The fresh 72-h spicules consist only of type 2 ACC, crystalline biogenic calcite, or combinations thereof, with no type 1 ACC phase (Fig. 2E), whereas the spectra obtained from the 10-month-old spicule are uniformly similar to crystalline biogenic calcite (Fig. 2F). This result implies that the type 2 phase is less stable than calcite. We infer that the crystalline phase grew at the expense of amorphous domains and/or the amorphous phase, having higher solubility, was preferentially removed. The 10-month-old spicule has an etched appearance when imaged in the scanning electron microscope (SEM) (Fig. 2D), resembling an aggregate of spheres of 40–100 nm in diameter. The etched spicule exhibits topographic features that are coarser than nonetched spicules. The homogeneity of the spectra extracted from this sample thus demonstrates that specimen topography cannot produce the spectral diversity shown in Fig. 2E.

Fig. 2.

X-PEEM micrographs and Ca L-edge XANES spectra of a fresh 72-h spicule, and the same spicule after 10 months. (A–C) X-PEEM micrographs of the 72-h spicule. (A) Fresh spicule: the dark region immediately below the spicule is its shadow. (B) The same spicule measured 10 months later. The yellow lines indicate the pixels from which the spectra in E and F were extracted. (C) High magnification of the area in A: colored pixels are those from which the corresponding colored spectra in E and F were extracted. (D) SEM micrograph of the same spicule taken after 10 months. (E and F) Ca L-edge spectra extracted from the lines in A and B, respectively. The blue curve in E corresponds to the type 2 ACC phase, which clearly became calcitic in F. Scale bar in A also applies to B, and scale bar in C also applies to D. Pixel size is 100 × 100 nm². Highlighted in black is the average spectrum.

Selected spectra from the 48- and 72-h spicules in Figs. 1 and 2 and from the reference standards were peak-fitted. The peak-fitting results, presented in Fig. S4, highlight the spectroscopic differences among type 1, 2, and 3 mineral phases, and the similarity of types 1 and 3 with synthetic ACC and calcite.

We further characterized the occurrence of these distinct carbonate phases in 72-h spicules by repeating the measurement at higher magnification (40-nm pixels vs. 200- and 100-nm pixels in Figs. 1 and 2) to gain more insight into the dimensions of the individual domains. Indeed, isolated spectra of amorphous domains can be detected in the middle of crystalline domains. We observed abrupt transitions between calcite and ACC in immediately adjacent pixels, as well as more gradual transitions (Fig. 3). The smallest domains observed are 1 pixel wide, and others are 3 pixels wide (40–120 nm).

Fig. 3.

Ca L-edge XANES spectra from a fresh 72-h spicule. The spectra are extracted from individual pixels along a straight line. Each pixel represents 40 × 40 nm². The 2 series of spectra show different patterns of mineral phase distribution in adjacent pixels. (A) We observe large blocks of 5–10 adjacent spectra of type 3 calcite interspersed with smaller series of spectra of type 2 ACC. (B) The transitions between type 2 and 3 spectra are abrupt between the bottom 2 spectra and gradual for the others. Highlighted in black are the average spectra. Blue highlights 1 of the type 3 calcite spectra, green highlights type 2 ACC.

Because the phases may be distinguished by the crystal field splitting at the L₂ and L₃ edges, we define an empirical splitting ratio (SR) to be the ratio of the maximum intensity of the crystal field peak (2 or 4) and the intensity of the minimum separating this peak from the corresponding main peak (1 or 3) (Fig. 4 and Fig. S5). For synthetic ACC, both peaks are poorly split and thus both SR values are less than unity. For biogenic and synthetic calcite, both peaks are well resolved, and the SRs are larger than 2 and 1.3, for L₂ and L₃, respectively. The single-pixel SR values for spicules are varied, and structural trends in the data from the biogenic samples are illustrated by plotting L₃ SR vs. L₂ SR (Fig. 4B). Each of the 3 phase types identified above falls in a different quadrant of this plot (Fig. 4C): ACC type 1 with both SRs < 1; ACC type 2, with L₂ SR > 1 and L₃ SR < 1; and calcitic with both SRs > 1.

Fig. 4.

