In this study we analyzed specimens of two species of foraminifera, A. gibbosa and S. marginalis grown under a range of controlled pCO₂, as well as larger benthic foraminifera collected from an Indo-Pacific reef aquarium. The experimental design of the culture experiment under different pCO₂ has been described in detail in Van Dijk et al. (2017b). In short, different species of larger benthic foraminifera collected in the Caribbean Sea were incubated under controlled conditions at the Caribbean Netherlands Science Institute (CNSI; St. Eustatius). Foraminifera were cultured in groups in 70 ml Falcon^® tissue bottles in four batches of seawater with added Calcein (5 mg/l), which were in equilibrium with four different pCO₂ (350, 450, 760, and 1,200 ppm named, respectively, treatment A, B, C, and D) by a pCO₂-control system developed in-house (described in detail in Webb et al., 2017). Culture media was replaced every 4 days. At the start and termination of the experiment, 125 ml samples from the stock seawater solutions (one for each experimental treatment) were collected to analyze the concentration of dissolved inorganic carbon and total alkalinity on a Versatile INstrument for the Determination of Titration Alkalinity at the CNSI. Physico-chemical parameters of the culture conditions can be found in Table 1 . Foraminifera were cultured at 25 ± 0.2°C for 21 days, during which they build on average 2.4 and 5.0 chambers for A. gibbosa and S. marginalis, respectively. After termination of the experiment, specimens were rinsed three times with de-ionized water and dried at 40°C. Foraminiferal samples were transported in slides to the Royal Netherlands Institute for Sea Research (NIOZ) to investigate the Mn/Ca of the shell and growth parameters. To assess the number of chambers added during the experiment, fluorescent calcein-stained chambers were counted using a ZEISS Axioplan 2 fluorescence microscope equipped with appropriate excitation and emission optics. Pictures were taken using a ZEISS Axiocam MRc 5 camera.

At Burgers’ Zoo in Arnhem, Netherlands, coral debris, rich in foraminifera (Ernst et al., 2011) were collected from the Indo-Pacific coral reef aquarium, one of the largest coral reef aquaria in the world. From these coral debris, hyaline (Amphistegina lessonii and Heterostegina depressa) and porcelaneous (Sorites orbiculus, Spiroculina angulata, Spiroculina communis, Quinqueloculina pseudoreticulata, Quinqueloculina sp.) species of foraminifera were isolated to analyze the shell chemistry. Foraminifera from both the experiment and the aquarium were cleaned following an adapted version of the Barker protocol (Barker et al., 2003), described in Van Dijk et al. (2017a, 2017b). In short, foraminifera were cleaned using an oxidizing step in which organics were removed with a 1% H₂O₂ solution (buffered with 0.1 M NH₄OH), and consequently, after gentle ultrasonication, rinsed with ultrapure water and dried in a laminar flow cabinet.

The sampling strategy and general seawater composition (Ca, Mg, Na, Sr and Ba) of the media used in the pCO₂ controlled experiment has been described in Van Dijk et al. (2017b). In short, at the start and end of the experiment and during replacement of the culture media every four days, subsamples were collected and acidified with three times Quartz distilled HCl to pH ∼ 1.8. Additionally, seawater from the aquarium in Burgers’ Zoo was sampled during collection of coral debris at two different occasions. In total, we collected three seawater samples (total n = 6) in 50 ml Falcon tubes, which were returned to the laboratory, acidified and stored at 7°C upon analysis. The seawater composition of the experimental samples as well as the Burgers’ Zoo seawater was analyzed on an Element-2 SF-ICP-MS run in medium resolution mode. International Association for the Physical Sciences of the Ocean Standard Seawater was used as a drift monitor. Analytical precision (RSD) was 3% for Ca, 4% for Mg, 1% Na, 1% for Sr and 5% Ba. For the samples of the culture experiment, we obtained average values of 5.25 ± 0.06 mol/mol for Mg/Ca, 44.6 ± 0.6 mol/mol for Na/Ca, 8.63 ± 0.05 mmol/mol for Sr/Ca, and 9.04 ± 0.47 μmol/mol for Ba/Ca. For Burgers’ Zoo seawater, Mg/Ca was 6.1 ± 0.9 mol/mol.

