We analyzed deep-sea scleractinian coral skeletons (n=63) and bamboo coral skeletons (n=13) collected from the Labrador Sea (RRS Discovery DY081), the Rockall Bank in the Northeast Atlantic (RRS James Cook JC136), the tropical North Atlantic (RRS James Cook JC094), the southwest Atlantic (collection from National Historical Museum, Brazil), the Drake Passage region of the Southern Ocean (R/V Nathaniel B. Palmer 1103), the northeast Pacific (R/V Western Flyer in the year 2004 and 2007), the Marquesas Islands in the central equatorial Pacific (R/V Alis MUSORSTOM 9; collection from National Museum of Natural History, France), and the Galápagos Archipelago (R/V Melville MV1007; Alucia AL1508; E/V Nautilus NA064) ( Figure 2 ). Sample depths ranged from 60 to 2800 m. The majority of deep-sea corals in this study were collected alive with organic tissue attached using a Remotely Operated Vehicle (ROV). If live samples were not available, selected sub-fossil specimens preserved in pristine condition with radiometric ages < 50 years were used (Galápagos samples, n=3; Northeast Atlantic samples, n=4; Equatorial Atlantic samples, n=6), which can be treated as “modern” corals (dating method taken from Chen et al., 2015). The ages of the California margin’s bamboo corals are adapted from radiocarbon analyses (Frenkel et al., 2017). All the corals were paired with available hydrographic data from seawater observational data (see below). Coral samples in this study are subgrouped by location, and the dominant water masses are listed in Table 1 .

We grouped DSC by taxonomic order and their distinct mineralogy for further comparison. First, scleractinian coral samples include solitary genera Balanophyllia, Caryophyllia, Desmophyllum, and the colonial genus Enallopsammia. The skeletal mineralogy of these scleractinian corals is aragonite. Second, bamboo corals belong to the Isidella and Keratoisis subfamilies with high-Mg calcitic mineralogy. The full coral sample metadata are included in Supplementary Table 1 .

Concentrations of seawater iodate and iodide data were taken from published sources (Wong and Brewer, 1977; Elderfield and Truesdale, 1980; Bluhm et al., 2011; Cutter et al., 2018; Moriyasu et al., 2020; Rapp et al., 2020). Station sites were selected in the closest proximity to the coral site, and the representative value was interpolated from the two closest available measurement depths. Measured seawater iodate/iodide data for the modern ocean, especially depth profiles, remain limited. As a result, some locations and deeper depths lack paired seawater values (e.g., Vayda Seamount in the Equatorial Atlantic). The time-series seawater iodine records have supported the use of single-point sources of seawater data (Campos et al., 1996). The largest published analytical precision for seawater data was 1% (RSD) (Rue et al., 1997). The uncertainty for each seawater iodate/iodide value was estimated from vertical variation between the three most proximal measurements at each coral location (± 2 SD).

The paired dissolved oxygen and hydrographic data (temperature, salinity, [CO₃ ^2-], pH, concentrations of nitrate, phosphate, and silicate) for coral samples were taken from sensor-mounted CTD/ROV measurements or extracted from the GLODAP2 bottle database (Lauvset et al., 2022). The uncertainty on dissolved oxygen estimates is based on the three most proximal measurements. The oxygen content of the Galápagos site is from the nearest published site (Salinas-de-León et al., 2020). These measurements were all within 300 km and 100 m water depth of coral sampling sites. The seawater [CO₃ ^2-] was obtained using the CO2sys.xls program (Pelletier et al., 2007). The input parameters were alkalinity and dissolved inorganic carbon from the GLODAP2 bottle database with dissociation constants taken from Dickson and Millero (1987). All paired seawater data are reported in Supplementary Table 1 .

Coral samples were treated in bleaching solution (30% v/v sodium hypochlorite solution (reagent grade, Fisher Scientific) in water; ~5% active Cl) after collection and air-dried before physical scraping to remove attached organic matter. A diamond-coated rotary blade was used to cut pieces of each coral for analysis. We applied coral sampling criteria as described by Stewart et al. (2020): Only coral theca walls were sampled to avoid microstructural bias, such as from highly contrasting trace metal chemistry in centers of calcification (Gagnon et al., 2007). Coral fragments (~1 g) were then finely crushed using a pestle and mortar to provide a homogeneous powder. Most coral samples were prepared at the University of Bristol, and the bamboo coral samples from the California margin were prepared at the University of California, Davis.

