To determine Ba / Ca partitioning, benthic foraminiferal culture experiments were set up with five different seawater Ba / Ca ratios (54–92 µmol mol-1). Media were prepared by increasing [Ba2+]sw while keeping the [Ca2+]sw constant. The range of [Ba2+] values used in these experiments exceeds the range of concentrations found naturally and allows the testing of the applicability of partition coefficients under conditions with artificially high seawater Ba / Ca. Seawater is only slightly undersaturated with respect to barite (BaSO4) and an increase in [Ba2+] in the sea water will cause barite precipitation (Langer et al., 2009). To be able to increase [Ba2+] beyond its natural range, artificial seawater was prepared with lower sulphate contents. All other salts were added according to the recipe of Kester et al. (1967) to produce a total of 5 L of medium for each treatment. As Amphistegina lessonii and Heterostegina depressa do not grow well in 100 % artificial seawater, the prepared media were mixed with natural seawater in a ratio of 9 : 1 (Mewes et al., 2014). To double check concentrations and determine potential loss of elements due to precipitation, sorption and/or scavenging, element concentrations of the culture media were determined using inductively coupled plasma optical emission spectrometry (ICP-OES) at the Alfred Wegener Institute in Bremerhaven, except for Ba, which was measured using ICP-MS at Utrecht University (Table 1).

Culture media pH was adjusted to 8.0 by adding NaOH (1 M) to the prepared media. Before the start of the experiments, dissolved inorganic carbon (DIC) and total alkalinity were measured at the Alfred Wegener Institute. DIC was measured photometrically in triplicates with a TRAACS CS800 QuAAtro AutoAnalyzer with an average reproducibility of ±10 µmol L-1. Alkalinity was calculated from linear Gran plots (Gran, 1952) after triplicate potentiometric titration (Bradshaw et al., 1981) using a TitroLine alpha plus autosampler (Schott Instruments). Parameters of the total carbonate system were calculated from temperature, salinity, DIC and alkalinity using the program CO2SYS adapted to Excel by Pierrot et al. (2006). The equilibrium constants K1 and K2 from Mehrbach et al. (1973), as reformulated by Dickson and Millero (1987), were used (Table 1).

Living specimens of A. lessonii and H. depressa were isolated from sediment collected at the tropical aquarium of Burger's Zoo (Arnhem, the Netherlands) in August 2012 and transferred to the Alfred Wegener Institute for the culture experiments. Healthy individuals of A. lessonii showing pseudopodial activity, a dark brown cytoplasm and minimal signs of bleaching were handpicked with a small brush under a Zeiss Stereo microscope and transferred to well plates. Adult specimens of H. depressa were picked directly from the aquarium with soft tweezers. After 2 weeks several individuals of both species underwent asexual reproduction. Individual H. depressa parent cells produced sufficient numbers of juveniles to study separate clone groups. Approximately 20 juveniles with two or three chambers from the same parent were selected for every treatment and divided over two petri dishes (diameter 55 mm, containing approximately 10 mL of culture medium). In total, two clone groups were used in the experiments, resulting in a total of at least 40 individuals per treatment. Specimens of A. lessonii did not produce sufficient numbers of juveniles for analysis of separate clone groups. Therefore, approximately 60 juveniles with two or three chambers from different parents were selected per treatment and distributed evenly over three petri dishes. All experiments were carried out in an adjustable incubator (RUMED Rubarth Apparate GmbH) at a constant temperature of 25 ∘C. As both species are symbiont-bearing, a 12 : 12 light : dark cycle was applied with a constant photon flux density of approximately 250 µmol photons m-2 s-1 during light hours. Pictures were taken weekly under a Zeiss Axiovert 200M inverted microscope and maximal diameters of the shells were measured with the AxioVision software to allow the determination of the chamber addition rates of the foraminifera in the experiments. The experiments were terminated after 6 weeks.

