pmc.ncbi.nlm.nih.gov

Amorphous calcium carbonate particles form coral skeletons

Significance

Whether coral skeleton crystals grow by attachment of ions from solution or particles from tissue determines (i) corals’ growth rate, (ii) how they survive acidifying oceans, and (iii) the isotopes in the crystals used for reconstructing ancient temperatures. Our data show that two amorphous precursors exist, one hydrated and one dehydrated amorphous calcium carbonate; that these are formed in the tissue as ∼400-nm particles; and that they attach to the surface of coral skeletons, remain amorphous for hours, and finally crystallize into aragonite. Since these particles are formed inside tissue, coral skeleton growth may be less susceptible to ocean acidification than previously assumed. Coral bleaching and postmortem dissolution of the skeleton will occur, but a calcification crisis may not.

Keywords: mesocrystal, PEEM, calcification crisis, vital effects, ocean acidification

Abstract

Do corals form their skeletons by precipitation from solution or by attachment of amorphous precursor particles as observed in other minerals and biominerals? The classical model assumes precipitation in contrast with observed “vital effects,” that is, deviations from elemental and isotopic compositions at thermodynamic equilibrium. Here, we show direct spectromicroscopy evidence in Stylophora pistillata corals that two amorphous precursors exist, one hydrated and one anhydrous amorphous calcium carbonate (ACC); that these are formed in the tissue as 400-nm particles; and that they attach to the surface of coral skeletons, remain amorphous for hours, and finally, crystallize into aragonite (CaCO₃). We show in both coral and synthetic aragonite spherulites that crystal growth by attachment of ACC particles is more than 100 times faster than ion-by-ion growth from solution. Fast growth provides a distinct physiological advantage to corals in the rigors of the reef, a crowded and fiercely competitive ecosystem. Corals are affected by warming-induced bleaching and postmortem dissolution, but the finding here that ACC particles are formed inside tissue may make coral skeleton formation less susceptible to ocean acidification than previously assumed. If this is how other corals form their skeletons, perhaps this is how a few corals survived past CO₂ increases, such as the Paleocene–Eocene Thermal Maximum that occurred 56 Mya.

The classical model for coral skeleton formation assumes that ions from seawater are either actively or passively (1) transported through the living tissue, they are concentrated (2), and delivered to the growing skeleton surface (3), where they are deposited ion by ion; thus, the coral skeleton should be in equilibrium with bulk seawater (4). Elemental Sr/Ca and Mg/Ca ratios and isotopic δ¹¹B, δ¹³C, and δ¹⁸O analyses of corals, however, reveal significant deviations from equilibrium, termed “vital effects” (5–8), which cannot be reconciled with the classical model. The alternative is coral formation from amorphous calcium carbonate (ACC) precursor particles formed in the animal tissue, as hypothesized for sea urchins (9), and thus, away from bulk solution and from equilibrium thermodynamics. Why is it important if the coral aragonite crystals form ion by ion from solution or by attachment of amorphous particles?

Coral is one of many marine organisms that biomineralize, that is, form hard crystalline shell or skeletal structures (10). Calcium carbonate (CaCO₃) biomineralization is an intensely studied, widespread natural phenomenon, with major implications in paleoclimate reconstructions (11, 12). Coral biomineralization is important to understand how corals responded to past environments (11, 13), thus clarifying their use as paleoenvironmental proxies, their vital effects, and their fate in future changing environments (14). Coral morphology and biomineralization have been studied for decades (15), including the animal tissue and the skeleton that it deposits (16). However, many of the basic mechanisms underlying the aragonite mineral formation remain unknown (1, 17).

ACC in coral was long suspected but never shown (17). Previous experiments aimed at revealing its presence in forming corals did not show evidence of ACC but could not discount it either (18). Three previous lines of evidence suggested that coral skeletons form via amorphous precursors. First, high-resolution scanning electron (19) and atomic force (20) microscopies (SEM and AFM, respectively) revealed that coral skeletons consist of nanoparticles on the order of 100 nm in size. Nanoparticulate texture persisting in mature mineralized tissues has been associated with crystallization by attachment of amorphous particles in vitro (21), in sea urchin spicules (22), and in nacre (23), in contrast with vaterite tunicate spicules, which are not nanoparticulate and likely form ion by ion from solution (24). Second, the spatial distributions of Sr and Mg in coral skeletons are very different from one another, and Mg has been suggested to stabilize ACC (25). Third, crystal growth in vitro in the presence of coral organic matrix showed formation and stabilization of ACC (26). Interestingly, two forms of ACC were observed, with different amounts of water and thermal stability (26, 27). ACC was first identified as a precursor phase in sea urchin spicules (28) and spines (29), larval mollusk shells (30), and a variety of other biominerals. Two different ACC precursor phases were detected on the surface of sea urchin spicules by Politi et al. (31), in their cross-sections by Gong et al. (32), and later, in nacre by DeVol et al. (23).

Can amorphous precursors be observed during coral formation? We addressed this question using photoemission electron spectromicroscopy (PEEM) and X-ray absorption near-edge structure spectroscopy with 20-nm resolution (23) to directly observe all mineral phases present in adult coral and newly settled spat skeletons, all freshly killed. We found clear spectroscopic evidence of two amorphous precursor phases, hydrated amorphous calcium carbonate (ACC-H₂O) and anhydrous ACC, present only at the newly deposited surface of growing coral skeletons, at the centers of calcification (CoCs), and in particles that appear to be within the tissue near the growing surface.

Results and Discussion

In Figs. 1–4, we present direct evidence of amorphous precursor phases in fresh, forming coral skeletons of Stylophora pistillata in various developmental stages and coral locations. Figs. 1–4 show “component maps” obtained from PEEM data by fitting each single-pixel spectrum with a linear combination of spectral “components,” displayed in Fig. 1A, and assigning to that pixel a color, which displays quantitatively the proportion of each component. The same region, at the interface of tissue and coral skeleton, is also shown in visible light microscopy in Figs. S1 and S2.

Fig. 1. — Component spectra and component maps of tissue and fresh, forming coral skeleton in a spine that appears separate from the rest of the skeleton, as it is curved and incompletely exposed in this polished surface. These data were acquired 25 h postmortem. (A) The reference Ca spectra used as components in all component maps in this work. (B) Component map of forming coral at the tissue–skeleton interface in *S. pistillata*. In a component map, colors indicate mineral phases, as distinguished by their spectra in A, and are correspondingly colored. In A–E, blue pixels are mostly or completely aragonite; ACC-H₂O, ACC, and pAra are displayed in red, green, and yellow, respectively. (C–E) Magnified component maps from the correspondingly labeled boxes in B. C and D show three particles (arrows) surrounded by tissue, which clearly contain more ACC-H₂O, ACC, and pAra pixels than any other region in this work. In the forming coral skeleton, instead, there are only ACC-H₂O and aragonite, very little ACC, and no pAra at all, as shown in E. The spectra in A are normalized to have the same area under the curve between 345 and 355 eV. They are vertically displaced here for clarity, not during component mapping. The maps in B–E were obtained from a stack of PEEM images, fitting every pixel with a linear combination of the four component spectra in A and displaying in each pixel the proportion of each spectrum that best fits the spectrum of that pixel in red, green, blue, and yellow colors. Since yellow never occurs as a mixture of red and green, displaying pAra as yellow is unambiguous (Figs. S8 and S9). Important additional data for this region are in Figs. S1–S3, S8, and S9, and repeat component maps are in Fig. S3. Quantitative analysis for the abundance of each mineral phase is presented in Table 1.

Fig. 4. — Amorphous precursors in a growing coral septum. These data were acquired 31 h postmortem. (A and B) Cross-polarizer micrographs, with magenta arrows and boxes showing the region magnified in C and D. (A) The whole coral branch, termed nubbin. (B) A flower-shaped holey structure termed corallite, with four fully formed and two forming septa. (C) Component map of a forming septum, showing that, in this adult coral, the forming skeleton has a lot of ACC-H₂O, very little ACC, and almost no pAra. (D) PIC map showing aragonite crystals in the septum, distinguished by different colors, which quantitatively identify their crystal orientations. By the time that the PIC map was acquired (32 h postmortem), all of the amorphous precursor phases had already crystallized to aragonite because of time delay and radiation damage. The repeat component map for this region, acquired immediately after the first and before PIC mapping, is in Fig. S3, and quantitative phase analysis is in Table 1.

Fig. S1. — Fresh, forming coral at the tissue–skeleton interface. (A) Cross-polarizer micrograph showing mature coral, corallite holes, and the growing coral front. (B) Zoomed in micrograph where the coral skeleton and the calicoblastic tissue (green arrows) depositing the mineral are visible. Magenta arrows and boxes in A and B indicate the area magnified in C–E. (C) Average of 121 PEEM images acquired across the Ca L₂₃ edge. The boxes indicate the areas magnified in Fig. 1 C–E. (D) Component map identical to the one presented in Fig. 1B using the component spectra in Fig. 1A. The boxes indicate the regions magnified in Fig. 1 C–E. (E) PIC map of the same region in D, indicating different crystal orientations with different colors. Noncrystalline pixels are black. Additional images for this region are in Fig. S2, and the repeat component map is in Fig. S3.

Fig. S2. — The coral tissue–skeleton interface is analyzed in detail in Fig. 1 and Fig. S1. (A) Average of all PEEM images acquired across the Ca L₂₃ edge. (B) Cross-polarizer micrograph showing the spine, the tissue immediately adjacent to it, ectoderm and endoderm, and the symbiont zooxanthellae.

The forming coral skeleton in Fig. 1 is as fresh as possible (25 h postmortem). It contains a remarkable amount of ACC-H₂O: more than 20% of the pixels contain at least some ACC-H₂O (as shown in Table 1) and are, therefore, red or magenta pixels interspersed with aragonite. The observation of red or magenta pixels is consistent across various skeletal locations: a growing spine in adult coral (Fig. 1 B and E), one in a settled and calcifying spat (Fig. 3), and a growing septum in a corallite of an adult coral (Fig. 4). Very little ACC is observed (∼2% green pixels in adult and ∼10% in the spat skeletons). This differs from what was observed in sea urchin spicules (32) and nacre (23), where the most abundant amorphous phase observed was anhydrous ACC, not ACC-H₂O as observed here.

Table 1.

Occurrence of red (R), green (G), blue (B), and yellow (Y) pixels in each component map

Figure	R pixels (%)	G pixels (%)	B pixels (%)	Y pixels (%)
Fig. 1, coral skeleton only	21	2	100	0
Fig. 1, particles appearing in coral tissue only	63	20	98	12
Fig. 1	22	2	100	0
Fig. 2	15	6	100	1
Fig. 3	25	10	100	2
Fig. 4	16	1	98	1

Fig. 3. — Amorphous precursor phases in a forming spine in a 2-wk-old spat (that is, a coral larva right after it settled and started forming a skeleton). These data were acquired 28 h postmortem. ACC-H₂O is abundant, ACC is sparse, and pAra is absent from all skeleton regions (spat here, adult in Figs. 1 and 4), except for the CoCs (Fig. 2). (A and B) Cross-polarizer micrographs of a spat spine, with magenta arrows and boxes indicating the area magnified in C and D. (C) Component map obtained with the same four reference spectra as in all other data in this work, although here, we omitted the pAra label to stress that we found no pAra. C, *Inset* shows a zoomed in map for the white box region. (D) Average of 121 stacked PEEM images (termed movie), from which the component map in C was obtained. The repeat component map for this region is in Fig. S3, and quantitative phase analysis is in Table 1.

