During 1970s, coral skeleton oxygen isotope composition (δ18O) was regarded as the isotopic thermometer following thermodynamic rules. Recently, coral aragonite oxygen isotopic fractionation could appear to be controlled by biology, its rate being accelerated by an enzyme (carbonic anhydrase or CA). Such a new concept results of an original approach involving coral culture in controlled conditions. Environmental factors, temperature and also light have been tested on macrosize scale samples (some mg), and δ18O revealed vital effects, anomalies compared with chemical and isotopic equilibrium, related to metabolic activity. δ18O analyses at microsize scale (some μm), using ion microprobe, could display the mechanism of crystallisation, δ18O fractionation responding to biological kinetic effects. The understanding of coral aragonite δ18O is the absolute prerequisite to develop the first model of a coral proxy.
- oxygen isotopic fractionation
- coral skeleton
- controlled conditions
- organic matrix
Coral colonies built the most important bioconstruction made of calcium carbonate (CaCO3) of the world, with a calcification of about 2–6 kgCaCO3 m−2 year−1 covering an area of about 284,300 km2 . This construction built from Jurassic results from the work of multiple small colonial organisms. The reefs, the biotic mound structure essentially made of corals as the Great Barrier Reef in Australia, are of major importance for marine ecosystems and biodiversity because they are the most productive and they host almost a third of all world fishes.
Corals are marine animals forming an aragonite (a polymorph of CaCO3) skeleton. They are developed in two distinct ecosystems, essentially zooxanthellate corals or symbiotic ones living in shallow water and solitary colonies or integrated in elaborate reef framework in deeper depth than 50 until 2000 m. More than 793 coral species are spread over marine tropical zone . Branched corals
Epstein [4, 5] demonstrated that skeletal carbonates of marine shells display similar oxygen isotopic composition (δ18OCaCO3) relationship versus SST than inorganic calcium carbonate (CaCO3) deposited from seawater at the same temperature, following thermodynamic laws.
The relationship was expressed as:
However, after preliminary studies , Weber and Woodhead deduced, despite apparent isotopic disequilibrium between coral skeletal carbonate and ambient seawater, that δ18Ocoral was temperature dependent. To support this assumption , the authors conducted isotopic analyses of coral skeleton collected over wide range of temperatures. This data series still constitutes the most exhaustive oxygen isotopic database existing for corals. Weber and Woodhead concluded that the calibration between annual δ18O and annual SST differed following each coral genus, and the isotopic disequilibrium was attributed to vital effect, anomalies compared with chemical and isotopic equilibrium, related to metabolic activity. Several models of mineralisation were proposed to explain the geochemical specificities of coral skeletons based on kinetic fractionation [7, 8, 9] disturbed by “vital effects”. Other models, based on precipitation efficiency  or Rayleigh fractionation [11, 12, 13], were suggested.
We developed drastically different approach considering that corals are animals living in symbiosis with algae, building aragonitic skeleton intimately related to biological activity. In collaboration with biologists from CSM (Centre Scientific de Monaco), Stéphanie Reynaud and Christine Ferrier-Pagès, and Claire Rollion-Bard geochemist from IPGP (Institut Physique du Globe de Paris), we developed an innovative strategy on cultured
Our demonstration is structured as followed: first, the main coral features are highlighted; second, we describe coral culture proceeding; third, temperature and light test results are presented at microscopic size scale; and finally, we display the stable oxygen isotopic ratio δ18O as indicator of deposit mechanism.
2. Main coral features
Shallow corals, because they are leaving in symbiosis with micro algae need light to benefit from photosynthetic activity.
2.1 Notions of coral morphology and biological activity
Coral skeleton is extracellular, located at the base of coral tissue, constituted of similar units, the polyps. Each polyp looks like a bag made by two layers of cells (Figure 1). Polyps are linked together by the coenosarc. Most of zooxanthellae are located within an internal layer (Figure 1).