All specimens were fed Dunaliella salina algae every 3 to 4 days. Although A. lessonii hosts symbionts, this foraminiferal species does not exclusively rely on nutrients from their symbionts but also ingests algae (Lee, 2006). To avoid changes in the barium concentration of the culture media, the water in the dishes containing foraminifera was diluted as little as possible by the solution containing the food for the foraminifera. For this purpose, foraminifera were fed 50 µL of a solution containing algae that was centrifuged at 2000 rpm for 10 min. Algae concentrated at the bottom of the tube were transferred to an empty tube with a pipette. To prevent changes in the culture media's carbonate chemistry by algal photosynthesis the algae were killed by heating the concentrated solution in an oven at 90 ∘C for 10 min. The cultures were transferred to new petri dishes every week to avoid excessive bacterial growth, potential build-up of waste products and shortage of ions or nutrients. To prevent changes in salinity by evaporation, media were refreshed 3 days after the cultures were transferred to new dishes by pipetting approximately 5 mL of the old media out of the petri dish and replacing it with the same volume of media from the prepared batch.

At the end of the culture experiment, specimens were cleaned by placing them in a 7 % NaOCl solution for approximately 30 min until completely bleached and organic material was removed from the tests. This cleaning method is shown to have a similar impact on average foraminiferal Ba / Ca values as cleaning with H2O2 and is relatively small (2–3 µmol mol-1) compared to cleaning with deionized water only (Pak et al., 2004). Specimens were then rinsed three times for approximately 60 s in deionized water to remove the NaOCl and any residual salts from the culture solutions. Cleaned foraminifera were put in an oven at 42 ∘C until completely dry and mounted on sample holders using double-sided adhesive tape.

Element composition of the calcite was determined using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) at Utrecht University (Reichart et al., 2003). The system consisted of a GeoLas 200Q 193 nm Excimer laser (Lambda Physik) connected to a sector field ICP-MS (Element2, Thermo Scientific). Samples were ablated in a single-volume chamber and the aerosol was carried to the ICP-MS by a helium flow. Monitored masses included 23Na, 24Mg, 26Mg, 27Al, 43Ca, 44Ca, 55Mn, 88Sr, 138Ba, and 238U and calibration was performed using a glass standard (NIST 610) that was ablated three times after every 10–12 foraminiferal samples. Diameter of the ablation crater was set to 80 µm for all specimens and the pulse repetition rate was 6 Hz. The ablated calcite was measured and integrated with respect to time. Energy density for the glass was higher than for the foraminifera (5 and 1 J cm-2, respectively). Although the resulting difference in ablation characteristics is not likely to affect obtained foraminiferal element concentrations (Hathorne et al., 2008), foraminiferal element concentrations were compared to those from a calcite standard made in-house with known element concentrations and ablated at the same energy density as the foraminifera (Dueñas-Bohórquez et al., 2009). Relative standard deviation for Mg / Ca and Sr / Ca based on repeated measurements on this material was < 5 % for both ratios. Due to the lamellar nature of Rotaliid foraminifera, the final chambers are thinnest and are therefore characterized by the largest uncertainty in the estimated average element / Ca ratio. Therefore, the F chamber was not considered and instead, the F-1 chamber of A. lessonii was ablated for every specimen. For H. depressa, walls of the final two chambers were commonly too thin for reliable chemical results and, therefore, the F-2 chamber was analyzed. In addition, for each species, the final 6–7 chambers of 10 sufficiently large specimens (two from each of the five treatments) were ablated to analyze intra-specimen variability in Ba / Ca, to analyze variability within chamber walls as a function of thickness and to detect potential ontogenetic trends in Ba incorporation.

Elemental concentrations were calculated from the ablation profiles with the Glitter software, using 43Ca as an internal standard and values from Jochum et al. (2011) for concentrations of elements in the NIST 610. This program integrates the ablation signal after subtracting the background signal to calculate the elemental concentrations. To avoid contaminated intervals of the ablation profile, sections with high 27Al and 55Mn counts were excluded from the analysis since these parts are also often characterized by unusually high Mg / Ca not reflecting the actual shell carbonate. Ablation profiles with a duration shorter than 5 s were rejected as such short profiles are unreliable due to poor counting statistics. Nine out of 188 ablation profiles were rejected for A. lessonii and 7 out of 140 profiles from H. depressa were discarded, which is less than 5 %.