Most interestingly, since in Fig. 1B, both the coral tissue and the growing coral skeleton are imaged, we captured three mineral particles surrounded by tissue (Figs. S1 and S2) with spectra that show mostly ACC-H₂O (63%) but also, ACC (20%) and poorly crystalline aragonite (pAra; 12%) (Table 1). These three particles, magnified in Fig. 1 C and D, are surprisingly large [0.4 ± 0.2 µm in size (median ± SD)] and could be in the skeleton in structures above or below the image plane. Based on the fact that they contain so much more ACC-H₂O, ACC, and pAra than is ever found in any skeletal elements (Table 1), however, we deduce that they must be in tissue. Synthetic ACC-H₂O crystallizes readily, especially in contact with water. Thus, the observation of long-lived ACC-H₂O implies inhibition of the transition from ACC-H₂O to ACC in the tissue and in the growing coral skeleton. Because repeat acquisitions show some of the ACC-H₂O transformed into ACC as a consequence of radiation damage (Fig. S3), we presume that this thermodynamically downhill transition (33) is accelerated by radiation exposure but would happen spontaneously in the undisturbed coral. In the particles within the tissue regions, we observe greater concentrations of ACC than in the forming coral skeleton, suggesting that this phase is stabilized in the tissue but not in the skeleton, where it is extremely short-lived. Greater stabilization of ACC in tissue is consistent with preventing crystallization more in the tissue (that is, at the wrong time and place) than in the skeleton, which makes sense.

Fig. S3. — Component maps from repeat movies for the areas in Figs. 1–4. A, C, E, and G are first movies; B, D, F, and H are second movies. *Insets* show small portions of the same maps at higher magnification, such that individual pixels are visible. Comparing A with B, some ACC-H₂O transformed into ACC, and some transformed into aragonite. The same occurs when comparing E with F or comparing G with H. No appreciable change is observed between C and D, which was acquired in mature coral skeleton.

Can dissolution of ACC-H₂O and reprecipitation as aragonite at the biomineral formation site occur, instead of solid-state transformation from ACC to crystal? Based on the paucity of ACC observed in all coral skeletons [<10% of pixels have any ACC in the coral skeleton (Table 1), and even in those pixels, ACC is never pure but 80% ACC at most], dissolution of ACC-H₂O or direct precipitation from solution is possible. However, this cannot be the only phenomenon occurring, as that would imply complete absence of ACC, which was never observed in any areas of forming coral skeletons (Table 1). The existence of some ACC leads to the conclusion that only or mostly solid-state transformations occur from ACC-H₂O to ACC to aragonite.

A pAra phase was previously observed in coral. This spectrum is similar to aragonite around 348 eV (peak 4 in Fig. S4) and to ACC around 351.5 eV (peak 2 in Fig. S4); hence, it was previously assigned to a protoaragonite or poorly crystalline aragonite mineral (23). With the normalization used in Fig. S4, it is clear that pAra has a much more intense peak 1, and thus, it is a crystalline phase, far more ordered than ACC or ACC-H₂O but less ordered than aragonite. It is certainly not a mixture of ACC and aragonite (23); thus, it can only be a poorly crystalline aragonite phase. The disorder may be caused by, for example, incorporation of organic molecules in the crystal (34). The pAra mineral was previously observed in the CoCs of a 10-y-old Madracis coral (23). Here, we confirmed this observation in S. pistillata, finding abundant pAra localized in the CoCs, as shown in Fig. 2C, and not anywhere else in the coral skeletons (Figs. 3 and 4). Interestingly, the pAra phase was also observed in the tissue particles in Fig. 1. Thus, the pAra phase, stabilized in the living tissue, is delivered and remains stabilized in the skeleton but only in the CoCs, where it persists for years. This is consistent with observations of CoC-specific organic molecules by Cuif et al. (35) and Mass et al. (20) and with the observation of Mg in CoCs (25, 36). Since the pAra phase in CoCs was never observed in either Stylophora or Madracis (23) to crystallize to aragonite, we conclude that it cannot be an “immature aragonite” or “ACC” phase as proposed by Von Euw et al. (36), nor one of the “polyamorphous” precursor phases described by Cartwright et al. (37), or the “protoaragonite” described by Farhadi‐Khouzani et al. (38). The pAra phase in CoCs is stable or stabilized poorly crystalline aragonite.

Fig. 2. — CoCs in a more mature part of the coral skeleton. (A) Visible light microscopy image obtained with crossed polarizers. In A and B, the magenta boxes indicate the region magnified in C and D. (B) Zoomed in cross-polarizer micrograph showing crystals that radiate out of the CoCs. (C) Zoomed in component map showing abundant ACC-H₂O, ACC, and pAra in the CoCs, indicated by the white arrows. The repeat component map for this region is in Fig. S3. (D) PIC map of the same CoCs shown in B, with white arrows in precisely the same positions. Notice that, where the amorphous precursors are localized in C, this PIC map shows no crystallinity; that is, no polarization dependence, displayed as black pixels.

Our data show that the pAra particles found in the CoCs were already pAra in the tissue (Fig. 1); thus, there is no evidence of nucleation events in the CoCs, as was proposed by Von Euw et al. (36).

The other ACC-H₂O and ACC phases do not exhibit any of the spectral features characteristic of aragonite (peaks 2 and 4 in Fig. S4); thus, they cannot be considered protoaragonite phases (37, 38). They are spectroscopically similar to those detected in forming calcite biominerals (31, 32, 39); thus, they could be termed “protocalcite” phases.

Additional evidence that the precursor phases ACC-H₂O and ACC observed in forming coral or the CoCs are indeed amorphous is provided by polarization-dependent imaging contrast (PIC) mapping, which displays in different colors the orientation of crystalline c axes in carbonate crystals (40, 41). The PIC map in Fig. 2D, for example, shows that, where pixels are red or green in the component map (Fig. 2C), they are black in the PIC map (Fig. 2D), confirming lack of crystallinity. White arrows in Fig. 2 C and D point precisely at the same CoC locations in the two maps, and these appear rich in amorphous phases in the component map and mostly black in the PIC map. Black in a PIC map may also indicate orientation: aragonite crystals or nanocrystals with their c axes pointing into the beam appear black (e.g., near the scale bar in Fig. 2D). The colocalization of amorphous spectra and black pixels in the PIC map (Fig. 2 C and D, respectively) make the interpretation of black in the PIC map unambiguous. Amorphous regions in Figs. 3 and 4 had already crystallized by the time that PIC maps were acquired.

Can the large amorphous particles identified in fixed and embedded tissue regions in Fig. 1 be detected in living corals? Confocal microscopy data in Fig. 5 show that this is indeed possible. The blue calcein-stained particles, outlined in Fig. 5D, vary in size between 0.4 and 9.4 µm, with 400 ± 700 nm being both the median and the mode ± SD, in the forming mineral (FM) and in the living tissue (LT). The LT has some particles greater than 1 µm, as previously observed (42) and as seen in Fig. 1C. The 400-nm most frequent particle size in LT is in agreement with that observed in Fig. 1 and in some FM regions in Figs. 2–4; therefore, it must be correct. It also coincides with the size of type I vesicles Clode and Marshall observed to be the most abundant in calcifying cells near the FM after cryofracturing them (43), but it is considerably larger than the 100 nm expected from SEM observations of coral skeletons (19–21, 44).

Fig. 5. — (A) Confocal micrograph of a living spat. Calcein fluorescence is displayed in blue; native green fluorescent protein (GFP) and chlorophyll are in green and red, respectively. (B) Map of the same spat, where regions of FM and LT are filled with blue and green, respectively. Notice that they are interspersed, with LT in FM and FM in LT. (C) Differential interference contrast micrograph acquired with crossed polarizers, showing a similar region of the same spat, where a basal plate (BP), a growing septum (S), and a layer of stinging cells (SC) are better recognizable, and therefore, they could be placed in B. (D) Outlines of all 2,726 blue-stained particles analyzed, which vary in size between 0.4 and 9.4 µm.

In Fig. 6, we show particles identified by their crystal orientations in PIC maps of both a natural coral skeleton and a synthetic aragonite spherulite grown in vitro by particle attachment. Fig. S5A shows 400-nm particles, and Fig. S5B shows that crystal orientation domains have edges with diameter of curvature ∼400 nm, consistent with attachment of particles of this size. In Figs. 3 and 4, amorphous domains are also on the order of 400 nm in the forming skeleton, again confirming that such large particles originating in LT are not unreasonable.

Fig. S5. — PIC maps of the same regions as in Fig. 6. White squares numbered 1–8 indicate similarly sized particles in all panels. Each particle is identified by a square (above) and a number (below) and is precisely between the square and the number, except for particle 7 in D, which is below the square and left of the number 7. The white squares are 400 nm in A and B and 100 nm in C and D. (Zoom in on the online figure if the particles or the squares are not visible to you.) (A) Mature coral skeleton region at the CoCs. Notice the eight labeled and many other 400-nm particles. (B) The same region of coral at lower magnification. Away from the CoCs, domains of identical crystal orientation (color) are larger, but their edges are jagged at the 400-nm scale, as shown by the eight labeled and many other 400-nm edge wiggles. This is consistent with attachment of 400-nm particles. (C and D) The smallest particles identifiable in coral (C) and synthetic aragonite spherulite (D) are labeled 1–8, with 100-nm squares. Notice that this is not the particle size, because PIC mapping is surface-sensitive [∼5 nm at the oxygen K edge where the PIC maps are acquired (96, 97)], and both coral skeleton and synthetic spherulites were polished surfaces; thus, these may have been larger particles with their center above or below the imaged plane. The precise sizes of the particles in C and D were analyzed by measuring the FWHM across the eight labeled particles. These were 143 ± 52-nm particles in coral skeleton and 72 ± 26 nm in synthetic spherulite (mean ± SD). These are in agreement with previous observations of ∼100-nm particles in coral skeletons (20, 44) but not with those observed in LT or with the most frequent 400-nm particles observed in these same PIC maps; thus, we deduce that these particles appear smaller, because their centers were away from the polished and imaged plane.

Monochromatic microdiffraction of two S. pistillata mature skeletons finds coherence lengths in different locations of 250 nm in the CoCs and 300 nm in the large crystalline domain areas. These are the minima observed; the maxima, averages of all data, and their errors are presented in SI Materials and Methods. These coherence length minima are similar to the smallest particle size observed in PIC maps and confocal microscopy in those same regions. Where a larger coherence length is observed in the large domain areas, uniformly colored in PIC maps, the particles need not necessarily have been larger; it is possible that multiple 400-nm particles became perfectly cooriented as they crystallized in those locations. In SEM images of coral skeletons, we see particle sizes varying between 100 nm and more than 1 micron (Fig. S6 E–H), consistent with this interpretation.