Due to the low concentration, [Mn] of subsamples from the culture experiment and tropical aquarium had to be analyzed using a commercially available pre-concentration system, SeaFAST S2. With the SeaFAST system elements with low concentrations are pre-concentrated to values above detection limit of the SF-ICP-MS. Accordingly, we measured Cd, Pb, U, B, Ti, Mn, Fe, Co, Ni, Cu, and Zn. In short, 10 ml of sample was mixed with an ammonium acetate buffer to pH 6.2 and loaded on a column containing NOBIAS chelating agent. After rinsing the column with a diluted ammonium acetate buffer the metals were eluted in 750 µL of quartz distilled 1.5 M HNO₃ before being quantified on the SF-ICP-MS. Analytical precision (RSD) was 4% for Mn and we obtained an average seawater Mn/Ca value of 5.4 ± 0.3 μmol/mol for the culture experiment, and of 14.8 ± 0.2 μmol/mol for the Burgers’ Zoo samples.

For the hyaline species A. gibbosa, Mn/Ca is 4.4 ± 2.1 μmol/mol for all measurements on average. When pCO₂ increases from 350 to 1,200 ppm, average Mn/Ca increase significantly (p < 0.0025; R ² = 0.95) with pCO₂ ( Figure 1C ). From lowest to highest pCO₂ condition, Mn/Ca increases from 3.2 to 5.9 μmol/mol ( Figure 1A ), an overall increase of factor 1.9, or 22% per 100 ppm CO₂. For the correlation between [CO₃ ²⁻] and pH with Mn/Ca, see Supplementary Figure S1 . Mn/Ca of the porcelaneous species S. marginalis are ∼11 times higher than the hyaline species A. gibbosa. For this species, Mn/Ca increases significantly with pCO₂ as well (p < 0.01; R ² = 0.91; Figure 1D ) by a factor of 2.1, or 25% per 100 ppm CO₂, from 42.7 to 90.3 μmol/mol. For both species, standard deviation (SD) per treatment increases with increasing pCO₂, and hence increasing Mn/Ca. For A. lessonii, the relative SD (RSD) is on average 40% for the different conditions, while for S. marginalis, the RSD increases from 35 to 51% with increasing pCO₂.

Individual measurements of Mn/Ca and Mg/Ca for both species are shown in Figure 2 . Average Mg/Ca is 27.4 ± 5.3 and 141.9 ± 5.3 mmol/mol for A. gibbosa and S. marginalis, respectively ( Table 3 ), which fits very well with earlier measurements on the same specimens, where the average Mg/Ca values obtained were 27.7 and 145.8 mmol/mol, respectively (Van Dijk et al., 2017b), illustrating the good reproducibility of the LA-ICP-MS analyses. Partitioning of Mg and Mn is, on average respectively 4.7 and 13.5 times higher in S. marginalis compared to A. gibbosa.

On average for all conditions, ∼80% of the specimens of S. marginalis (for example, see Figure 3B ) added new chambers during the experiment, vs. ∼90% of A. gibbosa individuals (for example, see Figure 3C ). There is no significant trend of chamber addition rate (number of chambers added per day) with pCO₂ for both species studied ( Figure 3A ). On average, specimens of S. marginalis added more than two times as many chambers during the culture period, 0.28 ± 0.12 chamber per day vs. 0.12 ± 0.05 for specimens of A. gibbosa.