Peak splitting anaylsis of Ca L-edge XANES spectra from 48-h and 72-h spicules, as well as those from ACC and calcite. (A) Ca L-edge XANES spectra extracted from single pixels of synthetic ACC (bottom red curve) and calcite (top blue curve) and 3 spectra from a 48-h spicule. The 3 spicule spectra are representative of the 3 mineral phases identified in Fig. 1: red is type 1, green is type 2, and blue is type 3. (B) A plot of SR(L₃) vs. SR(L₂) (see Methods and Fig. S5). The spicule samples indicated by triangles (48-h spicule tip, light blue; middle, purple; 72-h spicule fresh, green; 10 month olds, blue), and the adult sea urchin spine (squares, brown), span 3 quadrants, and synthetic ACC (red circles) is located in the bottom left quadrant, where L₂, L₃ SR < 1. Calcite (blue diamonds) is located in the top right quadrant where L₂, L₃ SR > 1. (C) Spicule ratios are shown separately with the relevant quadrants shaded in gray. Color code is as in B.

The spectra are often a mixture of phases. This mixture occurs when a pixel includes, for instance, part of a type 2 and part of a type 3 particle. Thus the corresponding spectrum and SRs are intermediate between types 2 and 3. The spicule SRs tend to become more calcite-like with increasing distance from spicule tip to the middle, and with growth time from 48 to 72 h after fertilization (Fig. 4). All SR values obtained from the 72-h spicule after 10 months fall in the calcite-like top-right quadrant, as expected. However, the values are much smaller than those of synthetic calcite, indicating greater structural disorder. This result might be attributed to the presence in the spicules of ≈5 mole% magnesium and the occluded matrix proteins. To support this hypothesis, we measured the SRs of adult sea urchin spines, considered to be composed of crystalline calcite exclusively. The spine's SRs (Fig. 4B), with Mg concentration similar to the spicules, also do not reach the values of synthetic calcite, although they are shifted toward calcite relative to the spicule. Consistently, spectra of biogenic minerals containing higher amounts of Mg have even lower SRs, but with both SRs > 1 (Yurong Ma, personal communication).

The present data thus show that spicule development involves 2 amorphous precursor phases. The freshly deposited mineral is similar to hydrated synthetic ACC. This phase is short-lived and can be detected only in areas of fast growth (the tip), rather than where slower radial thickening occurs (17). This type 1 ACC rapidly transforms into a second phase that appears amorphous from the XANES data. This type 2 ACC transforms more slowly into biogenic calcite.

Prior studies on spicules extracted at an advanced developmental stage have shown that the amorphous phase in this system is mostly anhydrous (19). However, it has been suggested that water molecules may be present as part of an initial hydrated ACC phase that subsequently transforms into the anhydrous phase (7, 18). Our data are in good agreement with this mechanism, and we suggest that they represent direct observation of the dehydration step for freshly deposited ACC. Because this type 2 phase is the most abundant in fresh spicules it is likely to be the same anhydrous ACC phase observed with bulk methods (18, 19) and now confirmed to be formed from an earlier transient phase. We note that the probing depth in these experiments is only 3 nm. Thus only the fresh material that is deposited on the spicule surface upon thickening is sampled.

A recent EXAFS study showed that the transient ACC phase at this stage has a short-range order that resembles the mature crystalline phase (18), which might be the origin of the prominent peak 2. The type 2 mineral is therefore intermediate between fully disordered, probably hydrated ACC and crystalline calcite with respect to both spicule development stage and crystallinity.

The present data show the presence of juxtaposed crystalline and amorphous phases in the surface layer of growing spicules, raising the fundamental question of how these heterogeneous mineral domains transform into the single crystal found in the mature spicule. Insight is obtained from the higher-resolution analysis of the size of single domains. In all specimens, we observe discontinuity in mineral phases in immediately adjacent pixels, suggesting that the precursor mineral phase is present in small units. Analysis of 72-h specimens showed that homogeneous domains of type 3 calcite as large as 1 μm are present, interspersed with smaller domains of type 2 ACC. The smallest ACC domains (40–120 nm) observed here with X-PEEM, are consistent in size with the previous observation of 50–100 nm spherules (Fig. 2) (3, 17) and the coherence length of the mature spicule observed in X-ray diffraction (26).

From these observations, we propose that a transformation from type 2 amorphous to type 3 crystalline phase propagates through the spicule via secondary nucleation, in which the crystallization of 1 amorphous unit stimulates the transformation of the domains in contact with it (27, 28). The overall crystallographic orientation is determined by the initial central crystal. Because type 2 ACC is anhydrous, no volume change occurs during the transformation to type 3 calcite, and spicule morphology is unaffected.