The I/Ca coral sample preparations were modified from Lu et al. (2022). Approximately 5 mg of the coral powder was taken for each sample. The cleaning procedure only involved MQ water cleaning for potential clay contamination. This cleaning step is critical in I/Ca measurements of foraminifera (Winkelbauer et al., 2021) and we included it for completeness; however, we note for coral samples clay contamination is generally negligible. Neither oxidative nor reductive cleaning steps were applied to best preserve volatile iodine from loss during sample preparation. Reductive cleaning was considered unnecessary as Fe-Mn coatings contain a limited iodine (Zhou et al., 2014). Organic matter in DSCs was removed in previous steps via physical scraping and treatment in dilute sodium hypochlorite solution.

For most samples, Mg/Ca measurements were conducted alongside iodine measurements. However, for those samples that required separate analysis due to the use of a different instrument (see below), trace element analysis (e.g., Mg/Ca, Mn/Ca) was performed using separate sample aliquots. Those samples were prepared following methods by Barker et al. (2003) and Stewart et al. (2020). Briefly, samples (~5 mg) underwent reductive cleaning (warm bath in 80˚C, 3.5% hydrous hydrazine buffered in citric acid/ammonia), oxidative cleaning (warm bath in 80˚C, 1% H₂O₂ buffered in ammonia), and weak acid polish (0.0005 M HNO₃). The samples were then dissolved in 0.5 M HNO₃, with a final diluted aliquot containing 4 mM [Ca] taken for analysis. The Mg/Ca of the California Margin bamboo corals were taken from Geyman et al. (2019).

To prepare the samples for iodine measurement, cleaned coral samples were dissolved in 3% HNO₃ buffered by 2% tetramethylammonium hydroxide (TMAH) to stabilize the iodine. Dissolved samples were then stored at 4°C. Iodine concentrations were measured the next day in diluted solutions containing ~50 ppm Ca, 5 ppb internal standards (Sc, In, Cs), 0.5% TMAH, and 0.5% nitric acid. The pH of the solution was controlled (pH = ~2) to minimize iodine volatilization (Cook et al., 2022). DSC I/Ca analyses were carried out on a Thermo Scientific iCAP Q reverse quadrupole inductively coupled plasma mass spectrometer (ICP-MS) at Woods Hole Oceanographic Institution and a Thermo Scientific iCAP triple quadrupole (TQ) ICP-MS at Michigan State University, employing similar techniques. For the latter, Ca, I, and Mg were each measured in both single-quadrupole and triple-quadrupole modes. Freshly made iodine calibration solutions (iodide solution, I.V. Labs, Inc., with 50 ppm [Ca]), calcium calibration solutions standards and blank solutions containing 0.5% TMAH and 0.5% nitric acid were analysed routinely to monitor contributions from processing blanks and instrument memory, which were all matrix-matched with all the samples. ICP-MS standards (Cs, In, Sc) were added to each sample and calibration solution as internal standards and used to normalize the target analytes. A calibration line was produced every 5 to 8 blocks to correct instrumental drift, and data were corrected using the slope of the drift monitors over time. Reference material in-house standard (Porites coral; aragonite) and international standard JCp-1 (Porites coral; aragonite) were measured after every 5 to 8 samples.

In-house coral standard (Porites coral; aragonite) showed consistency within the run, and the I/Ca value yielded 3.58 ± 0.36 µmol/mol (1 σ, n = 24). To compare data between labs, the I/Ca value of JCp-1 for all batches measured at WHOI is 3.91 ± 0.36 µmol/mol (1 σ, n = 34), and at Michigan State University is 4.09 ± 0.34 µmol/mol (1 σ, n = 7). Both results are within the uncertainty of published I/Ca values, where 3.70 ± 0.27 µmol/mol (1 σ, n = 2280) in Lu et al. (2020c) and 4.12 ± 0.26 µmol/mol (1 σ, n = 41) in Zhou et al. (2022). These standard data are shown in Supplementary Table 3 .

For I/Ca data generated using the iCAP RQ ICP-MS, paired Mg/Ca data were separately measured on a Thermo Scientific Element2 ICP-MS at the University of Bristol. An aliquot of the dissolved sample (separate powder from I/Ca data) was analysed using well-characterized, matrix-matched, synthetic standard solutions to yield Mg/Ca ratios. Both NIST RM 8301-coral (simulated coral solution) and JCp-1 (uncleaned) carbonate reference materials were measured repeatedly and yielded a long-term analytical precision of < ± 2% (1 σ).