On average partitioning of Mn (D_Mn) and Mg (D_Mg × 10³) are, respectively: 0.8 ± 0.4 and 5.5 ± 0.1 for A. lessonii, 1.9 ± 0.8 and 26.7 ± 0.3 for H. depressa, which are both hyaline species; and 16.2 ± 3.2 and 26.7 ± 0.4 for S. orbiculus, 11.8 ± 1.6 and 25.3 ± 0.4 for S. angulata, 18.6 ± 1.9 and 22.5 ± 0.4 for S. communis, 23.0 ± 1.8 and 25.9 ± 0.4 for Q. pseudoreticulata and 20.4 ± 2.7 and 23.5 ± 0.4 for Quinqueloculina sp., which are five porcelaneous species (see Table 4 ). The D_Mn and D_Mg values are plotted together to visualize possible trends between the two groups of foraminifera ( Figure 4 ). In general, porcelaneous species incorporate more Mn than hyaline species and the species investigated have similar Mg content (D _Mg × 10³ of 22.5–26.7). In contrast, hyaline species comprise a wider range in D _Mg compared to the porcelaneous species.

For both species studied here, the hyaline A. gibbosa and the porcelaneous S. marginalis, we observe that shell Mn/Ca increases with increasing pCO₂ (and hence decreasing sea water pH, [CO₃ ²⁻] or Ω). A similar trend was observed for the same species for Zn and Ba, but was not found case for elements such as Mg, Na and Sr (Van Dijk et al., 2017b). We do not suggest that Mn/Ca values can be used as recorders of the carbonate system, since Mn/Ca in foraminiferal shells is hypothesized influenced by seawater [Mn] (Barras et al., 2018), but this could hamper the precision of this proxy (cf. Chapter 4.3). Still, comparing sensitivity of Mn incorporation to pCO₂ to that of other elements provides information on the underlying mechanisms involved in organismal trace element uptake. Even though the slope of the Mn/Ca-pCO₂ calibrations differs between species (0.003 for A. gibbosa and 0.05 for S. marginalis), the incorporation of Mn as a function of pCO₂ is, in terms of relative sensitivity, similar for A. gibbosa (22% increase in Mn/Ca per 100 ppm CO₂) and S. marginalis (25% increase in Mn/Ca per 100 ppm increase in CO₂). This is surprising since these species are known to have contrasting calcification strategies (e.g., Bé et al., 1979; De Nooijer et al., 2009a; Hemleben et al., 1986). This could imply that Mn incorporation as a function of pCO₂ is governed by identical processes in calcification. This might relate to the (active) uptake of trace elements or the subsequent precipitation of carbonate. The first factor inherently includes e.g., symbiont and enzyme activity, speciation of Mn in seawater and pH-dependent absorption of Mn to calcite. The second factor, actual carbonate precipitation, includes both abiotic factors such as precipitation rate and also potentially biological controls during this phase.

Since speciation of Mn is sensitive to carbonate chemistry of the seawater ( Figure 5A ) and also at the SOC, offsets in internal pH during calcification between species might be responsible for observed difference in Mn incorporation. Since internal (De Nooijer et al., 2009a) and external pH (Glas et al., 2012b) are linked due to proton pumping (Toyofuku et al., 2017), a higher internal pH would correspond to a lower pH in the foraminiferal microenvironment. This will lead in turn to a higher activity of free Mn outside the organism, which could be transported through Ca-channels. Following this hypothesis, foraminifera with higher Mn/Ca, like H. depressa ( Figure 4 ), might have a higher passive transport of Mn and/or a higher internal pH during calcification compared to A. gibbosa. No independent evidence exists, however, for such a difference in internal pH between species or between the groups precipitating a hyaline or porcelaneous shell. For example, for hyaline species, also changes in the regulation of Ca-ATPase (Prazeres et al., 2015) between species might be responsible for the observed differences in Mn incorporation.