Fig. S6. — SEM images of cryofractured synthetic aragonite, biogenic aragonite from coral, and geologic aragonite. (A–D) Micrographs at increasing magnifications show a synthetic spherulite grown by particle attachment using the ammonium carbonate diffusion method and an Mg:Ca ratio of 5:1. Particles ∼100 nm in diameter are visible along with larger structures. (E–H) Micrographs of cryofractured fibers in an *S. pistillata* coral skeleton. Various regions imaged at high magnification exhibit nanoparticles at the tips of fibers (E–G), at fractured fiber cross-sections (E and G), and at the sides of the fibers (F–H). Particle with ∼100-nm size are visible, but so are other structures at the ∼400-nm scale (E) and even greater than a micron (F); thus, as expected, SEM alone cannot provide evidence of particle sizes. *Insets* in E–H show the same regions at lower magnification. (I–L) Micrographs of cryofractured geologic aragonite at increasing magnifications. These micrographs are representative of 30 other regions also acquired from geologic aragonite samples, cryofractured in liquid nitrogen, and coated with 20 nm Pt. They all show flat surface and straight edges resulting from cryofracturing; they never show nanoparticles either 100 or 400 nm in size, as do synthetic aragonite spherulites (A–D) and coral skeletons (E–H).

All of these data combined present direct experimental evidence that the coral skeleton indeed forms by attachment of amorphous precursor particles as previously hypothesized but never directly observed; that these are ACC-H₂O and anhydrous ACC; and that, after they crystallize to aragonite, the particles may retain distinct orientations. This implies that the biomineral is not formed in equilibrium with a large body of water but under direct biological control inside the tissue and behind tightly sealed membranes, as summarized in the model of Fig. 7 and described below.

Model for CaCO₃ Biomineralization via Amorphous Precursor Particles.

A model for calcium carbonate biomineral formation via amorphous precursor particles is presented in Fig. 7. This is based, in part, on these data and in part, on previous work, and it encompasses all that is currently known about this highly investigated topic to the best of our knowledge.

Marine biomineral crystal formation takes place in five steps.

i) Seawater is rich in Ca²⁺ ions but also, Mg²⁺, Sr²⁺, bicarbonate, a few carbonate, and many other ions, all of which are omitted from Fig. 7 for clarity. Seawater is captured by endocytosis, that is, a process in which the cell membrane invaginates, such that the external seawater is incorporated into an intracellular vesicle (on the left in Fig. 7). This is consistent with movies presented by Erez and Braun (45), in which calcein-stained Ca in seawater enters the cells in vesicles and is ultimately incorporated into the growing coral. Additional strong evidence by Tambutté et al. (46) supports this statement in coral, by Khalifa et al. (47) supports this statement in foraminifera, and by Vidavsky (9) supports this statement in sea urchin embryos. The biomineral-forming cells are calicoblastic cells in coral, primary mesenchyme cells in sea urchins, endothelial cells in the mantle of shell-forming mollusks, and the only possible cell in single-cell foraminifera. These are the cells in immediate contact with the deposited mineral (48). If seawater was incorporated by other cells at the surface of a larger, more complex organism, such as coral polyp or a whole-sea urchin embryo, and therefore, far away from these cells, then the vesicle may have been transferred from cell to cell across a variety of tissues. We show only one unstructured tissue in Fig. 7 for simplicity. The color gradient is purely ornamental, not meant to indicate anything.

ii) ACC particles are gradually formed in the vesicle. This implies the addition of CO₃²⁻ ions and the stabilization of amorphous precursor phases, which would otherwise transform into crystals instantaneously but are found to be stable for up to 32 h in this paper, both in the tissue (Fig. 1 C and D) and in the growing biomineral (Figs. 1 B, C, and E and 2–4). We note that, in this work, we see mineral particles in the tissue, both in component maps and in confocal microscopy. These are surprisingly large [400 nm as observed in both methods, once in the fixed (Fig. 1 C and D) and once in the LT (Fig. 5 A and D)], but we cannot say that mineral is localized in vesicles. Previous work by Clode and Marshall (43) showed that 400-nm vesicles were the most abundant objects observed inside cells immediately adjacent to the growing skeleton, when the cells were cryofractured and imaged by cryo-SEM (44). Venn et al. (42) observe half-micrometer mineral particles in coral LT near the skeleton. We, therefore, deduce that all of these separate observations refer to the same objects: ∼400-nm vesicles, delivering ACC to the FM skeleton. Mass et al. (20) used immunolocalization to show that acidic proteins are indeed present in the coral skeleton, as expected, but also throughout the coral tissue layers. Reggi et al. (26) and Falini et al. (27) showed that the organic matrix proteins extracted from coral skeletons stabilize ACC in vitro in the absence of cells. ACC stabilization by proteins was also observed in sea urchin spicules (32) and in fish intestines (49). These proteins, therefore, may be in the same vesicles as ACC.

iii) The carbonate ions are injected into the vesicle one at a time by an active multicomponent biological pathway that remains to be elucidated. Zoccola et al. (50) showed that, in coral, bicarbonate transporters deliver HCO₃⁻ to the calcification site and may, therefore, be one component of this transport into vesicles pathway. The vesicles may be the site where the calcifying fluid studied by Comeau et al. (51) resides. Vidavsky et al. (52) showed that, in sea urchin embryos, intracellular vesicles contain the mineral that will ultimately form the spicule. In our model, the vesicle, initially containing only seawater, gradually and actively is injected with carbonate ions; thus, ion pairs of CaCO₃ form (second, third, and fourth vesicles in Fig. 7). After enough Ca²⁺ ions are paired with carbonate ions, an amorphous hydrated form of calcium carbonate aggregates (fifth vesicle at the bottom in Fig. 7).

iv) The vesicle containing amorphous precursors is ultimately transported toward the cell membrane near the biomineral, and its ACC content is ejected by exocytosis. Thus, it is deposited directly on the growing surface of the biomineral, which is in immediate contact with the cell (or cell processes, as in sea urchin embryos) (9, 52, 53). At this point, deposited particles are still amorphous and, at least in part, hydrated. They then attach to one another and fill space (54). Memory of particle attachment is retained in the mature biominerals, which cryofracture in liquid nitrogen with a granular fracture surface in sea urchin spicules (22), mollusk shell nacre (23), and coral skeletons (20) but not in ascidian vaterite spicules that likely grow ion by ion from solution and present a smooth cryofracture figure at the nanoscale (24).

v) After a day or so, most or all of the ACC-H₂O transformed into ACC and then, into crystalline calcium carbonate (CCC) as shown by the ordered pattern at the bottom of Fig. 7. The ACC-H₂O, ACC, and CCC phases are progressively more stable and thus, energetically downhill (23, 32, 33). The pAra phase has not been characterized energetically, as it is impossible to isolate. Based on peak 1 amplitude at the Ca L-edge (Fig. 1A and Fig. S4), however, we deduce that this phase is more ordered than ACC and less so than aragonite. It is not at all clear, however, that this phase will ever transform into aragonite (23); hence, it cannot be a proto- or a precursor phase (38). In Madracis corals, it does not transform into fully crystalline aragonite in the CoCs, even after a decade (23). One possibility is that this poorly crystalline form of aragonite is disordered by occluded organics (34). In coral and other biominerals, there may be concomitant partial dissolution and reprecipitation, as shown in vitro by Gal et al. (21), but in sea urchin spicules, there is only direct solid-state transformation of ACC into calcite as shown by extensive anhydrous ACC regions in direct contact with calcite in sea urchin spicules (32). In coral skeletons, the anhydrous ACC phase is much more sporadically observed; thus, we conclude that this phase is short-lived (Fig. S7B).

Fig. S7. — Energy landscapes in the tissue and the CoCs (A) and in the coral skeleton (B). These phases are progressively more stable and thus, energetically downhill (32, 33). (A) In the tissue, both barriers are high; thus, both ACC-H₂O and ACC are long-lived. (B) In the coral skeleton, the first is high, and the second low; thus, once deposited in the skeleton, ACC-H₂O is long-lived, and ACC is short-lived. This is deduced from the fact that in the skeleton we find many red pixels (R), many blue pixels (B), and mostly magenta (M = R + B) where the two coexist. In the skeleton we infrequently find green pixels, hence ACC is short-lived, and the activation barrier between ACC and aragonite must be lower than that between ACC-H₂O and ACC.

We now discuss this model.

In biological systems, such as coral, myriad genetic, biochemical, or metabolic complications exist and could, therefore, be included, but they would not help convey a few key ideas; thus, we omit them from this model of CaCO₃ assembly by particle attachment. One key feature that this model describes is that of the separate origins of metal and carbonate ions, which is key to reconciling and explaining extensive and apparently contradictory results on isotopic compositions.

Metal Ions.

The fact that, in the model of Fig. 7, seawater provides the metal ions Ca²⁺, Mg²⁺, and Sr²⁺ satisfactorily explains why the elemental ratios Sr²⁺/Ca²⁺ and Mg²⁺/Ca²⁺ in corals, both modern and fossil, so faithfully represent the environmental temperature. Beautiful data by Beck et al. showed this first (11), and many others followed, including Schrag and Linsley (55). There are deviations from environmental parameters termed vital effects on Sr²⁺/Ca²⁺ and Mg²⁺/Ca²⁺ ratios (56), which various groups ascribed to photosynthesis by the zooxanthellae (6, 57), pH (42, 58), location in the coral (7, 59), or Rayleigh fractionation (60, 61), but when Sr²⁺/Ca²⁺ and Mg²⁺/Ca²⁺ ratios work as temperature proxies, they work extremely well; thus, they cannot be ignored. Ca²⁺, Mg²⁺, and Sr²⁺ ions must come from seawater, and when CaCO₃ first assembles in the vesicle, it must be in the presence of at least a few Mg²⁺ and Sr²⁺ ions for their T-dependent incorporation into the biomineral.

Carbonate Ions.

Many authors lament over vital effects (5, 62) on isotopic ratios δ¹⁸O and δ¹³C in coral, but Adkins et al. (7) showed magnificent data on δ¹⁸O and δ¹³C, which strongly correlate with one another. Again, when these isotopic ratios work, they work extremely well; thus, we propose a possible explanation: C and O are both part of the carbonate –CO₃²⁻ ion, which in our model, is formed in the tissue and by the organism, not from atmospheric CO₂, –HCO₃⁻, or –CO₃²⁻ from seawater.

Since both C and O result from biological functions, any parameter affecting C also affects O in the same direction and by the same amount. This is also in agreement with Spero et al. (63), who showed that that C and O isotopes covary with carbonate ion concentration and hence, with each other, providing a simple explanation for δ¹⁸O and δ¹³C data, which is source-independent. It is possible that the coral of Adkins et al. (7) is simpler, because it is a deep sea coral without photosynthetic symbionts, and photosynthesis has been shown to strongly affect δ¹³C values in coral (6). The major source of carbonate ions is metabolic CO₂ (15, 64, 65), and bicarbonate may exit the cells via a bicarbonate transporter (50), whereas Ca²⁺ ion transport is dependent of voltage-gated calcium channels (66). A schematic summarizing ion transport, physiological, and molecular components of coral biomineralization was presented by Bhattacharya et al. (ref. 67, figure 2).

Metal and Carbonate Ions Have Distinct Origins.