Incorporating iodine species in calcium carbonate strongly depends on the mineral structure. Previous studies determined the site location of iodine within carbonates. In contrast to most of the major/trace cations that exchange with Ca²⁺ in the carbonate structure, negatively charged iodine (as an iodate ion) has been confirmed to substitute for the carbonate ion in calcium carbonate (Podder et al., 2017). Results from computational simulations show that the iodate substitution into calcite involves a lower energy penalty than substitution into aragonite, meaning that iodate is preferentially partitioned into calcite over aragonite (Feng and Redfern, 2018). While incorporating iodate ion, the calcite lattice of the iodine-bearing structure remains the same (rhombohedral), but for iodine-bearing aragonite, the original orthorhombic cell reduces in symmetry to monoclinic. This leads to a large partitioning difference between calcite to aragonite.

Surprisingly, our results show that the iodine concentration is higher in aragonitic scleractinian corals (mean I/Ca = 8.5 µmol/mol) compared to calcitic bamboo corals (mean I/Ca = 2.4 µmol/mol; Figure 3 ). This contradicts the results from computational simulations that predict a higher iodine relative abundance in calcite than in aragonite (Feng and Redfern, 2018). We emphasize that this comparison is done between I/Ca in scleractinian corals and I/(Ca+Mg) in bamboo corals, but since the iodate is substituting for the carbonate ion, the effect of higher Mg content in bamboo corals is minimal ( Supplementary Figure 2 ). Additionally, the computational simulations were based on a concentrated iodate system with I/Ca ≈40 mmol/mol, almost 5,000 times higher than mean scleractinian DSC I/Ca. Despite these substantial concentration differences, the results of iodate incorporation from our DSCs do not align with the enthalpic penalties and local distortion effects predicted by theoretical simulations. Previous studies noted inconsistencies between thermodynamic calculations and synthesis experiments in iodine-bearing carbonate systems. For example, if thermodynamics are the only influence of iodate incorporation, increased temperature is expected to favor iodate incorporation due to the entropic advantage. However, the synthetic calcite experiment showed lower partition coefficients for iodate at higher temperatures (Zhou et al., 2014). In addition, the energetic penalty for iodate incorporation is less significant in vaterite than in calcite (Feng and Redfern, 2018), but synthesis experiments suggest that iodine is preferentially incorporated into calcite over vaterite (Podder et al., 2017). Moreover, simulation studies in calcite suggest a mechanism of surface segregation. From an entropy point of view, iodate could be more prone to substitute for CO₃ ^2- in aragonite than calcite, as aragonite is more soluble than calcite (Mucci, 1983). These discrepancies, however, are largely unexplained and potential mechanisms remain a subject of speculation. These mechanisms include temperature-dependence of iodine speciation in aqueous solutions (Burns et al., 1990), and/or kinetic effects associated with carbonate dissolution and reprecipitation reactions.

Biological or ‘vital’ effects are another plausible control on iodine partitioning in DSCs. During coral calcification, scleractinian coral increases the Ca concentration and pH of their calcifying fluid through the activity of the Ca/proton-pumps (Cohen and McConnaughey, 2003). This leads to a higher carbonate saturation, which favors aragonite precipitation. By contrast, bamboo corals experience a more complex anion exchange (bicarbonate), which includes a combination of Ca/proton pumping and bicarbonate active transport, the latter of which has been proposed for sulfate incorporation (Flöter et al., 2022). Bicarbonate active transport delivers bicarbonate to the calcifying fluid to keep the calcite saturation at a high level and increases carbonate ion concentration by shifting the calcifying fluid pH. This process leads to a decrease in the calcifying fluid sulphur:carbonate ratio and thus a decrease in bamboo coral skeletal S:Ca ratio. This is also supported by a calcifying fluid study and boron isotope measurements that suggested bamboo corals undergo a minimal modification of their internal pH compared to scleractinian corals (Farmer et al., 2015). As sulfate also replaces the carbonate ion in the carbonate lattice, an analogous mechanism may be involved in iodate incorporation. The additional bicarbonate active transport in bamboo corals may be responsible for their lower I/Ca values compared to scleractinian corals. We thus argue that the biological control on iodate partitioning between different DSC families represents a more robust control than mineralogical differences.

Further studies are required to better understand iodine incorporation, including the divergent results from the thermodynamic calculation and experimental measurements, the factors controlling over iodine incorporation (e.g., pH, temperature), the growth rate effects of biogenic carbonate during calcification, and DSC culture experiments of iodine uptake. This also highlights the need for biogenic carbonate studies to consider a group-specific classification, such as family, genus, or species when developing proxy applications (e.g., Kershaw et al., 2023).