Biological activity might be quantified. Photosynthesis and respiration were measured using the respirometry technique, which measured the changes in oxygen concentration at different light levels. Rates of net photosynthesis and respiration were estimated using a linear regression of O2 against time . By using two light intensities, at three temperatures (Figure 2), it appeared that increasing temperature enhanced photosynthetic activity, the effect arising with light intensity (Figure 2a) . Additionally, coral growth rates might be estimated. Corals were weighed regularly using the buoyant weight technique and the surface expansion of the new skeleton formed was estimated . It was generally accepted that calcification was light-enhanced (LEC or light enhanced calcification) during the day . Zooxanthellae density and pigment concentration were determined under the light microscope using a counting chamber . Coral symbiont distribution was not homogeneous on the skeleton and it differed following different coral genera, different depths in the fields. For example, tips of coral branches or other exposed surfaces were sun-adapted while most of the lowest parts were shade-adapted . Iluz and Dubinsky  listed all the strategies developed by coral to optimise the light impinging on the zooxanthellae.
2.2 Microstructures of coral skeletal
It was admitted that the coral skeleton such as coral
These crystalline elements were differently distributed according to morphology [19, 20, 21]. Each microstructure is preferentially present in some morphological parts, which were more or less developed following the genus . However, we know that they are composed by identical microstructures and only differ by their relative amounts.
2.3 Consequences on δ18O
The oxygen isotopic composition of coral skeleton was measured on conventional spectrometer at mm size scale and might be also measured at micrometre scale by using ion microprobe.
Coral skeleton δ18O was impacted by biology, which was essentially responsible of the vital effect. Most of the models proposed by geochemists neglect biology effects on δ18O, isotopic fractionation only depending on seasurface temperature and isotopic composition, δ18Oseawater following Eq. (1). All climate reconstructions are derived according to this rule, including the estimate of both temperature and salinity based on the use of paired δ18O and Sr/Ca measured on the same sample [23, 24, 25]. Salinity values deduced by this method are systematically misleading, d18O and Sr/Ca SST calibrations being based on classical thermodynamics. Whereas consequences of temperature and light on coral growth rates is well known by biologists , light effect is ignored by geochemists because the demonstration of this influence cannot be established from field data and needs culture tests.
δ18O differed following microstructures at microscopic size scale (Figure 3j and k) . This was confirmed later on [28, 29]. COC δ18O was lower, while fibre δ18O was higher and variable between equilibrium and COC value (Figure 3j and k) .
3. Calibrations of annual and monthly δ18O
3.1 Weber and Woodhead (1972) data set and annual calibrations revisited
(a) and (b) being constants (instead of A and B in ).
with R2 = 0.83, N = 22, p < 0.001, only significant over the SST range from 24 to 30°C .
However, the correlation linking δ18O directly to temperature showed a higher coefficient :
A similar procedure was conducted for
with R2 = 0.87, N = 24, p < 0.001, significant over the temperature range from 21 to 30°C (Figure 4e).
the correlation linking δ18O directly to temperature showed a higher coefficient :
with R2 = 0.98, N = 27, p < 0.001, significant over the temperature range from 21 to 30°C.
Slopes (a) shown by
After introducing δ18Oseawater into dataset , for
In the calibrations depending only on temperature, temperature might act first, according to thermodynamic law [5, 15] and second, through the photosynthetic process  (Figure 2a), which was enhanced by a temperature increase. Therefore, an increase in temperature induced a decrease in δ18O following the first process while the second mechanism caused a rise in δ18O, confusing the global isotopic effect. Temperature influences δ18O twice, explaining that temperature is the main factor, which does not exclude the role of δ18Oseawater.
Calibrations taking into account δ18Oseawater exhibited a slope value close to isotopic equilibrium of inorganic aragonite with water, suggesting that under quasi-uniform light, the isotopic offset of coral δ18O is constant, regardless of temperature (Figure 4c). Eqs. (3) and (4) confirm that, to a lesser degree than temperature, δ18Oseawater may be included in a calibration.
with R2 = 0.95, N = 37 and p < 0.001. (a) corresponds to a disequilibrium indicator compared to −0.19, the slope value derived from the theoretical δ18O at equilibrium . Such a relationship was not hazardous, but reflected inherent features of annual δ18O-annual SST calibrations. Linear calibrations determined from single genus deduced from figures or table of , showed strong correlation coefficients: R2 = 0.99 (Figure 4e).