The propagation pathway through the spicule is inferred to be complex and tortuous, implying that the rate of transformation depends on the size and interface structure of the amorphous domains. These are probably determined by the presence, location, and concentration of the organic additives. Within larger calcitic regions of 72-h spicules, individual amorphous domains of 40 × 40 nm² were occasionally identified, indicating that the propagation pathway may leave small domains untransformed. Mapping the distribution of single-phase domains, as demonstrated here, will enable the testing of various hypotheses that may account for the patterns of crystalline-phase propagation that are akin to fractal network percolation (29, 30).

The transformation mechanism presented here may well represent a common strategy in biomineralization, bearing in mind the widespread use of precursor ACC phases in biology and the many cases in which 30- to 50-nm spherulitic structures have been observed in biogenic calcium carbonate minerals from diverse phyla (31–35).

Methods

More detailed descriptions of the experimental procedures are provided in SI Text.

Sea Urchin Larval Culture.

Strongylocentrotus purpuratus embryos were grown in artificial sea water containing Gentamycin (20 mg·L⁻¹) at 15 °C, following established methods (36, 37).

Extraction of S. purpuratus Spicules.

Embryos were disrupted in a Polytron homogenizer. The spicules were collected by centrifugation and extracted with SDS and 3.5% NaOCl. The spicules were washed with CaCO₃-saturated solution and rinsed with ethanol and acetone, frozen in liquid nitrogen and kept at −80 °C for up to 2 days until the measurements.

Synthetic calcite crystals were grown in Nunc multiwell dishes by diffusion of ammonium carbonate vapor into 10 mM calcium chloride (Merck; A grade) solutions (38). Synthetic ACC was synthesized following Koga et al. (39) by mixing solutions of calcium chloride (0.1 M) with sodium carbonate (0.1 M) and sodium hydroxide (1 M).

X-PEEM Sample Preparation.

Forty-eight- or 72-h spicule samples were resuspended in ethanol. A drop of the suspension was deposited on a 10 mm × 10 mm silicon chip and air-dried. Synthetic ACC and calcite powders were pressed into indium foil. All samples were sputter-coated with 1 nm Pt (40).

X-PEEM Experiments.

We used the spectromicroscope for photoelectron imaging of nanostructures with X-rays (SPHINX), which is an X-PEEM (Elmitec), installed on the VLS-PGM beamline at the Synchrotron Radiation Center. The instrument and its performances are described in detail in ref. 20. Briefly, the sample was mounted vertically and illuminated from a side (16° grazing incidence angle) with monochromatic soft X-rays. The photoelectrons emitted by the sample were accelerated toward an electron optics column and onto a phosphor screen. The chamber was held at ultrahigh vacuum (10⁻¹⁰ Torr). The real-time, sample surface image was acquired by a slow-scanning cooled CCD camera. Movie stacks of 170 images were acquired while scanning the photon energy across the Ca L-absorption edge, so that each pixel in the movie contained a complete XANES spectrum. Images were acquired with fields of view ranging from 20 to 100 μm and corresponding pixel sizes of 40–200 nm. Because the samples are not flat, the spatial resolution may be lower than the pixel size.

Data Analysis.

Data processing was done by using routines we designed in National Institutes of Health Image 1.62 and Igor Pro 6.0.3 (WaveMetrics) for Macintosh. From each stack of 170 images Ca L-edge spectra were extracted from single pixels along a straight line on the spicule or standards and divided by a pre-edge linear background. All spectra presented here were normalized to a linear pre-edge fit. The position of peak 1 was set to be 352.6 eV for all spectra, following Benzerara et al. (41). The intensity of the pre-edge was then set to 0 and that of peak 1 to 1. All spectra were smoothed over 5 points and displaced vertically in all figures.

Peak Fitting.

Selected spectra were peak-fitted by using routines we designed in Igor Pro 6.0.3 (WaveMetrics). The most representative peak-fitted spectra are presented in Fig. S4.

Calculations of SRs.

For each peak we divided the intensity value, after approximated-baseline subtraction, of the split peak (spL₂, spL₃) by the intensity value of the minimum between these peaks and the main peak (dipL₂ and dipL₃, respectively). Such that:

See Fig. S5.

Scanning Electron Microscopy.

Samples were coated with 6 nm Cr and viewed with a SEM (Philips; XL30 FESEM FEG), operated at 10 kV.