With the large difference in iodine concentration in DSCs due to their carbonate structures reported here, we first focus on scleractinian coral (aragonite) and investigate the potential controlling factors of I/Ca. Scleractinian coral I/Ca increases with seawater iodate concentration, except for samples from high latitudes ( Figure 4B ). A similar relationship has been reported in synthesis experiments for calcite, with a significant linear correlation between calcite I/Ca and ambient water iodate concentration (Lu et al., 2010; Zhou et al., 2014). Aragonitic coral I/Ca is overall higher than synthetic calcite I/Ca for a given water/seawater iodate concentration ( Figure 4B ). With the exception of high latitude samples (North Atlantic and Southern Ocean), the slope of the linear regression line for the scleractinian corals is strikingly similar to that of the synthetic calcite ( Figure 4B ; Lu et al., 2010). This relationship suggests similar modes of incorporation of iodate into coral aragonite and synthetic calcite structures. Still, a few uncertainties may raise concerns about the findings. The calcite synthesis experiments were conducted in deionized water rather than seawater, which may affect the incorporation of iodate. Also, this process was found to be temperature dependent, with greater incorporation at low temperatures and less at higher temperatures (Zhou et al., 2014). Nevertheless, the consistency between iodate incorporation into coral aragonite and synthetic calcite suggests that iodine in marine carbonate is incorporated mainly from surrounding seawater and in turn that the iodate concentration of ambient seawater could be a controlling factor for coral I/Ca.

In contrast, corals from high latitudes, specifically the North Atlantic and Southern Oceans, exhibit I/Ca ratios that deviate from values expected based on their nearby seawater iodate concentrations (450 to 500 nM; Figure 5 ). To further investigate this effect, we explore the secondary environmental controls on these samples. The I/Ca values do not correlate with seawater pH, [CO₃ ^2-], dissolved phosphate concentrations, or dissolved nitrate concentrations at all sites ( Supplementary Figure 1 ). However, on a regional scale, corals from the North Atlantic show that I/Ca and temperature have a negative correlation ( Figure 5C ). Also, I/Ca is negatively correlated with salinity ( Figure 5D ) due to the fact that seawater temperature and salinity are highly correlated at these sites. A similar negative correlation between I/Ca and temperature has also been reported in synthetic calcite experiments (Zhou et al., 2014) and core-top benthic foraminifera from the Little Bahama Bank region where oxygen concentrations are high (150 to 200 µmol/kg) (Lu et al., 2022). We speculate that, similar to benthic foraminifera, corals may naturally incorporate more iodate into their carbonate structure at lower temperatures. Another possible influencing factor is seasonal seawater iodate changes at high latitudes. Seawater iodate concentration at coastal Antarctic sites varies from 250 to 400 nM, with an interannual variability of around 160 nM (three observation years at 15 m depth) related to iodate reduction by biological productivity (Chance et al., 2010). Due to the limited longer-term seawater data, it is difficult to explore further the impact of temperature and seasonal iodate variation on coral I/Ca. Thus, further investigation of iodine incorporation into DSCs from high-latitude regions is needed.

The concentration of seawater iodate varies with oxygen content, and coral I/Ca ratios reflect these iodate concentrations; hence, this allows coral I/Ca to be used as a proxy for seawater oxygen content. Our results show that the coral I/Ca-O₂ relationship undergoes an apparent “step change” at O₂ concentration ≈ 160 μmol/kg, whereafter coral I/Ca remains elevated at high oxygen levels ( Figure 4C ). This significant step change is also present in both benthic and planktic foraminifera ( Figure 3 ; Glock et al., 2014; Lu et al., 2020b; Lu et al., 2022), where a low O₂ threshold for planktics occurs at around 100 μmol/kg (I/Ca < 2.5 µmol/mol), and for benthics at about 50 μmol/kg (I/Ca < 3 µmol/mol). Although the semi-quantitative I/Ca-O₂ relationships remain empirical, the oxygen thresholds in DSC I/Ca could correspond to their metabolic limits. Laboratory studies of Lophelia pertusa (scleractinian) have shown it maintains respiratory independence at high oxygen concentrations, but that respiration rate declines rapidly below a critical oxygen level of ~140 μmol/kg (Dodds et al., 2007). These results also agree with field-based population distributions in the North Atlantic, which indicate that DSCs are absent at the local oxygen minimum of around 150 μmol/kg (Thiagarajan et al., 2013). Further paleohabitat studies in the Southern Ocean also suggest that corals were less prevalent in the deep Polar Front (~1200 m depth) during the Antarctic Cold Reversal (14.7-13 ka), when seawater oxygen levels fell below this threshold. (Stewart et al., 2021). Together, these studies suggest DSCs have a change in biological function around an O₂ threshold at ~160 μmol/kg, which could correspond to a potential change in iodate incorporation.