This suggests that the δ18O SST dependence is based on a unique rationale according to taxonomy, in turn inherent to the coral skeleton.
Dataset  revealed unique relationship between annual δ18O-annual temperature calibrations of each genus, because coral taxonomy is based on morphology. Land et al.  stressed the high δ18O variability following the longitudinal section on the calices of some species or the septa dentations of another one, inducing that according to coral morphology, some skeleton portions might be more or less developed, implying a large isotopic variability.
We underlined the relationship existing between the annual δ18O-annual SST calibration constants. However, identical feature was highlighted for annual Sr/Ca-annual temperature calibrations [35, 36, 37]. The link existing between δ18O and Sr/Ca is not straightforward, oxygen being a component of CaCO3 and Sr/Ca an impurity included in the skeleton. However, it is possible to recognise common δ18O and Sr/Ca behaviour relative to their microstructure distribution in the coral and the concept of taxonomy.
Coral skeleton presents composite mineral microstructures: centres of calcification (COC) and fibres, embedded in a few organic matter as a network , differently distributed according to morphology [19, 20, 21]. Latter authors show that COC and fibres are essentially present in morphological parts, more or less developed following the genus. On the one hand, δ18O signature differs according to the microstructures [19, 20, 21], COC δ18O being lower than fibre δ18O  (Figure 3j and k). On the other hand, Sr/Ca ratios measured on COCs are higher than those of fibres . Cohen et al.  examined synchronously deposited microstructures on
3.2 Identification of microstructures and their isotopic signatures using microsensor
Several small colonies of
The new skeleton, formed under unique controlled condition, was grown on the glass slide, and sampled for the calibration of the growth units (Figure 3j). COC- and fibre-enriched zones were identified using SEM . To characterise separately the isotopic signature of fibres and COC, analyses were focused on the microstructures earlier identified on the newly formed skeleton on two zones (Figure 3a–c; Figure 3j). We then focused our measurements around the theca of the newly formed skeleton (Figure 3k) where Gladfelter  recognised large amounts of “fusiform crystals” (Figure 3e). The sampling of the septae aimed at confirming the presence of both COC and fibres as they have been identified from SEM observations .
The present study, confirmed that there was a strong relationship between isotopic value, crystal shape and skeleton morphology . Crystals called “fusiform” by Gladfelter , according to their shape, show the same isotopic values as COC. We distinguished in septa both COC and fibres (Figure 3k). This confirmed microscopic observations of septa  showing discontinuous COC surrounded by fibres.
Isotopic fractionation was likely of kinetic origin, the rate changing according to microstructure. Skeleton microstructures δ18O shed in light how chemical and/or physical processes might be adapted by biology to form crystals characterised by specific shapes and distributed following a hierarchical arrangement. The present study demonstrated that the presence of organic molecules (the organic matrix located at the interface tissue-mineral (Figure 1) had the capability to control the mineral deposition mechanism. Probably, the influence of external factors should be superimposed on the chemical signature of coral biomineral genetically determined).
3.3 Monthly calibrations on coral
The preliminary step of climatic reconstruction using
At millimetre size scale, it was also possible to highlight the strongly significant linear relationship between the constants of seasonal δ18O-seasonal temperature calibrations and to relate behaviour of the constants of the seasonal-δ18O- and Sr/Ca-seasonal temperature calibrations to the presence of two crystallographic units. Following DeLong et al. , fibres insuring the thickening of a colony should be preferentially deposited during a less active photosynthesis, whereas COC insuring axial growth should be formed during high photosynthetic activity. Juillet-Leclerc and Reynaud agreed, however, they demonstrated that growth mode was not so simple .
In order to test seasonal δ18O-seasonal temperature calibration variability including the seasonal light effect, calculated for several coral cores collected on a given site, at different temperature ranges, we considered studies conducted on several
We assumed that calibrations measured on different coral colonies grown at a given site (New Caledonia, Clipperton or Indonesia) differed according to various light sensitivities due to depth or light incidence or acclimation because seasonality strongly affected light variations, and was likely different following site location (Figure 6a). However, calibration constants calculated from monthly data for
As seasonal δ18O-seasonal temperature calibrations presented similar behaviours, even in different sites characterised by distinct δ18Oseawater , they did not reflect classical thermodynamics.