The key point of our model is that the metal ions and the carbonate ions have completely different origins, are processed independently, and therefore, show inconsistent behavior, previously assumed to be a problem if one wants to use coral skeletons to measure paleoenvironmental temperatures. In fossils, one of the few if not the only way to validate temperatures measured by Sr²⁺/Ca²⁺ ratios is to compare them with temperatures measured by δ¹⁸O or the clumped isotope thermometer (68). Such comparisons may be extremely successful, as in the work by Ghosh et al. (68), but they also frequently fail. The model that we propose here predicts that the metal and the carbonate origins are different; thus, they should not behave similarly universally. In addition, when ACC is formed in intracellular vesicles, as shown here, the metal ratios may also be altered by small organic molecules, as observed in vitro for Mg²⁺/Ca²⁺ ratios by Dove and coworkers (69), or by deposition rates as shown in coccolithophorids and foraminifera (70, 71). Some of the vital effects, that is, deviations of elemental and isotopic compositions predicted by thermodynamics and environmental parameters alone, observed in corals (7, 48, 59, 60) may be reinterpreted in the light of the data and model presented here for coral formation via attachment of amorphous particles, in direct contrast with the previously assumed precipitation of coral minerals ion by ion from solution.

Selection of Phase.

One might ask why ACC-H₂O is produced rather than more stable phases, such as aragonite. It is actually common in natural systems that the least stable phase that is energetically available is the one that forms first. This observation is known as the Ostwald Step Rule or the Gay–Lussac–Ostwald Step Rule (72, 73). The high supersaturation produced by the injection of carbonate ions into the vesicle provides the energy needed to produce the less stable phase. We acknowledge that the Ostwald Step Rule only considers classical ion-by-ion pathways of nucleation and growth (73) and that much remains to be learned about the processes involved in crystallization by attachment of amorphous particles (22).

Particle Attachment.

The particle attachment mechanism described here is quite different from the oriented attachment of crystalline particles described by Penn and Banfield (74) in synthetic biomineral systems, Banfield et al. (75) in bacterial biomineral systems, and Cölfen and Antonietti (76, 77) in the formation of a variety of mesocrystals. In oriented attachment, nanoparticles are crystalline first, and then, they attach to one another, whereas in coral skeletons, nanoparticles are initially amorphous; then, they attach and fill space, and hours later, they crystallize.

What is the advantage to the animal of formation by attachment of amorphous particles as opposed to precipitation from solution? One hypothesis explored here involves aragonite growth rate. Corals with fast-growing skeletons outcompete their neighbors in crowded reefs by reaching farther for better illumination, exposure to nutrients, and disposal of waste or for faster regeneration after damage by tropical storms or predators (78). Thus, fast growth must be a terrific physiological advantage for the coral polyps and the symbiont zooxanthellae in the rigors of crowded and competitive reef ecosystems.

How do growth rates by attachment of ions or ACC particles compare? Aragonite growth experiments in the laboratory using realistic seawater pH, temperature, and concentrations show an average linear growth rate of 0.22 µm/d (Table 2). S. pistillata, the coral used here, elongates its skeleton an average of 40 µm/d (Table 3), thus more than 100 times faster than it would if its aragonite precipitated ion by ion from solution.

Table 2.

Abiotic aragonite ion-by-ion growth rate

Study	p (µm/d)	n	r (µm/d) at S = 3.695	T (°C)	pH	Other notes
Burton and Walter, 1987 (82)	0.033	1.70	0.18	25	NA	Salinity 35, filtered Gulf Stream water
Mucci et al., 1989 (83)	0.030	2.20	0.27	25	7.53–7.70	ASW without sulfate
Mucci et al., 1989 (83)	0.019	1.60	0.09	25	7.60–7.94	ASW with sulfate
Zhong and Mucci, 1989 (84)	0.011	2.26	0.10	25	7.48–7.67	ASW, salinity 35–44
Gutjahr et al., 1996 (85)	0.165	1.05	0.47	20	7.91	ASW
Average of all growth rates			0.22

Table 3.

Biogenic aragonite particle-by-particle growth rate

Study	S. pistillata linear growth rate (various units)	S. pistillata growth rate (mm/y)	S. pistillata growth rate (µm/d)
Liberman et al., 1995 (ref. 86, figure 2)	0.5–2.0 cm in 7 mo	8.6–34.3	24–94
Kotb, 2001 at 5-m depth (87)	9.24 mm/y	9.24	25
Shaish et al., 2006, Bal in table 1 in ref. 88	7.6–13.3 mm/y	7.6–13.3	21–36
Average of all growth rates			40

Is this astounding growth rate achieved biochemically, or is it an abiotic phenomenon? To tackle this question, we grew synthetic aragonite spherulites, which form by attachment of nanoparticles in vitro as shown in Fig. 6B and Fig. S6 in the absence of any organic molecules, and measured their growth rates, as presented in Table 4. Again, the growth rate is at least 100 times greater than for ion by ion growth. We, therefore, conclude that it is growth via ACC particle attachment that enables fast aragonite crystal growth. Biochemical growth modifiers may also play a significant role in coral skeleton growth rate, but in vitro particle-by-particle growth significantly increases growth rate compared with ion-by-ion growth and does so in the absence of any organic molecules. (Biological control is still all important in forming, stabilizing, and delivering ACC particles to the growing biomineral!)

Table 4.

Abiotic aragonite particle-by-particle growth rate

Study	Synthetic spherulite linear growth rate (µm/d)
This study, range across 15 spherulites	14–60
Average of all growth rates ± SD	29 ± 12

Coral skeletons formed via attachment of amorphous particles retain a nanoparticulate texture even years after formation, as is visible in PIC maps (Fig. 6A) and also by SEM of cryofractured coral skeleton (Fig. S6) but not in geologic aragonite (Fig. S6). Other authors saw ∼100-nm particles with SEM or AFM in corals (19, 20), as do we in Fig. S6 E–H along with other ∼400-nm or larger features. This reinforces the well-established notion (22) that morphology alone cannot provide accurate measurement of particle sizes, but combined with PIC mapping and component mapping, it provides a more complete and therefore, stronger case for coral crystallization by particle attachment. There may also exist concomitant ion-by-ion precipitation, but granular structure in SEM, in crystal orientation PIC maps, and in direct observation of particles in the tissue all concur to show that particle attachment plays a major role in coral skeleton formation. Even slow-growing corals in the slowest growth conditions of low light (79) still grow faster than ion-by-ion aragonite (Table 2).

Therefore, our hypothesis that growth by attachment of amorphous precursor particles confers the advantage of fast skeleton growth to corals seems to be confirmed by data: aragonite crystal growth via particle attachment in both synthetic spherulites and coral skeletons is more than 100 times faster than ion-by-ion growth from solution. This must be an advantage in the competitive and crowded reef environment. It also provides an avenue to fine-tune paleoclimate proxies using synthetic crystals grown by particle attachment. Coral bleaching (80) and postmortem skeleton dissolution (81) are increasingly occurring, but coral skeleton growth rate in S. pistillata will not necessarily be decreased by ocean acidification. Additional experiments on growth rates at lower pH will shed light on this possibility. If other corals also form their skeletons by particle attachment, a calcification crisis may or may not occur in acidifying oceans, because ACC particles are formed inside coral tissue.

SI Materials and Methods

Coral Sample Preparation.

Coral fragments from the zooxanthellate coral S. pistillata were obtained from corals growing in an 800-L aquarium system at the Marine and Costal Science Department at Rutgers University. The fragments were shipped to Berkeley and acclimated for 2 wk in an aquarium at 25 °C. Twenty hours before each experiment, a ∼5-mm coral nubbin was cut and rapidly fixed in 2% paraformaldehyde and 0.05 M sodium cacodylate buffer in 22 g/L Na₂CO₃ for 1 h. After fixation, the samples were washed twice for 5 min each with 0.05 M sodium cacodylate buffer in 22 g/L Na₂CO₃ and then dehydrated in 50, 60, and 70% mixtures of anhydrous ethanol and 1 g/L Na₂CO₃ followed by 80 and 90% anhydrous ethanol and 0.5 g/L Na₂CO₃ and finally, 100% anhydrous ethanol twice. At each dehydration step, the coral samples were left in the solution for 5 min. The samples were then embedded and polished following previously established protocols (96) and coated with 40-nm Pt masking off the area to be analyzed and then, 1-nm Pt (97) on the whole sample, as described in ref. 98. Specifically, we optimized the polishing process to preserve any ACC present by dialyzing all polishing and cleaning solutions against Na₂CO₃, which is 10⁴ times more soluble than ACC; thus, the solution is greatly supersaturated with respect to ACC, and ACC cannot dissolve. To prevent dissolution, the saturation needs not be for both Ca and CO₃: supersaturating only in CO₃ is sufficient. This method was first published in the work by Gong et al. (32), and we have successfully used this method in all following publications on any amorphous precursors, including this one on coral.

Growth of Synthetic Spherulites.

Synthetic aragonite spherulites were grown using the ammonium carbonate [(NH₄)₂CO₃] diffusion method. (NH₄)₂CO₃ vapor sublimating from the solid in a closed table top desiccator was allowed to diffuse for 24 h into 1 mM CaCl₂ and 5 mM MgCl₂ solution in a beaker placed into the same desiccator (99). Each beaker contained a glass slide, on the surface of which the spherulites nucleated and grew. They were then scraped off the glass slide for SEM analysis (Fig. S6 A–D) or for embedding in epoxy, polishing, and coating for PIC mapping following previously established protocols (96) and coated with 40-nm Pt around it and then, 1-nm Pt (97) in the area to be analyzed, as described in ref. 98. Spherulite diameters were measured with NIH ImageJ, selecting the largest diameter visible in a 2D polished cross-section of each spherulite and dividing it by two to obtain the radius, and thus, the linear growth rate became directly comparable with those in ion-by-ion growth from solution of aragonite (Table 2) and S. pistillata coral (Table 3). The spherulite growth rate along the radii of 16 spherulites, grown in four different batches, is presented in Table 4. For the PIC mapping analysis in Fig. 6B, spherulites were embedded, polished, and coated like the coral samples. We observe particles in spherulites in SEM images and PIC maps, but we did not study these particles with spectromicroscopy; thus, we cannot say that they were ACC before attachment to the growing spherulite. For this reason, we did not state anywhere here that these were ACC particles.

PEEM.

Photoelectron emission spectromicroscopy or PEEM experiments to obtain component maps or PIC maps were done using PEEM-3 on beamline 11.0.1.1 at the Advanced Light Source, Lawrence Berkeley National Laboratory (96).

Component maps.

For component maps, we acquired 121 images across the Ca L edge, with energy step 0.1 eV between 345 and 355 eV and 0.5 eV in the featureless regions below and above this range: 340–345 and 355–360 eV, respectively. The images were stacked and processed using the Gilbert Group Macros (GG Macros). This software for PC or Mac runs in Igor Pro Carbon and can be downloaded free of charge (100). Spectra from each pixel were analyzed and best fitted to a linear combination of four component spectra: ACC-H₂O, ACC, pAra, and aragonite (termed “refs 0823”). The linear combination was optimized using nonlinear least square analysis in GG Macros (100). At the end of the fit for all 10⁶ pixels in each stack, the proportion of each component was assigned a color, and RGBY maps were created, which represent the Ca mineral composition and distribution. The four component spectra in refs 0823 were obtained as follows:

ACC-H₂O from spicules, Gong et al. (32) (six single-pixel spectra);
ACC from spicules, Gong et al. (32) (eight single-pixel spectra);
pAra from Madracis coral, DeVol et al. (23) (157 single-pixel spectra); and
Aragonite from S. pistillata coral in this work (100 single-pixel spectra).