To compare the results from cultured specimens with Ba / Ca from specimens derived from natural conditions, a number of living specimens of both A. lessonii and H. depressa were isolated from the zoo's stock (i.e., sediment collected at the zoo from which the specimens were isolated; Sect. 2.2) and cleaned and prepared for LA-ICP-MS analyses as described in Sect. 2.3. From both species, seven specimens were ablated twice at the Royal Netherlands Institute for Sea Research (NIOZ) using a NWR193UC (New Wave Research) laser, containing an ArF Excimer laser (Existar) with deep UV 193 nm wavelength and < 4 ns pulse duration. Provided that the same reference material is used, the use of multiple laser systems (see above) is shown not to bias obtained foraminiferal element / Ca ratios (De Nooijer et al., 2014a). Laser ablation was performed with an energy density of 1 J cm-2 at a repetition rate of 6 Hz for calcite samples and an energy density of 5 J cm-2 for the glass (NIST610) standards. Helium was used as a carrier gas with a flow rate of 0.8 L min-1 for cell gas and 0.3 L min-1 for cup gas. From the laser chamber to the quadrupole ICP-MS (iCAP Q, Thermo Scientific), the He flow was mixed with ∼ 0.4 L min-1 nebulizer Ar. Before measuring the samples, the nebulizer gas, extraction lens, collision cell technology (CCT) focus lens and torch position were automatically tuned for the highest sensitivity of 25Mg by laser-ablating MACS-3. The masses measured by the ICP-MS were 23Na, 24Mg, 25Mg, 27Al, 43Ca, 44Ca, 88Sr and 138Ba. JCp-1, MACS-3 and an in-house (foraminiferal) calcite standard (NFHS) were used for quality control and were measured every 10 foraminiferal samples. Internal reproducibility of the analyses was all better than 9 %, based on the three different carbonate standards used. Intensity data were integrated, background subtracted, standardized internally to 43Ca and calibrated against the MACS-3 signal using a custom-built MATLAB routine within the program SILLS (Guillong et al., 2008). Since ablation of the NIST SRM 610 and NIST SRM 612 could increase the sodium background, they were only ablated and analyzed at the end of every sequence and cones were cleaned before the next sequence. Accuracy of the analyses was better than 3 %, based on comparison of the carbonate standards with internationally reported values (Okai et al., 2002; Wilson et al., 2008). Signals were screened for surface contamination and parts of the outside or inside of the shell with elevated Mg, Mn or Al values were eliminated from the area selected for integration.

Seawater samples from the zoo's aquarium were measured in duplicate using a sector field ICP-MS (Element2, Thermo Scientific). The ICP-MS was run in low-resolution mode (24 cycles) for 138Ba and in medium-resolution mode (24 cycles) for 43Ca. Calibration was performed through an external calibration series with increasing concentrations of Ba.

Average shell diameters increased considerably during the experimental period (Fig. 1). Overall, increase in shell diameter did not significantly differ between treatments. Treatment C (seawater Ba / Ca = 64 µmol mol-1) for A. lessonii, however, shows somewhat reduced chamber addition rates per incubated specimen. This may be the consequence of slightly higher mortality under these conditions and a relatively high number of specimens that did not add any chambers. Although not systematically investigated, two petri dishes from this treatment contained relatively many bleached (i.e., devoid of symbionts) specimens at the end of the 6-week period.

Calcite Ba / Ca increases linearly with seawater Ba / Ca for both species (Fig. 2, Table 2). ANOVA performed on the individual data points combined with regression analyses reveals a significant increase in Ba / Cacc with Ba / Casw for both species (Table 3). Calculated regression slopes result in a DBa of 0.326 (±0.005) for A. lessonii and 0.777 (±0.007) for H. depressa (Fig. 3, solid lines). Regression lines are forced through zero as it seems reasonable to assume that no Ba is incorporated into calcite when the Ba concentration in the seawater is zero. Without this forcing, regression slopes would be Ba / Cacc=0.34×Ba/Casw-1.1 for A. lessonii and Ba / Cacc=0.92×Ba/Casw-10 for H. depressa. The resulting partition coefficients ((Ba / Cacc)/(Ba / Casw)) are constant and significantly different between the species (ANOVA) (∼ 0.3 for A. lessonii and ∼ 0.8 for H. depressa) over the range of seawater Ba / Ca studied here. The regression line for Ba / Cacc as a function of Ba / Casw for A. lessonii corresponds well with that reported for a number of different low-Mg species (Lea and Boyle, 1989).