4. Coral cultures simulating different environmental conditions
4.1 Test of temperature
New technique of culture was developed to calculate δ18O-temperature calibration for
The calibration given by the experiment (Figure 5c) might be expressed as:
with N = 5 and R2 = 0.96.
As δ18Oseawater = 1.29 vs. SMOW = 1.02 vs. PDB was constant, the calibration might be expressed as:
4.2 Test of light intensity
Tips from 24 branches were sampled from a single parent colony of
In the first time , the averaged results showed that daily calcification, net photosynthesis significantly increased with increasing light and skeletal δ18Ocoral were more negative under low light than high light, −4.2 versus −3.8.
Another interpretation, considering each colony was later published . Following the evolution, δ13Ccoral increasing or decreasing, two populations appeared: one responding to light with increasing net photosynthesis associated to low linear extension and the other characterised by poor net photosynthesis associated to high linear extension (Figure 5d).
We suggested that different behaviours were due to different zooxanthellae amounts contained by colonies.
4.3 Factorial design of three temperatures and two light intensities
Calibrations calculated from the mean δ18O values for each temperature regime were consistent with those previously published (Figure 2c) [49, 50]. δ18O
By comparing our results with other culture experiments [51, 52, 53], differences appeared between various δ18O-temperature (°C) calibrations regarding both the slopes and the intercepts with the temperature scale. We suggested that this could be due to inter-species or inter-colony δ18O differences. Even two calibrations obtained on cultured
Mean δ18O values calculated for each temperature did not vary with light, which contradicted observations made of mean physiological parameters, in contrast with other proxies (δ13C, Sr/Ca and Mg/Ca). By changing the light intensity from low to moderate, Juillet-Leclerc and Reynaud  recorded a δ18O increase associated with skeletal infilling following a kinetic process. The present experiment, conducted under high light intensities, did not show a similar behaviour. We suggested that temperature and light effects on isotopic composition were competing. The results of the present experiment indicated that, under the chosen conditions, the temperature effect was more important than the light effect. This was illustrated by the weak discrepancy in the mean δ18O recorded at 28°C.
Previous culture experiments have been conducted to test the temperature effect on δ18O [38, 51, 52, 53, 54]. Due to the sensitivity of the photosynthesis to temperature, δ18O-temperature calibrations will always include temperature-dependent photosynthetic changes, enabling the vital effect due to temperature only to be deconvolved from the total signature. Therefore, all calibrations, even established on a single coral head or from cultured nubbins, are impacted by photosynthetic activity linked to zooxanthellae density. The universal calibration does not exist.
5. A new paradigm for δ18O in coral skeleton oxygen isotope fractionation response to biological kinetic effects
During the last experiment, we failed understanding the light effect on δ18O. We kept one colony cultured in each of six light and temperature conditions previously discussed. Knowing that standard error obtained in the first step of the experiment for six samples cultured in the same condition was between 0.02 and 0.12, we consider that values measured on one colony by using SIMS were representative for each environmental condition .
We discussed our results after listing all biological and biological advances, such as (i) conclusions derived from inorganic CaCO3 precipitation disturbed by biology, biased by non-realistic models [8, 55, 56]; (ii) the potential role of the calicoblastic layer composed of proteins, sugars and water [57, 58] and (iii) the role of carbonic anhydrase (CA), ubiquitous enzymes known to act as catalysts for the interconversion of CO2 and [59, 60].
The random SIMS measurements were made exclusively on the newly formed skeleton coenosteum (the skeleton portion separating corallites), avoiding newly formed corallites and spines . Samples were distributed along the growth axis from the initial branch to the rim of the expanded tissue (Figure 7a).
The δ18O SIMS measurements are displayed as histograms, with bin width of histograms, 0.25‰, depending on the precision of the measurements between 0.09 and 0.32‰ (1σ) (Figure 7b).