All spectra were obtained from multiple single 20-nm-pixel spectra (number in parentheses above), first aligned, then averaged, and peak-fitted. Only the four noise-free fitted spectra were used as components. Before component analysis, the four spectra were normalized so that the area under the curve was identical for all four between 345 and 355 eV (100). All obtained component maps were brightness enhanced and masked (100). Using Adobe Photoshop, all resulting red, green, blue, yellow (RGBY) component maps were cropped, and a scale bar was added. In all cases, we used four components, as shown in the color legend in Figs. 1 and 2. In Figs. 3 and 4, however, the color legends do not include yellow pAra to indicate that, in those cases, there were no yellow pixels detected.

Timing of component mapping.

To detect amorphous precursor phases, we must acquire Ca movies as soon as possible after the death of the animal. However, sample prep is not fast. Fixation takes 1 h; epoxy takes 13 h to cure; and a mirror-flat surface takes 6–8 h to polish and coat, 1 h to image the sample with visible light microscopy to enable navigation in PEEM, and another 1 h to transfer the sample into the vacuum chamber. The earliest that we can ever take data is 22 h after the death of the animal. We continuously take data until it is about 32 h old; then, we stop and move to another sample, which was prepared by some of us while others took data. This is done around the clock for several days during each beamtime. Then, we spend 6–12 mo processing the terabytes of data collected. Between 22 and 32 h postmortem, some ACC can occasionally be found. After 32 h, we never found any ACC. We, therefore, conclude that it spontaneously transforms into aragonite. What happened during the first 22 h, we cannot know, but given the decrease observed between hours 22 and 32, it is logical to deduce that more mineral was amorphous before analysis: maybe not 100% of the minerals but certainly, more than the observed maximum of 20%.

PIC maps.

A stack of 19 images was acquired for each PIC map (40, 41, 101, 102). In these images, the photon energy was fixed at the oxygen K-edge π* peak, which is a sharp, intense peak at 534 eV extremely sensitive to polarization in all carbonates (40). The linear polarization angle in the illuminating X-rays was then rotated by 90° (from horizontal to vertical) in 5° steps. PIC maps were produced from the stack of 19 polarization images using the Polarization Analysis Package in GG Macros (100).

Detailed descriptions of PIC mapping from stacks of images at varying polarization were previously published (40, 41, 103, 104). Briefly, the intensity vs. polarization angle χ curve from each pixel is fit to

f(χ)=a+b⁡cos2(χ−c′),

[S4]

where a, b, and c′ are fit parameters. The colors produced in PIC maps describe the angle of the projection of the c axis onto the polarization plane, termed the c′ axis. When this axis is vertical (in the laboratory and in the images), the angle is 0°, and the color is cyan; when the c′ axis is horizontal left (−90° from the vertical) and right (+90°), the color is red, and intermediate colors follow the color legends in Figs. 2D, 4D, 6 and Fig. S1E. Color PIC maps also show with brightness the c axis orientation off-plane (polarization plane): bright colors indicate in-plane c axes, darker colors are off-plane, and black is 90° off-plane (that is, the c axis is pointing directly into the X-ray beam) (40). For all PIC maps presented here, the default (100) parameter values were used: angle minimum = −90, maximum +90, scale bar colors 0°–360°, brightness set to maximum B = 100 in Fig. 4D, maximum B = 150 in Figs. 2D and 6 and Fig. S5, and maximum B = 70 in Fig. S1E. Using Adobe Photoshop, all PIC maps were cropped, and scale bars were added.

RGBY mapping.

Displaying four mineral phases with three colors, R, G, and B, may be ambiguous. However, in coral, we find that Y never occurs as a mixture of R + G = Y. We checked this carefully in all coral component maps and show representative examples in Fig. S8. Thus, using an RGBY map is unambiguous. For additional clarification, we have presented the separate distribution maps for all four components in Fig. S9.

Fig. S8. — Comparison of RGB and RGBY maps from details of the regions in Figs. 1 and 2, the only ones that contain any pAra phase. Both maps were obtained analyzing each pixel with a linear combination of all four components, but in RGB and RGBY maps, only three or all four of them are displayed, respectively. Notice that, in the RGB maps, there are no Y pixels (*Upper*) or nearly no Y pixels (*Lower*), which if present, would result from R + G = Y. Thus, the presence of Y pixels in the RGBY map can unambiguously be interpreted as pAra.

Fig. S9. — Column 1 shows the RGBY maps for each of Figs. 1–4, and columns 2–5 show the separate R, G, B, and Y, respectively. Each of the separate maps shows the spatial distribution of R, G, B, and Y pixels on a grayscale from 0 to 255, corresponding to black (0), and pure R, G, B, or Y to white (255), respectively. Table 1 shows the numeric data on these maps.

Pixel counting in component maps.

The component analysis was done in Igor using the four reference spectra in Fig. 1A as components. After RGBY component mapping, the unscaled proportion maps (called pMaps in Igor) were exported for Figs. 1–4 and all other component maps here, and the RGBY maps were exported and displayed here in Fig. S9, column 1. Counting of all pixels with each color was done in NIH ImageJ 1.49v using the four eight-bit grayscale pMaps for each component as displayed in Fig. S9.

Each pMap was opened in ImageJ as an eight-bit grayscale image, with pixel values from 0 to 255 (Fig. S9). The pixel values and their coordinates were exported from each pMap while selecting the entire image (Analyze > Tools > Save XY Coordinates). The pixel values and coordinates were then combined into one table, and the black pixels (with R = G = B = Y = 0) were removed. The histogram of all pixels with each color (Analyze > Histogram) showed the distribution of pixels with gray scale values from 0 (pure black) to 255 (pure red, green, or blue or yellow). Each histogram showed a high proportion of pixels in the lowest 5% of the color range. The lowest 5% (pixels with values from 0 to 12) had insignificant proportions of components, resulting from fitting noise in the spectra, and therefore, were disregarded. For the remaining pixels, in Table 1, we report the abundance of red, green, blue, and yellow pixels as the percentage of all colored pixels containing at least 5% of that component in the component map.

Radiation damage.

The effect of radiation damage is crystallization as described in detail by Gong et al. (32) and as observed in all repeat acquisitions here in Fig. S3. We, therefore, are very careful to always only acquire data on virgin, never exposed before, areas. As in most areas, on this septum, we first acquired two Ca movies for component analysis and then, two O movies for PIC mapping. For the septum in Fig. 4, for example, by the time that we acquired the first O movie, all of the amorphous minerals had already crystallized to aragonite. This happened in the majority of areas analyzed. In fact, the few cases in which we could beat radiation damage are those presented here and a few additional areas. It is important to note that, if ACC-H₂O or ACC is detected, their presence cannot be an artifact of radiation damage. If aragonite is detected, however, it may be either an artifact of radiation damage or the result of natural coral skeleton crystallization, or both. The two are indistinguishable, as radiation damage-induced crystallization is epitaxially cooriented with the immediately adjacent crystalline skeleton. The radiation dose imparted to the sample by acquiring O movies for PIC mapping is greater than that for Ca movies for component mapping. In CoCs, all amorphous phases are more stable and remain detectable in repeated component mapping and PIC mapping, as shown in Fig. 2. Notice that synthetic ACC crystallizes in seconds, especially in wet environments (33); thus, our observation that ACC-H₂O is detectable after more than 30 h implies that something must be stabilizing it.

Confocal Microscopy.

Coral spats settled on the glass bottom of Petri dishes 9 d before analysis. All observations were made either with an inverted fluorescent microscope equipped with polarization equipment (Nikon ECLIPSE TI-E) for the image in Fig. 5C or with an inverted confocal laser-scanning microscope (Nikon A1-Red) for the image in Fig. 5A.

For localization of calcium particles (Fig. 5) in the spat LTs or FM, the spats were incubated overnight in 2.6 µM blue calcein (Sigma 54375–47-2) dissolved in seawater. They were then rinsed for 1 h in sea water before observation with fluorescence confocal microscopy. The image in Fig. 5A was obtained from a stack of 10 images, acquired focusing at different heights from the bottom of the Petri dish. Excitations at 405, 488, 561, and 638 nm in addition to differential interference contrast (DIC) filter were used. Blue calcein is displayed in blue, GFP (native in S. pistillata and all other corals) is displayed in green, and chlorophyll in red (native in coral symbionts).

In Fig. 5, the top one-half of the FM is part of a basal plate (BP), which eventually will cover the entire Petri dish with mineral. Green islands of tissue are clearly visible in this region of forming skeleton. The bottom one-half of the FM is a growing septum (S). At the edge of the spat is a layer of stinging cells (SCs).

The micrograph in Fig. 5C was acquired 5 d after the one in Fig. 5A; thus, it shows a similar region, not an identical one, as the animal kept growing and became thicker and as the BP was more complete. In this micrograph, however, mineral crystals are easier to identify, as they appear as bundles of acicular crystals fanning from CoCs. From this and other similar images acquired the same day as the one in Fig. 5A, it was possible to recognize the forming BP, the forming S, and the layer of SCs.

Particle Size Analysis.

The blue calcein-stained particles in Fig. 5 were analyzed as follows. We opened the blue calcein only component of the confocal micrograph in Fig. 5A in ImageJ 1.49v, applied a threshold, then analyzed the 2,726 particles, and thus, obtained their outlines and a listing of particle areas. We processed the list in Microsoft Excel, where we calculated the square roots of all areas and then, calculated their means and SDs. Many particles are not squares, but for simplicity, we assumed that they were squares and reported the side of the square as particle size. We did this identically also for the three particles in Fig. 1 C and D and calculated the mean, the median, and the mode of all particle sizes.

Four hundred nanometers is the most frequently observed (mode) and the median particle size in both LT and FM. To be precise, these are 417 ± 711 nm (mode = median ± SD). This is also the resolution of confocal micrographs; thus, this size is to be considered an upper bound of fluorescent particle sizes. Because LT has more particles greater than 1 µm, the average particle size is 1.0 ± 0.9 µm in the LT and 0.7 ± 0.6 µm in the FM, although the median and the mode are identical and 400 nm in both LT and FM. The difference in particle size distribution between LT and FM is statistically significant (P = 1.1 × 10⁻⁷, Kolmogorov–Smirnov test). The difference in particle size distributions between LT and FM regions in Fig. 5 was tested using the Kolmogorov–Smirnov test (105) in a Microsoft Excel statistical package following the method explained in www.real-statistics.com/non-parametric-tests/two-sample-kolmogorov-smirnov-test/. Both data from LT and FM were first binned with a 20-nm bin size from 0 to 10⁴ nm. With a P value of P = 1.1 × 10⁻⁷, the null hypothesis, that the two sets of data come from the same distribution, is rejected.

To analyze the size of differently oriented particles in solid skeleton or spherulites, however, this method did not work: there is no rigorous way to threshold the image to highlight the smallest particles; thus, the above method could not be used.

To identify and measure the smallest particles in Fig. 6 and Fig. S5 C and D, we displayed them in Adobe Photoshop CC 2014, zoomed in to find the smallest particles of any color isolated from the surrounding color, and labeled them by hand with a 100-nm square just above each particle. We then numbered each particle with a number label right below it. We then opened the image in ImageJ 1.49v and acquired a horizontal line profile across each particle. We then displayed all line profiles in Kaleidagraph 4.5, zoomed in, and carefully measured the FWHM of each particle. Eight particles in the coral skeleton and eight particles in the spherulite were analyzed; then, the mean and SD were calculated in Microsoft Excel, which were 143 ± 52-nm particles in coral skeleton and 72 ± 26 nm in synthetic spherulite (mean ± SD).