The aquarium-derived specimens (aquarium samples) had a diameter ranging from 550 to 1180 µm (with an average of 975 µm) for A. lessonii and from 1380 to 2340 µm (average: 1936 µm) for H. depressa. They had an average Ba / Ca of 15.4 (±2.3 SD) µmol mol-1 for A. lessonii and 35.7 (±14 SD) µmol mol-1 for H. depressa. In combination with the measured aquarium's seawater Ba / Ca of 35.7 (±3.9 SD) µmol mol-1, the partition coefficients for Ba vary between 0.43 and 1.0 for A. lessonii and H. depressa, respectively. The aquarium-derived data are consistent with the controlled-growth-derived data, but they were not used in the regression analysis (Fig. 2) since the conditions (e.g., carbonate chemistry) under which the specimens from the aquarium were grown were not determined as precisely and accurately as in our culturing experiment. Including these data in the linear regression (Fig. 2) would change the sensitivity from 0.78 to 0.77 for H. depressa and from 0.33 to 0.32 for A. lessonii.

From both species, 10 specimens were used to quantify the relation between ontogeny (i.e., size-dependent) and Ba incorporation into foraminiferal calcite. For this purpose, the final 6–7 chambers of these individuals were ablated (Fig. 3). With the selected spot diameter (80 µm), ablation of a small amount of material of adjacent chambers could not always be avoided. Some chamber walls, particularly of the youngest (i.e., built latest) chambers, were too thin for reliable measurements and were excluded from further consideration.

Since these specimens were cultured at different Ba / Casw ratios, the interchamber variability is expressed as the difference of a single-chamber Ba / Ca and the individual's average Ba / Ca. Positive single-chamber values indicate higher-than-average values, whereas negative values indicate single-chamber Ba / Ca below that individual's average Ba / Ca (Fig. 3).

In H. depressa, Ba / Cacc increases significantly with subsequently new chambers added (Fig. 3). Regression analysis reveals an average increase of 1.43 µmol mol-1 Ba / Cacc with every chamber added (Table 4). Ba / Cacc appears to decrease with chamber position in A. lessonii, although the ANOVA p value shows that this is statistically not significant. Still, removing one single outlier already results in a p value lower than 0.01, indicating that the current data set does not allow the rejection of the presence of a trend for A. lessonii.

Combining data from all five treatments, average Mg / Ca of A. lessonii was 64 mmol mol-1, with a relative standard deviation of 47 %. Within treatments, the variability in Mg / Ca is considerably lower (between 27 and 37 %). Average Mg / Ca in H. depressa was 152 mmol mol-1, with a standard deviation of 25 mmol mol-1 (16 %). Within treatments, the relative standard deviation ranged from 4.1 % (treatment E) to 17 % (treatment D). The species-specific single-chamber Mg / Ca and Ba / Ca ratios combined for all treatments are positively and significantly related (Fig. 4). For A. lessonii, Mg / Ca = 3.1 × Ba / Ca - 3.6 (t value = 12.2, p < 0.01 for the slope of the regression) and for H. depressa, Mg / Ca = 1.1 × Ba / Ca + 92 (t value = 14.8, p < 0.01 for the slope). The slopes of these two regressions (3.1 and 1.1) are significantly different: this is calculated by z=(aHeterostegina-aAmphistegina)/√(SEa,Heterostegina2+SEa,Heterostegina2), in which a is the value for the regression's slope and SEa is the slope's associated standard error. For the slopes of the Mg / Ca–Ba / Ca regressions for Amphistegina and Heterostegina, the resulting z score is higher than > 7, indicating that the two slopes are significantly different.

When comparing the single-chamber DBa with DMg of all data combined, the partition coefficient for Mg is over 30 times lower than that of Ba (Fig. 4). Over the range in Ba / Casw studied here, the relation between DBa and DMg is linear within both species. For A. lessonii, DMg=40×DBa-2.0 (t value = 7.3, p < 0.01 for the slope of the regression) and for H. depressa, DMg=29×DBa+3.8 (t value = 6.5, p < 0.01 for the slope). The slopes of these two regressions (40 and 29) are not significantly different (z score 1.6). When combining the data from both species, the regression equals DMg=34×DBa+0.073 (t value = 29.9, p < 0.01 for the slope).