5.2 Isotopic results
At 22°C, although the mean values were identical within the analytical error, light had a significant effect on the δ18O distribution. Under LL, the single high bar was surrounded by values spread over 2‰ (Figure 7b). By contrast, under HL, values spread over 4‰, δ18O distribution was bimodal, two high bars were observed, one bar centred on −0.01‰, followed by decreasing values that exceeded the expected δ18O values for aragonite precipitated in oxygen isotope equilibrium in water.
At 25°C, the bimodal δ18O distribution in the two samples exhibited two high bars, with the more depleted in 18O peak being the same in the two light conditions (Figure 7b).
At 28°C, δ18O distribution was bimodal, the isotopic amplitude being slightly higher under LL than under HL (Figure 7b).
Assuming that two high bars observed in histograms were significant, we used Ashman’s D test  to strengthen the bimodality of the LL-25°C, LL-28°C, HL-25°C and HL-28°C.
The histograms showed bimodal distributions (except for the colony grown in LL-22°C) caused by distinct kinetic processes. All the colonies were submitted to the diurnal cycle of 12 h light and 12 h dark necessary to grow healthy coral. Therefore, we assumed that, for the high isotopic bar corresponding to the values more depleted in 18O (Figure 2g), corresponding to the highest kinetic fractionation, only depending on temperature is associated with nighttime. By contrast, for the other high isotopic bar corresponding to the values less depleted in 18O, in turn to the weakest kinetic process, depending both on temperature and light could be associated with daytime calcification (Figure 7b and c).
Therefore, assuming that the dual high δ18O bars exhibited in the histograms (Figure 7b) were related to the Ca2+-pump activity, this could modify the internal pH [62, 63]. If the pH is lower in the dark than in the light [62, 63], in line with McCrea’s  calculations, illustrated by Adkins et al. , high δ18O bars less depleted in 18O should be associated with nighttime calcification, and high δ18O bars more depleted in 18O should correspond to daytime conditions (Figure 7b and c).
However, our experiment demonstrated an opposite distribution: High δ18O bars less depleted in 18O were identified as daytime skeleton deposits, and those more depleted in 18O were identified as nighttime deposits. Therefore, we concluded that the two distinct high δ18O bars could not be caused by internal pH diurnal variations. In contrary to the common assumption , we demonstrated that mineralisation is not controlled by classical thermodynamic rules, that is, pH, but rather should obey biological kinetic effects, following a mechanism that remains to be identified.
Now we need to identify a mechanism that allowed daytime and nighttime mineralisation to be distinguished, knowing that the second assumption given by Moya et al. , a modification of the biochemical compounds of the organic matrix, remained to be examined.
5.4 Influence of organic matter on crystallisation
It was supposed that the photosynthetic supply of precursors might modify the biochemical composition of an organic matrix , which necessitated an internal rearrangement related to the secretion of specific proteins defending the observed lag. The formation of the organic matrix, controlled by the calicoblastic cells, appeared to be a prerequisite for crystallisation [58, 59, 60, 61, 62, 63, 64, 65, 66, 67] Recently, 36 proteins were extracted from the skeletal organic matrix (SOM) embedded within aragonite crystals, constituting a bio mineralisation toolkit including at least two Carbonate Anhydrase  accelerating mineralisation. From this toolkit, four unique proteins, coral acid-rich proteins (CARP), catalysed the precipitation of CaCO3 in vitro . Moreover, some proteins appeared to be differentially expressed between day and night [68, 69]. Therefore, the two different proteins caused different kinetic fractionation processes, inducing during the night higher kinetic isotope fractionation than during the day. We note that calcification rate and isotope fractionation kinetics were drastically different concepts.
Results derived from our last geochemical experiment should drove to responses also addressed by biological study. The fact that classical geochemistry rule, such as pH, could not explain isotopic behaviour led to look for another assumption. Therefore, it highlighted that coral mineralisation could be controlled by proteins secreted by organic matrix. This evidence is now well admitted, supported by multiple biological studies [68, 69].