Microdiffraction.

Microdiffraction experiments were done on beamline 12.3.2 at the Advanced Light Source using powder diffraction with monochromatic 8-keV X-rays, illuminating an S. pistillata skeleton with 5 × 5-µm beam size in 60 locations on CoCs and acicular aragonite larger crystals. The coherence length was measured from the broadening of powder diffraction rings using the Scherrer equation:

where S is the size of a coherent domain, λ is the wavelength of the incident X-ray beam, θ is the Bragg angle of the powder diffraction ring, and w is the FWHM of the ring integrated along the azimuthal direction and corrected with instrumental broadening.

The data show coherence lengths of 250–700 nm in the CoCs and 300–1,700 nm in the large crystalline domain areas. The average coherence lengths in the CoCs and large crystal areas are 350 ± 100 and 800 ± 400 nm, respectively (average ± SD).

Materials and Methods

Detailed methods are provided in SI Materials and Methods. Briefly, S. pistillata corals, adults and spats, were fixed, dehydrated, embedded, polished, and coated for PEEM analysis as described previously (23, 32). For confocal microscopy, they were exposed to blue calcein in seawater, washed, and observed alive, so that blue calcein particles could be imaged in the LT and in the forming biomineral. For PEEM experiments, sample preparations were rapid and repeated to capture the transforming amorphous phases; PEEM data acquisition was done exposing the sample to as little radiation damage as possible, which has the effect of crystallizing amorphous phases. Data processing was done using Ca L-edge component mapping with four spectral components (ACC-H₂O, ACC, pAra, and aragonite) after normalizing these spectra to have the same area under the whole spectrum between 345 and 355 eV. With this approach, quantitative analysis of the concentration of each component can be done, giving equal weight to crystalline and amorphous phases. This normalization differs from the ones used previously (23, 32).

Note Added in Proof.

Two recently published articles are closely related to this work and strongly support it. Sun et al. (106) demonstrate that S. pistillata corals grow their skeletons spherulitically. Walker et al. (107) characterize in vitro a solid state transformation from ACC to aragonite similar to the biogenic one observed here.

SI Growth Rates

Growth Rate of Aragonite Deposited Abiotically Ion by Ion from Solution.

We want to estimate how quickly aragonite can grow abiotically directly from solution rather than by attachment of amorphous particles. Growth rates may be estimated from the widely used (85, 89, 90) empirical relationship of growth rate r with saturation S:

where p is the kinetic rate constant and n is the reaction order, which are two parameters measured in conditions of temperature and pH realistic for seawater; S is the saturation defined as

with K as the solubility product of aragonite at 25 °C, which is K = 6.65 × 10⁻⁷ mol²/kg² from the work by Morse et al. (91). Several experiments measured abiotic growth rates in seawater conditions of seeded aragonite crystals as a function of saturation and report a relatively small range of values for p and n (Table 2).

Rate constant p in all of the studies in Table 2 is originally normalized to surface area (moles area⁻¹ time⁻¹). To compare area-normalized aragonite growth rates with the linear growth rates measured in coral skeletons, we convert all P values to micrometers per day using aragonite molecular weight (100.1 g/mol) and density (2.93 g/cm³). An example of this unit conversion for p from Gutjahr et al. (ref. 85, table 3) is given below:

p=0.056×10−10molcm2 s⋅100.1gmol⋅1 cm32.93 g⋅10,000 μmcm⋅86,400 sd=0.165μmd.

[S3]

Using Eq. S2 and the standard seawater concentrations of Ca and carbonate given by Millero et al. (ref. 92, table 4), which are 0.010282 and 0.000239 mol/kg, respectively, we obtain a saturation with respect to aragonite, S = 3.695.

Using the obtained value p = 0.165 µm/d and n = 1.05 from Gutjahr et al. (85) in Eq. S1, we obtain a growth rate r = 0.47 µm/d.

This is an upper-limit growth rate calculated from the largest measured P value in Table 2 and unconstrained by the presence of divalent cations other than Ca or any other inhibitors. In Table 2 we show the results of similar calculations for this and four more experiments on abiotic ion-by-ion aragonite growth in vitro, and average their results to obtain 0.22 μm/d.

Growth Rate of Biogenic Aragonite in Stylophora pistillata Coral Skeletons.

The linear extension (how quickly branches elongate) measured by various authors in S. pistillata living corals is reported in Table 3. Other authors report growth rates in units not convertible to linear growth rate with the information available in their papers; thus, these were not included in Table 3 (93–95).

Growth Rate of Aragonite Deposited Abiotically by Attachment of ACC Nanoparticles.

We grew aragonite spherulites using a 5MgCl₂:1CaCl₂ solution and letting ammonium carbonate vapors in a closed desiccator diffuse through the solution for 24 h. The solution did not contain any organic molecules. Using this method, we obtained elongated aragonite radial crystals in each spherulite grown by attachment of particles, not individual ions from solution, as shown in Fig. S6 A–D. We then embedded the spherulites and measured the longest diameter in PIC maps of their cross-sections, with the results reported in Table 4.

Comparison of Growth Rates.

Comparing the growth rates in abiotic aragonite (Table 2) and biogenic aragonite in S. pistillata corals (Table 3), we find a ratio varying between 45 and 1,044 times faster growth for coral. Comparing the averages, which given the scatter of the data, are more reasonable, we obtain 40/0.22 = 182 times faster for coral vs. abiotic growth. We thus conclude that the growth rate is at least 100 times faster in S. pistillata coral grown by attachment of amorphous particles than in aragonite grown abiotically ion by ion.

Comparing the growth rates in abiotic aragonite grown by particle attachment (Table 4) and by ion attachment (Table 2), we obtain 29/0.22 = 132.

Even in the absence of any organic molecules, therefore, aragonite crystals grow at least 100 times faster when they attach one particle at a time rather than one ion at a time.

Some of the vital effects (that is, deviations of elemental and isotopic compositions predicted by thermodynamics and environmental parameters alone) observed in corals (7, 48, 59, 60) may be reinterpreted in the light of the data presented here for coral formation via amorphous precursor particles, in direct contrast with the previously assumed precipitation of coral minerals from solution.

We stress, however, that the high growth rate alone cannot explain the vital effect, as it can be reproduced in the absence of any organic molecules. The vital effects must be associated with the formation of ACC nanoparticles in the coral tissue, which must occur under biological control.

SI Phases and Their Transformations

Since both ACC-H₂O and ACC are abundant in the coral tissue particles, they must both be long-lived in the tissue, presumably stabilized by organic molecules. In the forming skeleton, ACC-H₂O is abundant, but ACC is rare; thus, the transition to crystalline aragonite must not be inhibited by organics after particles are attached to the growing skeleton. These deductions are summarized in the energy landscapes in Fig. S7. These differ from those observed in sea urchin spicules (32) and nacre (23), where ACC-H₂O and ACC are short- and long-lived, respectively. A theoretical framework for these energy landscapes was described by De Yoreo et al. (ref. 22, figure 4).

In coral and other biominerals, there may be concomitant partial dissolution and reprecipitation, as shown in vitro by Gal et al. (21), but in sea urchin spicules, there is only direct solid-state transformation of ACC into calcite, as shown by extensive anhydrous ACC regions in direct contact with calcite in sea urchin spicules (ref. 32, figure 2B or figure 3, 48 h).

In coral, the anhydrous ACC phase is much more sporadically observed; thus, we conclude that this phase is short-lived, as shown in Fig. S7B. Dissolution and reprecipitation would imply no ACC ever anywhere. We find a few ACC pixels in every acquisition in coral (Table 1); thus, there must be a solid-state transformation from anhydrous ACC to crystalline aragonite. This rapid, solid-state transformation may be the only mechanism of crystallization, or it may occur along with partial dissolution and reprecipitation as aragonite. The present data cannot rule out this possibility.

Acknowledgments

We thank Andrew H. Knoll, Steve Weiner, Lia Addadi, and Paul Falkowski for reading the manuscript and suggesting improvements; Elia Beniash for introducing T.M. to P.U.P.A.G.; Paul Falkowski for supplying the corals used for this study; and Jonathan Stillman for use of his aquarium facility at the University of California, Berkeley. We also thank Andreas Scholl (Advanced Light Source) and Palle Von Huth (Center of Microscopy and Imaging, University of Haifa) for their assistance during experiments. T.M. acknowledges support from Israel Science Foundation Grant 312/15 and United States–Israel Binational Science Foundation Grant BSF-2014035. P.U.P.A.G. acknowledges 80% support from US Department of Energy, Office of Science, Office of Basic Energy Sciences, Chemical Sciences, Geosciences, and Biosciences Division Award DE-FG02-07ER15899; 19% from National Science Foundation Grant DMR-1603192; and 1% from United States–Israel Binational Science Foundation Grant BSF-2010065. PEEM and microdiffraction experiments were done at the Advanced Light Source, which is a Department of Energy Office of Science User Facility supported by Grant DE-AC02-05CH11231.

Footnotes

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

²Previously publishing as Gelsomina De Stasio.