The range of Ba concentrations used in the experiments did not influence the increase in shell diameter of either foraminiferal species (Fig. 1). Compared to H. depressa, increases in shell diameter (which is proportional to the chamber addition rate) for A. lessonii were slightly more variable. To prevent barite precipitation it was necessary to reduce the sulphate concentration below that typically measured in natural seawater. Sulphate concentrations between 0.1 and 1 mmol L-1 do not affect inorganic calcite growth (Reddy and Nancollas, 1976), but a decrease in growth rates of approximately 30 % was observed in coccolithophores growing in artificial seawater with a sulphate concentration 10 % that of natural seawater (Langer et al., 2009). Although coccolithophores and foraminifera may respond differently to lowered sulphate concentrations, this reduction could have hampered growth of the specimens in our culturing experiment. Chamber addition rates of A. lessonii in a culture setup with sulphate concentrations similar to those of natural seawater (Mewes et al., 2014) were approximately 20 % higher than chamber addition rates observed in our experiments. Since these experiments were not performed simultaneously using specimens from the same batch, it is not straightforward to compare absolute rates and therefore the 20 % difference cannot unambiguously be attributed to sulphate concentration (Hoppe et al., 2011). Unfortunately no data exist on the effect of reduced sulphate concentrations on the uptake of trace elements in foraminiferal calcite. However, Langer et al. (2009) demonstrated that sulphate limitation had no discernible effect on Ba incorporation in coccolithophore calcite.

The variability in Ba / Ca between individual ablation craters is considerable, but the average foraminiferal Ba / Ca shows a consistent relation with seawater Ba / Ca. This implies that the observed variability is a reflection of the inhomogeneous distribution in the test and is hence filtered out when averaging. This is similar to the behavior for Mg and Sr (Sadekov et al., 2008; Wit et al., 2012; De Nooijer et al., 2014a) and underscores the power of single-chamber analyses. If present, inhomogeneity in test wall Ba / Ca in combination with different cross sections sampled during the ablation potentially account for the observed variability. This would imply that although large differences are observed within a test wall, the average still reliably reflects seawater concentration (this paper) and for Mg, still reflects seawater temperature (Hathorne et al., 2009). Comparing within-specimen and between-specimen variability, De Nooijer et al. (2014a) showed that within-specimen variability does not account for all of the observed variability in Mg / Ca in Ammonia tepida. This seems to be similar for Ba / Ca (compare Fig. 4 in this paper with Fig. 5 from De Nooijer et al., 2014a), which would mean that at least 20 chambers need to be analyzed to reach a 5 % relative precision (De Nooijer et al., 2014a). This is not limited by the analytical precision, but rather is due to inherent biological interchamber and interspecimen variability. To reduce ontogenetic variability (in, for example, paleoceanographic applications where complete specimens are measured), a narrow size fraction should be analyzed.

Incorporation of Ba in H. depressa shows a partitioning that is about 2.5 times higher than in A. lessonii. Such a large offset of DBa between benthic species fits previously reported (differences in) partition coefficients for Ba. Lea and Boyle (1989) found DBa=0.37±0.06 for Cibicidoides wuellerstorfi, Cibicidoides kullenbergi and Uvigerina spp. for a series of core tops, comparable to the partition coefficient reported here for A. lessonii (0.33 ± 0.022, Fig. 2). In contrast, partition coefficients for Ba in planktonic foraminifera are roughly only twice as low as these benthic foraminiferal partitioning coefficients (0.14–0.19; Hönisch et al., 2011; Lea and Boyle, 1991; Lea and Spero, 1992). Although temperature, pH, salinity and pressure were initially proposed as potential explanations for the offset between planktonic and benthic DBa (Lea and Boyle, 1991; Lea and Spero, 1992), studies by Lea and Spero (1994) and Hönisch et al. (2011) showed no significant impact of temperature, pH and salinity on Ba incorporation into planktonic foraminiferal calcite. This would leave hydrostatic pressure to explain the difference between benthic and planktonic species. Conversely, van Dijk et al. (2017) showed that in a number of larger benthic foraminifera, Ba / Ca is positively influenced by pCO2. Our observations show, however, that the observed differences in DBa between H. depressa and A. lessonii and also the offset with the planktonic species are inherent to these species. A small impact of environmental parameters other than seawater Ba / Ca may account for the slightly higher DBa in the foraminifera taken from the aquarium compared to the cultured ones (Fig. 2). The overall differences in partitioning seem to coincide with different taxonomic groups, which may indicate that foraminifera may differ in their controls on transporting ions from seawater to the site of calcification. For example, the contribution of transmembrane transport versus that of seawater transport (i.e., leakage, Nehrke et al., 2013, or vacuolization, Erez, 2003) may vary between species and thereby account for differences in Mg / Ca, Ba / Ca, and so forth (Nehrke et al., 2013).