When investigating Mn incorporation across a wider range of foraminiferal species, we observe that overall Mn/Ca values seem linked to Mg content which seems to primarily reflect calcification strategy ( Figure 4 ). Porcelaneous species incorporate in general more Mn and Mg, however we observe no correlation between shell Mg/Ca and Mn/Ca within this group of species. In contrast, within the hyaline species, we observe that species that have a higher Mg content also incorporate more Mn. Based on inorganic precipitation experiments, higher precipitation rates would lead to higher Mg partitioning (Mavromatis et al., 2013), but would also lower Mn/Ca_CALCITE due to a decrease in partitioning (Lorens, 1981). This could potentially explain the observed increase of Mg/Ca_CALCITE and simultaneous decrease of Mn/Ca_CALCITE with temperature in cultured A. lessonii (Van Dijk et al., 2019a). However, it does not explain the trends we observe for different hyaline and porcelaneous species ( Figure 4 ), in which both Mg/Ca and Mn/Ca increase in concert. Whatever the cause of the offset in Mg content between species (discussed in more detail in Van Dijk et al., 2019a), difference in Mg/Ca can potentially by itself alter the incorporation of other elements, through crystal lattice distortion. Based on inorganic experiments, Mg incorporation alters the structure of the crystal lattice (Mucci, 1987) and the resulting strain on the destabilized crystal lattice allows for higher incorporation of other trace metals. Lattice distortion has been proposed before to explain an increase in Na with Mg for larger benthic foraminifera (e.g., Evans et al., 2015). Similarly, this might explain the observed trends for hyaline species, for which we observe a co-variation of Mn/Ca and Mg/Ca between individuals ( Figure 2 ) and between species ( Figure 4 ). However, this would also imply that porcelaneous, and high-Mg hyaline species that have the same D_Mg as high-Mg porcelaneous species (like H. depressa), should have a similar D_Mn. However, hyaline species have lower calcite Mn/Ca values compared to porcelaneous species, even when looking at species with similar Mg content, for instance H. depressa (D_Mg × 10³ = 26.7, D_Mn = 1.9) compared to S. marginalis (D_Mg × 10³ = 26.6, D_Mn = 13.2), which incorporates ∼7 times more Mn while Mg content is similar ( Figure 4 ). This suggests there is an additional process, probably fundamentally related to calcification of either hyaline or porcelaneous shells, which respectively decreases or increases the Mn/Ca of the precipitated calcite.

In this study we observed that partitioning of Mn is species-specific and that Mn/Ca is positively correlated with pCO₂ ( Figure 1 ) and/or negatively correlated with pH, [CO₃ ²⁻], Ω ( Supplementary Figure S1 ). Although it is complicated to incisively explain this empirical relation with a single process, these observations have to be taken into account when applying Mn/Ca to field or (sub) geological samples as a proxy for oxygenation (Groeneveld and Filipsson, 2013; Ní Fhlaithearta et al., 2018; Petersen et al., 2018; Guo et al., 2019). Calibrations of the Mn/Ca oxygenation proxy for foraminifera require quantifying the direct impact of carbonate parameters on the proxy sensitivity, especially since covariation between pCO₂ and oxygenation can occur in benthic environments. However, we expect that for benthic species the impact of differences in carbonate chemistry on foraminiferal Mn/Ca is limited compared to the effect of oxygen-induced changes in Mn²⁺ and subsequently foraminiferal Mn/Ca, due to the low sensitivity observed in this study ( Figure 1 ). Further studies are needed to decouple the effect of both environmental parameters, like obtaining sensitivity of Mn/Ca of larger benthic species to changes in O₂/seawater [Mn²⁺]. Furthermore, for calibration studies, in the water column, parameters of the carbonate chemistry will behave similar to this culture experiments, but in the sediment, characterized by high alkalinity, changes in carbonate chemistry might be buffered, and fluctuations in e.g., pH could therefore become negligible. More importantly, our observations of major differences in Mn incorporation between hyaline, both low and high Mg species, and porcelaneous species ( Figure 4 ) is crucial to take into account for calibration and application of a potential O₂ proxy. Reconstructing oxygen concentration requires species-specific calibrations for Mn/Ca as well as insight into factors potentially affecting Mn speciation.