References

1.Allemand D, Tambutté É, Zoccola D, Tambutté S. Coral calcification, cells to reefs. In: Dubinsky Z, Stambler N, editors. Coral Reefs: An Ecosystem in Transition. Springer; Berlin: 2011. pp. 119–150. [Google Scholar]
2.Tambutté S, et al. Coral biomineralization: From the gene to the environment. J Exp Mar Biol Ecol. 2011;408:58–78. [Google Scholar]
3.Davy SK, Allemand D, Weis VM. Cell biology of cnidarian-dinoflagellate symbiosis. Microbiol Mol Biol Rev. 2012;76:229–261. doi: 10.1128/MMBR.05014-11. [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Constantz BR. Coral skeleton construction; a physiochemically dominated process. Palaios. 1986;1:152–157. [Google Scholar]
5.Weiner S, Dove PM. An overview of biomineralization processes and the problem of the vital effect. Rev Mineral Geochem. 2003;54:1–29. [Google Scholar]
6.Cohen AL, Owens KE, Layne GD, Shimizu N. The effect of algal symbionts on the accuracy of Sr/Ca paleotemperatures from coral. Science. 2002;296:331–333. doi: 10.1126/science.1069330. [DOI] [PubMed] [Google Scholar]
7.Adkins JF, Boyle EA, Curry WB, Lutringer A. Stable isotopes in deep-sea corals and a new mechanism for “vital effects.”. Geochim Cosmochim Acta. 2003;67:1129–1143. [Google Scholar]
8.Blamart D, et al. Correlation of boron isotopic composition with ultrastructure in the deep‐sea coral Lophelia pertusa: Implications for biomineralization and paleo‐pH. Geochem Geophys Geosyst. 2007;8:1–11. [Google Scholar]
9.Vidavsky N, et al. Calcium transport into the cells of the sea urchin larva in relation to spicule formation. Proc Natl Acad Sci USA. 2016;113:12637–12642. doi: 10.1073/pnas.1612017113. [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Lowenstam HA, Weiner S. On Biomineralization. Oxford Univ Press; Oxford: 1989. [Google Scholar]
11.Beck JW, et al. Sea-surface temperature from coral skeletal strontium/calcium ratios. Science. 1992;257:644–647. doi: 10.1126/science.257.5070.644. [DOI] [PubMed] [Google Scholar]
12.Zachos JC, et al. Rapid acidification of the ocean during the Paleocene-Eocene thermal maximum. Science. 2005;308:1611–1615. doi: 10.1126/science.1109004. [DOI] [PubMed] [Google Scholar]
13.Gothmann AM, et al. Fossil corals as an archive of secular variations in seawater chemistry since the Mesozoic. Geochim Cosmochim Acta. 2015;160:188–208. [Google Scholar]
14.Anthony KR, Kline DI, Diaz-Pulido G, Dove S, Hoegh-Guldberg O. Ocean acidification causes bleaching and productivity loss in coral reef builders. Proc Natl Acad Sci USA. 2008;105:17442–17446. doi: 10.1073/pnas.0804478105. [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Pearse VB. Incorporation of CO2 into coral skeleton. Nature. 1970;228:383–383. doi: 10.1038/228383a0. [DOI] [PubMed] [Google Scholar]
16.Raz-Bahat M, Erez J, Rinkevich B. In vivo light-microscopic documentation for primary calcification processes in the hermatypic coral Stylophora pistillata. Cell Tissue Res. 2006;325:361–368. doi: 10.1007/s00441-006-0182-8. [DOI] [PubMed] [Google Scholar]
17.Cohen AL, McConnaughey TA. Geochemical perspectives on coral mineralization. Rev Mineral Geochem. 2003;54:151–187. [Google Scholar]
18.Clode PL, Lema K, Saunders M, Weiner S. Skeletal mineralogy of newly settling Acropora millepora (Scleractinia) coral recruits. Coral Reefs. 2011;30:1–8. [Google Scholar]
19.Stolarski J. Three-dimensional micro-and nanostructural characteristics of the scleractinian coral skeleton: A biocalcification proxy. Acta Palaeontol Pol. 2003;48:497–530. [Google Scholar]
20.Mass T, Drake JL, Peters EC, Jiang W, Falkowski PG. Immunolocalization of skeletal matrix proteins in tissue and mineral of the coral Stylophora pistillata. Proc Natl Acad Sci USA. 2014;111:12728–12733. doi: 10.1073/pnas.1408621111. [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Gal A, et al. Particle accretion mechanism underlies biological crystal growth from an amorphous precursor phase. Adv Funct Mater. 2014;24:5420–5426. [Google Scholar]
22.De Yoreo JJ, et al. Crystallization by particle attachment in synthetic, biogenic, and geologic environments. Science. 2015;349:aaa6760. doi: 10.1126/science.aaa6760. [DOI] [PubMed] [Google Scholar]
23.DeVol RT, et al. Nanoscale transforming mineral phases in fresh nacre. J Am Chem Soc. 2015;137:13325–13333. doi: 10.1021/jacs.5b07931. [DOI] [PubMed] [Google Scholar]
24.Pokroy B, et al. Narrowly distributed crystal orientation in biomineral vaterite. Chem Mater. 2015;27:6516–6523. [Google Scholar]
25.Meibom A, et al. Distribution of magnesium in coral skeleton. Geophys Res Lett. 2004;31:1–4. [Google Scholar]
26.Reggi M, et al. Biomineralization in mediterranean corals: The role of the intraskeletal organic matrix. Cryst Growth Des. 2014;14:4310–4320. [Google Scholar]
27.Falini G, Fermani S, Goffredo S. Coral biomineralization: A focus on intra-skeletal organic matrix and calcification. Semin Cell Dev Biol. 2015;46:17–26. doi: 10.1016/j.semcdb.2015.09.005. [DOI] [PubMed] [Google Scholar]
28.Beniash E, Aizenberg J, Addadi L, Weiner S. Amorphous calcium carbonate transforms into calcite during sea urchin larval spicule growth. Proc R Soc Lond B Biol Sci. 1997;264:461–465. [Google Scholar]
29.Politi Y, Arad T, Klein E, Weiner S, Addadi L. Sea urchin spine calcite forms via a transient amorphous calcium carbonate phase. Science. 2004;306:1161–1164. doi: 10.1126/science.1102289. [DOI] [PubMed] [Google Scholar]
30.Weiss IM, Tuross N, Addadi L, Weiner S. Mollusc larval shell formation: Amorphous calcium carbonate is a precursor phase for aragonite. J Exp Zool. 2002;293:478–491. doi: 10.1002/jez.90004. [DOI] [PubMed] [Google Scholar]
31.Politi Y, et al. Transformation mechanism of amorphous calcium carbonate into calcite in the sea urchin larval spicule. Proc Natl Acad Sci USA. 2008;105:17362–17366. doi: 10.1073/pnas.0806604105. [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Gong YUT, et al. Phase transitions in biogenic amorphous calcium carbonate. Proc Natl Acad Sci USA. 2012;109:6088–6093. doi: 10.1073/pnas.1118085109. [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Radha AV, Forbes TZ, Killian CE, Gilbert PUPA, Navrotsky A. Transformation and crystallization energetics of synthetic and biogenic amorphous calcium carbonate. Proc Natl Acad Sci USA. 2010;107:16438–16443. doi: 10.1073/pnas.1009959107. [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Metzler RA, Tribello GA, Parrinello M, Gilbert PUPA. Asprich peptides are occluded in calcite and permanently disorder biomineral crystals. J Am Chem Soc. 2010;132:11585–11591. doi: 10.1021/ja103089r. [DOI] [PubMed] [Google Scholar]
35.Cuif J-P, Dauphin Y, Doucet J, Salome M, Susini J. XANES mapping of organic sulfate in three scleractinian coral skeletons. Geochim Cosmochim Acta. 2003;67:75–83. [Google Scholar]
36.Von Euw S, et al. Biological control of aragonite formation in stony corals. Science. 2017;356:933–938. doi: 10.1126/science.aam6371. [DOI] [PubMed] [Google Scholar]
37.Cartwright JH, Checa AG, Gale JD, Gebauer D, Sainz-Díaz CI. Calcium carbonate polyamorphism and its role in biomineralization: How many amorphous calcium carbonates are there? Angew Chem Int Ed Engl. 2012;51:11960–11970. doi: 10.1002/anie.201203125. [DOI] [PubMed] [Google Scholar]
38.Farhadi‐Khouzani M, Chevrier DM, Zhang P, Hedin N, Gebauer D. Water as the key to proto‐aragonite amorphous CaCO3. Angew Chem Int Ed Engl. 2016;55:8117–8120. doi: 10.1002/anie.201603176. [DOI] [PubMed] [Google Scholar]
39.Killian CE, et al. Mechanism of calcite co-orientation in the sea urchin tooth. J Am Chem Soc. 2009;131:18404–18409. doi: 10.1021/ja907063z. [DOI] [PubMed] [Google Scholar]
40.DeVol RT, et al. Oxygen spectroscopy and polarization-dependent imaging contrast (PIC)-mapping of calcium carbonate minerals and biominerals. J Phys Chem B. 2014;118:8449–8457. doi: 10.1021/jp503700g. [DOI] [PubMed] [Google Scholar]
41.Gilbert PUPA, Young A, Coppersmith SN. Measurement of c-axis angular orientation in calcite (CaCO3) nanocrystals using X-ray absorption spectroscopy. Proc Natl Acad Sci USA. 2011;108:11350–11355. doi: 10.1073/pnas.1107917108. [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Venn A, Tambutté E, Holcomb M, Allemand D, Tambutté S. Live tissue imaging shows reef corals elevate pH under their calcifying tissue relative to seawater. PLoS One. 2011;6:e20013. doi: 10.1371/journal.pone.0020013. [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Clode PL, Marshall AT. Low temperature FESEM of the calcifying interface of a scleractinian coral. Tissue Cell. 2002;34:187–198. doi: 10.1016/s0040-8166(02)00031-9. [DOI] [PubMed] [Google Scholar]
44.Gilis M, et al. Biomineralization in newly settled recruits of the scleractinian coral Pocillopora damicornis. J Morphol. 2014;275:1349–1365. doi: 10.1002/jmor.20307. [DOI] [PubMed] [Google Scholar]
45.Erez J, Braun A. Calcification in hermatypic corals is based on direct seawater supply to the biomineralization site. Geochim Cosmochim Acta. 2007;71(15 Suppl 1):A260. [Google Scholar]
46.Tambutté E, et al. Calcein labelling and electrophysiology: Insights on coral tissue permeability and calcification. Proc Biol Sci. 2012;279:19–27. doi: 10.1098/rspb.2011.0733. [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Khalifa GM, et al. Biomineralization pathways in a foraminifer revealed using a novel correlative cryo-fluorescence-SEM-EDS technique. J Struct Biol. 2016;196:155–163. doi: 10.1016/j.jsb.2016.01.015. [DOI] [PubMed] [Google Scholar]
48.Tambutté E, et al. Observations of the tissue-skeleton interface in the scleractinian coral Stylophora pistillata. Coral Reefs. 2007;26:517–529. [Google Scholar]
49.Schauer KL, et al. A proteinaceous organic matrix regulates carbonate mineral production in the marine teleost intestine. Sci Rep. 2016;6:34494. doi: 10.1038/srep34494. [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Zoccola D, et al. Bicarbonate transporters in corals point towards a key step in the evolution of cnidarian calcification. Sci Rep. 2015;5:9983. doi: 10.1038/srep09983. [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Comeau S, et al. Coral calcifying fluid pH is modulated by seawater carbonate chemistry not solely seawater pH. Proc Biol Sci. 2017;284:20161669. doi: 10.1098/rspb.2016.1669. [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Vidavsky N, Masic A, Schertel A, Weiner S, Addadi L. Mineral-bearing vesicle transport in sea urchin embryos. J Struct Biol. 2015;192:358–365. doi: 10.1016/j.jsb.2015.09.017. [DOI] [PubMed] [Google Scholar]
53.Vidavsky N, et al. Initial stages of calcium uptake and mineral deposition in sea urchin embryos. Proc Natl Acad Sci USA. 2014;111:39–44. doi: 10.1073/pnas.1312833110. [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Yang L, Killian CE, Kunz M, Tamura N, Gilbert PUPA. Biomineral nanoparticles are space-filling. Nanoscale. 2011;3:603–609. doi: 10.1039/c0nr00697a. [DOI] [PubMed] [Google Scholar]