6. Simple models
6.1 Model according to microstructure distribution
As early as 1982, Gladfelter  assumed that linear extension and infilling were two independent growth rates, an assumption supported by Juillet-Leclerc and Reynaud . The authors demonstrated that each growth rate was related to preferential deposition of microstructures, COCs ensuring linear extension and fibres, infilling.
Furthermore, geochemical investigations revealed that crystal isotopic signatures differed [27, 28, 29, 48]. COC formation should be related to temperature  and fibre deposit depends on both temperature and light . Therefore, temperature and light changes interplayed to determine skeletal isotopic composition.
Sampling conducted as it was described by DeLong et al.  included both COCs and fibres. Changes of relative amounts of microstructure as illustrated by X-rays and their respective δ18O were determined by their mechanisms of formation, unknown so far . Following isotopic laws, the combination of calcification processes and isotopic fractionation could be expressed as:
where xCOC and xfibre are the relative amounts of the crystal microstructures, with xCOC + xfibre = 1, and δ18OCOC and δ18Ofibre are their isotopic signatures depending on temperature and temperature and light, respectively. This expression is likely to be simplistic but closer to the truth than the thermodynamic formula. Temperature is the prominent factor because included both in the crystal amounts and the isotopic signatures. SSTintersection, the corresponding δ18Ointersection, should be related to morphology . When using Eq. (10), the intersection of calibration should be obtained when δ18Ointersection = (0.50 × δ18OCOC) + (0.50 × δ18Ofibre) or at SSTintersection, δ18Ointersection = (δ18OCOC + δ18Ofibre)/2. As long as temperature does not reach SSTintersection, more fibres are formed in the coral skeleton and when temperature exceeds SSTintersection, COC are progressively prevailing.
6.2 Model according to environmental parameters
In Pacific Ocean, local zones may be characterised by seasonal and/or interannual environmental parameter amplitude, as ΔSST. By this way, we are able to identify El Niño-Southern Oscillation (ENSO) occurrence , over past time.
ΔSST<2°C, seasonal conditions occurring in Tarawa atoll, in Galapagos or in Fiji , δ18O seasonal variability mimics SSS variability. In Fiji, δ18O is correlated to seasonal precipitation . In other sites, δ18O variability may indicate oceanic advection. Such events are directly related to El Niño.
If seasonal δ18O is recorded over several decades, interannual variability may be isolated. By removing the seasonal cycle and applying a 13-month running mean filter from monthly δ18O, interannual isotopic variability may be regarded as temperature, the greatest fluctuations revealing El Niño-Southern Oscillation (ENSO) occurrence , or the global warming over the twentieth and twenty-first century.
When ΔSST ≥ 5°C, as it is occurring off South Korea coast , temperature and δ18Oseawater are both involved in coral skeleton δ18O variability. Environmental parameters are difficult to separate. If δ18O shows strong decrease associated to great SST drops, it may be caused by the occurrence of La Niña (characterised by colder SST than the normal conditions) .
When 2°C ≤ ΔSST ≤ 4°C as it is recorded in the central tropical Pacific as in Palmyra , δ18O snapshots focused on crucial past periods demonstrate that a 2- to 7-yr bandpassed record (the lower-frequency of ENSO) , the interannual isotopic signal highlights El Niño and La Niña occurrences.
Coral skeletal δ18O is well-suited tool to shed in light climatic events before and after industrial era, to predict future events in the next decades .
Coral skeleton δ18O does not obey to classical thermodynamics but rather reflects aragonite microstructure distribution. We demonstrated that oxygen isotopic fractionation is essentially temperature dependent, due to two temperature effects, one following thermodynamic law, decreasing δ18O when temperature increases, and second, temperature acting through photosynthetic process, increasing δ18O when temperature increases. Consequently, when temperature changes, δ18O is affected in opposite senses, confusing the global isotopic effect.
δ18O measured at millimetre size scale on coral colonies cultured in controlled conditions under varying temperatures and/or light intensities, allows highlighting biologic and isotopic changes associated to environmental factors, acting as vital effect. Measured at microscopic size, δ18O reveals mineralisation processes. By using the last method coupled with biologic evidences, the role of proteins and enzymes, secreted by organic matrix at the interface tissue mineral is demonstrated, showing the potential biologic control on meralisation.