In both species cultured here, Ba / Cacc decreases significantly from the largest (i.e., built latest in life) towards the smaller chambers (Fig. 3). Observed trends were not significantly different between A. lessonii and H. depressa, suggesting that Ba / Cacc decreases at the same rate with size, despite the overall difference in Ba / Cacc (Fig. 3). Since we always analyzed chambers at the same position (F-1 for A. lessonii and F-2 for H. depressa) and since the final size of the cultured specimens was similar between treatments (Fig. 1), ontogenetic trends in Ba / Ca do not influence the trends in Ba / Ca between treatments (Fig. 2). Several other studies showed that element / Ca ratios can vary with chamber position. Raitzsch et al. (2011), for example, reported increasing B / Ca and decreasing Mg / Ca towards younger chambers in the benthic Planulina wuellerstorfi. Such patterns may be related to changes in the surface-to-volume ratio or relative changes in vital effects as foraminifera grow larger. For example, pH reduction in the foraminiferal microenvironment is related to the specimen's size (Glas et al., 2012) and may thereby affect the chemical speciation of minor and trace element, which in turn, may determine their uptake rates. Hönisch et al. (2011), however, showed that seawater pH has no noticeable effect on Ba incorporation in planktonic foraminiferal calcite, rendering changes in the pH of the foraminiferal microenvironment an unlikely explanation to account for the observed chamber-to-chamber variability in Ba / Ca. Alternatively, changes in the metabolic rate, the instantaneous calcification rate or a different partitioning between the impacts of the life processes may lead to the observed ontogenetic trend.

Bentov and Erez (2006) argued that decreasing Mg / Ca with foraminifera test size could be explained by relatively high Mg concentrations at or near the primary organic sheet (POS), which is the organic matrix on which the first layer of calcite precipitates during the formation of a new chamber. With the formation of a new chamber, a low-Mg calcite layer is deposited over all existing chambers, so that the high-Mg phase is diluted as more layers are deposited (Bentov and Erez, 2006). Future studies may indicate whether Ba / Ca is also heterogeneously distributed within chamber walls, by, for example, being enriched close to the POS (Kunioka et al., 2006). If this is the case, lamellar calcification mode may also result in changing Ba / Ca with chamber position.

If incorporation of Ba and Mg (and Na, Sr and B) are physically linked during biomineralization, interspecies differences in composition may likely be correlated across the various elements. The correlation between Mg / Ca and Ba / Ca within and between species (Fig. 4) suggests that these two elements are simultaneously affected during their incorporation. The relationship between Mg / Ca and Ba / Ca is different between the two species, which may be (partly) caused by the variability in seawater chemistry between treatments (i.e., seawater Ba / Ca and Mg / Ca; Table 1). Alternatively, incorporation of Mg in H. depressa may be close to the maximum concentration of Mg that can be incorporated into a calcite crystal lattice at ambient conditions (Morse et al., 2007). This may result in an overall asymptotic relationship between Mg / Ca and Ba / Ca as Mg / Ca approaches ∼ 200 mmol mol-1 (Fig. 4).

When correcting for the different seawater Ba / Ca and Mg / Ca between treatments, incorporated Ba and Mg correlate similarly within as well as between the two species studied here (Fig. 4). This suggests that these elements are coupled during biomineralization itself and that the ratio of Ba and Mg in seawater is preserved during calcification by these species of foraminifera. When comparing the relation between Ba / Ca and Mg / Ca from other benthic species (e.g., Lea and Boyle, 1989, Fig. 2; Lea and Spero, 1994; Hönisch et al., 2011; Evans et al., 2015), the coupling between Ba and Mg incorporation is likely similar across a wide range of benthic foraminiferal species.