55.Schrag DP, Linsley BK. Paleoclimate. Corals, chemistry, and climate. Science. 2002;296:277–278. doi: 10.1126/science.1071561. [DOI] [PubMed] [Google Scholar]
56.Saenger C, Erez J. A non-traditional stable isotope perspective on coral calcification. In: Goffredo S, Dubinsky Z, editors. The Cnidaria, Past, Present and Future. Springer; Berlin: 2016. pp. 181–205. [Google Scholar]
57.Lobel PS. Kinetic control of skeletal Sr/Ca in a symbiotic coral: Implications for the paleotemperature proxy. Paleoceanography. 2001;16:20–26. [Google Scholar]
58.Hönisch B, et al. The geological record of ocean acidification. Science. 2012;335:1058–1063. doi: 10.1126/science.1208277. [DOI] [PubMed] [Google Scholar]
59.Gagnon AC, Adkins JF, Fernandez DP, Robinson LF. Sr/Ca and Mg/Ca vital effects correlated with skeletal architecture in a scleractinian deep-sea coral and the role of Rayleigh fractionation. Earth Planet Sci Lett. 2007;261:280–295. [Google Scholar]
60.Gaetani GA, Cohen AL. Element partitioning during precipitation of aragonite from seawater: A framework for understanding paleoproxies. Geochim Cosmochim Acta. 2006;70:4617–4634. [Google Scholar]
61.Gagnon AC, Adkins JF, Erez J. Seawater transport during coral biomineralization. Earth Planet Sci Lett. 2012;329:150–161. [Google Scholar]
62.Urey HC, Lowenstam HA, Epstein S, McKinney CR. Measurement of paleotemperatures and temperatures of the upper cretaceous of England, Denmark, and the southeastern United States. Geol Soc Am Bull. 1951;62:399–416. [Google Scholar]
63.Spero HJ, Bijma J, Lea DW, Bemis BE. Effect of seawater carbonate concentration on foraminiferal carbon and oxygen isotopes. Nature. 1997;390:497–500. [Google Scholar]
64.Bertucci A, Tambutté S, Supuran CT, Allemand D, Zoccola D. A new coral carbonic anhydrase in Stylophora pistillata. Mar Biotechnol (NY) 2011;13:992–1002. doi: 10.1007/s10126-011-9363-x. [DOI] [PubMed] [Google Scholar]
65.Venn AA, et al. Imaging intracellular pH in a reef coral and symbiotic anemone. Proc Natl Acad Sci USA. 2009;106:16574–16579. doi: 10.1073/pnas.0902894106. [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Tambutte E, Allemand D, Mueller E, Jaubert J. A compartmental approach to the mechanism of calcification in hermatypic corals. J Exp Biol. 1996;199:1029–1041. doi: 10.1242/jeb.199.5.1029. [DOI] [PubMed] [Google Scholar]
67.Bhattacharya D, et al. Comparative genomics explains the evolutionary success of reef-forming corals. eLife. 2016;5:e13288. doi: 10.7554/eLife.13288. [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Ghosh P, et al. 13 C–18 O bonds in carbonate minerals: A new kind of paleothermometer. Geochim Cosmochim Acta. 2006;70:1439–1456. [Google Scholar]
69.Stephenson AE, et al. Peptides enhance magnesium signature in calcite: Insights into origins of vital effects. Science. 2008;322:724–727. doi: 10.1126/science.1159417. [DOI] [PubMed] [Google Scholar]
70.Stoll HM, Schrag DP. Coccolith Sr/Ca as a new indicator of coccolithophorid calcification and growth rate. Geochem Geophys Geosyst. 2000;1:1–24. [Google Scholar]
71.Lea DW, Mashiotta TA, Spero HJ. Controls on magnesium and strontium uptake in planktonic foraminifera determined by live culturing. Geochim Cosmochim Acta. 1999;63:2369–2379. [Google Scholar]
72.Van Santen R. The Ostwald step rule. J Phys Chem. 1984;88:5768–5769. [Google Scholar]
73.Navrotsky A. Energetic clues to pathways to biomineralization: Precursors, clusters, and nanoparticles. Proc Natl Acad Sci USA. 2004;101:12096–12101. doi: 10.1073/pnas.0404778101. [DOI] [PMC free article] [PubMed] [Google Scholar]
74.Penn RL, Banfield JF. Morphology development and crystal growth in nanocrystalline aggregates under hydrothermal conditions: Insights from titania. Geochim Cosmochim Acta. 1999;63:1549–1557. [Google Scholar]
75.Banfield JF, Welch SA, Zhang H, Ebert TT, Penn RL. Aggregation-based crystal growth and microstructure development in natural iron oxyhydroxide biomineralization products. Science. 2000;289:751–754. doi: 10.1126/science.289.5480.751. [DOI] [PubMed] [Google Scholar]
76.Cölfen H, Antonietti M. Mesocrystals: Inorganic superstructures made by highly parallel crystallization and controlled alignment. Angew Chem Int Ed Engl. 2005;44:5576–5591. doi: 10.1002/anie.200500496. [DOI] [PubMed] [Google Scholar]
77.Cölfen H, Antonietti M. Mesocrystals and Nonclassical Crystallization. Wiley; New York: 2008. [Google Scholar]
78.De’ath G, Fabricius KE, Sweatman H, Puotinen M. The 27-year decline of coral cover on the great barrier reef and its causes. Proc Natl Acad Sci USA. 2012;109:17995–17999. doi: 10.1073/pnas.1208909109. [DOI] [PMC free article] [PubMed] [Google Scholar]
79.Rosenfeld M, Yam R, Shemesh A, Loya Y. Implication of water depth on stable isotope composition and skeletal density banding patterns in a Porites lutea colony: Results from a long-term translocation experiment. Coral Reefs. 2003;22:337–345. [Google Scholar]
80.Hughes TP, et al. Global warming and recurrent mass bleaching of corals. Nature. 2017;543:373–377. doi: 10.1038/nature21707. [DOI] [PubMed] [Google Scholar]
81.Orr JC, et al. Anthropogenic ocean acidification over the twenty-first century and its impact on calcifying organisms. Nature. 2005;437:681–686. doi: 10.1038/nature04095. [DOI] [PubMed] [Google Scholar]
82.Burton EA, Walter LM. Relative precipitation rates of aragonite and Mg calcite from seawater: Temperature or carbonate ion control? Geology. 1987;15:111–114. [Google Scholar]
83.Mucci A, Canuel R, Zhong S. The solubility of calcite and aragonite in sulfate-free seawater and the seeded growth kinetics and composition of the precipitates at 25 C. Chem Geol. 1989;74:309–320. [Google Scholar]
84.Zhong S, Mucci A. Calcite and aragonite precipitation from seawater solutions of various salinities: Precipitation rates and overgrowth compositions. Chem Geol. 1989;78:283–299. [Google Scholar]
85.Gutjahr A, Dabringhaus H, Lacmann R. Studies of the growth and dissolution kinetics of the CaCO3 polymorphs calcite and aragonite I. Growth and dissolution rates in water. J Cryst Growth. 1996;158:296–309. [Google Scholar]
86.Liberman T, Genin A, Loya Y. Effects on growth and reproduction of the coral Stylophora pistillata by the mutualistic damselfish Dascyllus marginatus. Mar Biol. 1995;121:741–746. [Google Scholar]
87.Kotb MM. Growth rates of three reef-building coral species in the northern Red Sea, Egypt. Egypt J Aquat Biol Fish. 2001;5:165–185. [Google Scholar]
88.Shaish L, Abelson A, Rinkevich B. Branch to colony trajectory in a modular organism: Pattern formation in the Indo-Pacific coral Stylophora pistillata. Dev Dyn. 2006;235:2111–2121. doi: 10.1002/dvdy.20861. [DOI] [PubMed] [Google Scholar]
89.Nancollas G, Reddy M. The crystallization of calcium carbonate. II. Calcite growth mechanism. J Colloid Interface Sci. 1971;37:824–830. [Google Scholar]
90.Mucci A, Morse JW. The incorporation of Mg2+ and Sr2+ into calcite overgrowths: Influences of growth rate and solution composition. Geochim Cosmochim Acta. 1983;47:217–233. [Google Scholar]
91.Morse JW, Mucci A, Millero FJ. The solubility of calcite and aragonite in seawater of 35%. salinity at 25 C and atmospheric pressure. Geochim Cosmochim Acta. 1980;44:85–94. [Google Scholar]
92.Millero FJ, Feistel R, Wright DG, McDougall TJ. The composition of standard seawater and the definition of the reference-composition salinity scale. Deep Sea Res Part 1 Oceanogr Res Pap. 2008;55:50–72. [Google Scholar]
93.Bongiorni L, Shafir S, Rinkevich B. Effects of particulate matter released by a fish farm (Eilat, Red Sea) on survival and growth of Stylophora pistillata coral nubbins. Mar Pollut Bull. 2003;46:1120–1124. doi: 10.1016/S0025-326X(03)00240-6. [DOI] [PubMed] [Google Scholar]
94.Mass T, et al. Photoacclimation of Stylophora pistillata to light extremes: Metabolism and calcification. Mar Ecol Prog Ser. 2007;334:93–102. [Google Scholar]
95.Moya A, et al. Study of calcification during a daily cycle of the coral Stylophora pistillata: Implications for ‘light-enhanced calcification’. J Exp Biol. 2006;209:3413–3419. doi: 10.1242/jeb.02382. [DOI] [PubMed] [Google Scholar]
96.Gilbert PUPA. Photoemission spectromicroscopy for the biomineralogist. In: DiMasi E, Gower LB, editors. Biomineralization Sourcebook, Characterization of Biominerals and Biomimetic Materials. CRC; Boca Raton, FL: 2014. pp. 135–151. [Google Scholar]
97.Frazer BH, Gilbert B, Sonderegger BR, De Stasio G. The probing depth of total electron yield in the sub keV range: TEY-XAS and X-PEEM. Surf Sci. 2003;537:161–167. [Google Scholar]
98.De Stasio G, Frazer BH, Gilbert B, Richter KL, Valley JW. Compensation of charging in X-PEEM: A successful test on mineral inclusions in 4.4 Ga old zircon. Ultramicroscopy. 2003;98:57–62. doi: 10.1016/S0304-3991(03)00088-3. [DOI] [PubMed] [Google Scholar]
99.Falini G, Albeck S, Weiner S, Addadi L. Control of aragonite or calcite polymorphism by mollusk shell macromolecules. Science. 1996;271:67–69. [Google Scholar]
100.GG Macros 2017 Available at home.physics.wisc.edu/gilbert/software. Accessed August 10, 2017.
101.Metzler RA, et al. Architecture of columnar nacre, and implications for its formation mechanism. Phys Rev Lett. 2007;98:268102. doi: 10.1103/PhysRevLett.98.268102. [DOI] [PubMed] [Google Scholar]
102.Metzler RA, et al. Polarization-dependent imaging contrast in abalone shells. Phys Rev B. 2008;77:064110. [Google Scholar]
103.Olson IC, Kozdon R, Valley JW, Gilbert PUPA. Mollusk shell nacre ultrastructure correlates with environmental temperature and pressure. J Am Chem Soc. 2012;134:7351–7358. doi: 10.1021/ja210808s. [DOI] [PubMed] [Google Scholar]
104.Olson IC, et al. Crystal nucleation and near-epitaxial growth in nacre. J Struct Biol. 2013;184:454–463. doi: 10.1016/j.jsb.2013.10.002. [DOI] [PubMed] [Google Scholar]
105.Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical Recipes in Fortran. Cambridge Univ Press; Cambridge, UK: 1992. [Google Scholar]
106.Sun CY, et al. Spherulitic growth of coral skeletons and synthetic aragonite: Nature's three-dimensional printing. ACS Nano. 2017;11:6612–6622. doi: 10.1021/acsnano.7b00127. [DOI] [PubMed] [Google Scholar]
107.Walker JM, Marzec B, Nudelman F. Solid–state transformation of amorphous calcium carbonate to aragonite captured by cryoTEM. Angew Chem Int Ed Engl. July 25, 2017 doi: 10.1002/anie.201703158. [DOI] [PMC free article] [PubMed] [Google Scholar]

Amorphous calcium carbonate particles form coral skeletons

Significance

Abstract

Results and Discussion

Fig. 1.

Fig. 4.

Fig. S1.

Fig. S2.