We next compare the I/Ca in aragonitic foraminifera versus aragonitic corals. The only I/Ca dataset from modern/recent aragonitic benthic foraminifera is Hoeglundina elegans from OMZs off the Peru margin, which suggests no significant correlation between I/Ca and bottom water oxygen concentration (Glock et al., 2014). These authors concluded that this might be due to the dissolution and recrystallisation of metastable aragonite that can happen at the early stage of diagenesis, thus altering I/Ca ratios. However, their study only focused on extremely low dissolved oxygen content (< 10 μmol/kg), meaning iodate reduction had occurred and making iodide the prevalent iodine species in this setting. The I/Ca data only varies in a relatively small range (~0.1 to 0.3 μmol/mol), which makes it difficult to detect significant variations. Our results for DSCs argue against this finding and suggest that I/Ca in aragonite could reflect seawater oxygen content. The low aragonitic H. elegans I/Ca versus higher I/Ca in scleractinian corals at locations with O₂ concentrations < ~50 μmol/kgare possibly due to the difference in seawater iodate concentrations and/or regional water mass mixing processes related to their locations (i.e., H. elegans were collected at OMZs off Peruvian margin; scleractinian corals were collected at OMZ closer to the equator), or the difference between foraminifera and DSC calcification processes.

Lastly, we find no significant oxygen threshold in our bamboo coral I/Ca results, but a hint of a slight increase of I/Ca values at O₂ concentration < ~160 umol/kg ( Supplementary Figure 3 ). This could be explained by reduced vital effects in bamboo corals (Section 4.1) and/or sampling biases given the limited available data. Additional calibration studies of bamboo corals are required before they can be used to study paleo-oxygenation. Due to these limitations, we focus on the potential use of scleractinian corals for future proxy applications.

Here, we propose that iodate incorporation in deep-sea corals is linked to seawater oxygen content. Scleractinian corals exhibit a “step-change” in I/Ca at an O₂ concentration ≈ 160 μmol/kg, which may be further utilized as an indicator to reconstruct marine oxygen content similarly to I/Ca measurements in foraminifera (e.g., Lu et al., 2020c).

Decreasing marine dissolved oxygen content as a result of global warming is predicted to accelerate in the near future (Oschlies et al., 2018). Due to the probable threat to marine ecosystems, understanding the linkages between the carbon and oxygen cycles is vital. Scleractinian DSCs offer a promising, well-dated archive of past oceanic environments, and by applying an I/Ca-oxygenation proxy, the reconstructed oxygenation records could shed new light on key paleoclimate events. For example. during the last ice age, the deep Pacific is thought to have stored more respired carbon when oxygen-depleted deep waters were expanded in the eastern tropical Pacific (Hoogakker et al., 2018). However, redox proxies commonly used to tackle this issue, such as authigenic uranium, benthic foraminifera carbon isotope gradients and biomarker preservation, sometimes yield different oxygen-level reconstructions due to the limitations of each proxy (Jacobel et al., 2020). The DSC I/Ca proxy has the advantage of recording variations in intermediate water masses, providing an independent proxy perspective on oxygen reconstructions. For example, glacial to modern seawater O₂ concentration in the South Indian Ocean is suggested to have increased from ~100 to ~220 µmol/kg (Gottschalk et al., 2020). This range covers our coral I/Ca-O₂ threshold and provides an alternative archive to study deep-ocean oxygenation, focusing on abrupt events during the last millennium owing to the absolute radiometric dating potential of DSCs.

One potential obstacle to utilizing the I/Ca proxy in scleractinian coral is the biological behavior under low oxygen conditions (e.g., Hughes et al., 2020). Recent studies have shown that DSCs are able to thrive under hypoxic environments (e.g., Hebbeln et al., 2020; Salinas-de-León et al., 2020), but the question remains whether DSCs would modify their calcification to adopt to extreme conditions. For example, Lophelia pertusa would exhibit high mortality to maintain respiratory competence under low O₂ concentration by utilizing polyp expansion to increase tissue surface area and facilitate a diffusive gas exchange (Dodds et al., 2007). However, whether these processes would affect iodate incorporation during coral skeleton formation is unknown. Culture studies are therefore required to further explore the I/Ca ratio in DSCs as an oxygenation proxy. Nevertheless, DSCs serve as a potential archive that will allow us to indicate marine oxygen concentration above or below a specific level in the past.