Foraminiferal biomineralization determines the incorporation of many elements and fractionation of many isotopes during the production of new chambers, as indicated by overall large compositional differences between inorganically precipitated and foraminiferal calcite (Erez, 2003; Bentov and Erez, 2006; Nehrke et al., 2013; De Nooijer et al., 2014b). For example, Mg / Ca ratios in many species are orders of magnitude lower than what is expected from inorganic precipitation experiments. Additionally, Mg / Ca varies considerably between foraminiferal species and especially between species known to have different calcification strategies (Bentov and Erez, 2006; Toyofuku et al., 2011; Wit et al., 2012; De Nooijer et al., 2009, 2014b). Other elements such as Sr (e.g., Elderfield et al., 2000) and B / Ca (e.g., Allen et al., 2012) also vary significantly between species. Generally, concentrations for these elements correlate within taxa and hence species incorporating relatively much Mg also have high (for example) Sr / Ca, B / Ca and Na / Ca. Miliolids and many large benthic foraminifera (LBF) produce calcite with Mg / Ca of up to 100–150 mmol mol-1 (Toyofuku et al., 2000; Dueñas-Bohórquez et al., 2011; Sadekov et al., 2014; Evans et al., 2015), while most planktonic and symbiont-barren benthic foraminifera produce test calcite with Mg / Ca values ranging from 1 to 10 mmol mol-1 (e.g., Nürnberg et al., 1996; Elderfield et al., 2002; Lear et al., 2010; Wit et al., 2012; De Nooijer et al., 2014b). The same distinction is observed for B / Ca (compare, for example, Allen et al., 2012, and Kazcmarek et al., 2015), Li / Ca (Lear et al., 2010, versus Evans et al., 2015), Na / Ca (Wit et al., 2013, versus Evans et al., 2015) and Sr / Ca (e.g., Dueñas-Bohórquez et al., 2011). The correlation between relatively high (for example) Mg / Ca, Sr / Ca and B / Ca corresponds to the observed trends in the data presented here for Ba / Ca and Mg / Ca in H. depressa and A. lessonii (Fig. 4). The Mg / Ca in the former species is approximately 2.5 times that of the latter, which is similar to the difference observed in Ba / Ca ratios between these species and implies that Ba changes in concert with Mg, which is consistent with the single-chamber correlation between Mg / Ca and Ba / Ca (Fig. 4). Such a change could potentially be caused inorganically by differences in Mg opening up the crystal lattice in such a way that it can accommodate more or less Ba. Such a mechanism is described for Mg and Sr (e.g., Morse and Bender, 1990; Mucci and Morse, 1983; Mewes et al., 2015; Langer et al., 2016) and may also apply to Ba incorporation and the influence of Mg ions that increase stress in the calcite crystal lattice. Unless the strain of incorporated Mg ions does not increase linearly with its concentration, the covariance between Mg and in this case Ba may well be interrelated during an earlier stage of the biomineralization process, e.g., during their transport from the surrounding seawater into the site of calcification (Erez, 2003; De Nooijer et al., 2014b).

Interestingly, the partitioning of different elements is not the same between taxa. For example, Sr / Ca in LBF is approximately twice as high (Dueñas-Bohorquez et al., 2011; Evans et al., 2015) as in planktonic species (Elderfield et al., 2002; Dueñas-Bohórquez et al., 2009; Hendry et al., 2009), whereas the ratio between the DMg of these groups is between 10 and 100 (see above). Comparing the offset of D between groups as a function of D itself shows an approximate logarithmic correlation (Fig. 5). The distinction between the two groups on the basis of their element signature coincides with known differences in biomineralization controls. Element controls in low-Mg species are thought to be determined by (highly) selective trans-membrane ion transporters, (limited) leakage of seawater into the site of calcification and/or selective Mg2+ removal (Nehrke et al., 2013; De Nooijer et al., 2014b; Toyofuku et al., 2017). Miliolid foraminifera belong to the high-Mg foraminiferal group and are known to secrete their calcite within vesicles that are hypothesized to contain seawater, which may be modified after endocytosis (Hemleben et al., 1986; Ter Kuile and Erez, 1991; De Nooijer et al., 2009). These intracellular vesicles may therefore contain relatively high concentrations of Mg2+, Ba2+ and other ions present in seawater, although so far mainly Sr / Ca and Mg / Ca of Miliolid foraminifera have been published (Supplement). The biomineralization of non-Miliolid, intermediate- and high-Mg benthic foraminifera may employ characteristics of both these types of calcification and therefore incorporate moderate to high concentrations of elements (see Segev and Erez, 2006).