The temperature experiments were carried out with live Amphistegina lessonii, collected from turf algae on stones in the coral reef environment in the Gulf of Eilat, Israel. Uniform populations of ∼400 μm of diameter were placed in sealed 60-mL glass stoppered Erlenmeyer bottles filled with natural seawater. The natural seawater was surface-nutrient depleted-water from the Gulf of Eilat. This seawater has an initial salinity of 40.7‰ that was adjusted to 35‰ (by adding distilled water). The Erlenmeyer bottles were immersed for 2–3 months in temperature-controlled baths (at 18, 21, 24, 27, 30, and 33 °C respectively), keeping the assigned temperatures to ±0.1 °C. Water in each bottle was replaced twice a week with fresh seawater. The pH in these experiments ranged between 8.1 and 8.2 with lower pH values at the higher temperatures due to faster calcification. All bottles received the same amount of natural light (∼ 10–15 μmol photons m⁻² s⁻¹) and were examined daily for the development of algae. The foraminifera in these experiments increased their weight from 50 μg/individual to between 95 and 135 μg/individual, with optimal growth at 27 °C.

Amphistegina lobifera, collected similarly as above, were used for the cultures undertaken at various pH. The open Gulf of Eilat water used in the experiments has an average DIC of 2070 ± 10 μmol·kg⁻¹ (Silverman et al., 2007), and the original alkalinity is roughly 2500 μmol·kg⁻¹. These waters were diluted to a salinity of 35 ‰ with double-distilled water (DDW), and were boiled to remove CO₂. The DIC and the original alkalinity were thus diluted by the salinity factor of 35/40.7 (as explained above); hence the DIC was roughly 1780 μmol·kg⁻¹. These experiments were done at 24 ± 0.1 °C and the carbonate chemistry was modified by keeping the DIC constant, while the pH was lowered with HCl or raised with freshly made NaOH. The seawater reservoirs were kept out of contact with the atmosphere in plastic collapsible containers. Water was pumped with a peristaltic pump from the containers into tightly capped 60-mL Erlenmeyer flasks containing the foraminifera at flow rates between 160 and 180 mL/day. The cumulative alkalinity of the out-flowing water was measured every three days and the calcification of the foraminifera was estimated from the alkalinity difference between the reservoir and the outflow. In addition, the weight increase of the foraminifera was estimated from the final weight of the individuals compared to that of the initial control group. The weight increased on average by a factor of 5 (from 18 μg/individual to about 100 μg/individual). The different pH values for this experiment were 7.90, 8.10, 8.20 and 8.45 ± 0.10 (NBS scale).

Amphistegina lobifera were used for the cultures undertaken at various DIC. The diluted Gulf of Eilat seawater was acidified with HCl to pH 2.8 and air bubbled for 12 h to get rid of most of the dissolved inorganic carbon. Then, NaHCO₃ was added to get final concentrations of DIC of 1.0, 1.5, 2.0, 2.5, and 3.0 mmol·kg⁻¹, respectively. For comparison, the DIC range in the ocean is 1.8–2.4 mmol·kg⁻¹. The pH was adjusted with freshly prepared 1 N NaOH and 1 N HCl in order to obtain values of 8.15 for all of these waters with different DIC. The pH values ended up with a small variability of ≈ 0.01 pH units. The reagents used were the same as those for varying the pH in the previous experiments except for the NaHCO₃ (section 2.2). The water reservoirs were isolated from the atmosphere inside collapsible plastic reservoirs connected to a peristaltic pump that pumped the diluted modified seawater solutions through each Erlenmeyer flask where the foraminifera were growing. The water coming out of the Erlenmeyer was collected and weighted (every 2 or 3 days) and measured for its alkalinity. Occasionally, we also measured oxygen and pH in the water coming out of the Erlenmeyer flasks. The alkalinities were close to the original ones, but always slightly lower (because of the foraminiferal calcification). The pH of the outcoming waters was slightly different from that in the reservoir because of the metabolism of the foraminifera.

After the experiment, the cultured foraminifera were rinsed with double-distilled water (DDW) and dried overnight at 50 °C. Specimens were then placed in dilute sodium hypochlorite (1:3) for 7 h followed by several DDW rinses and overnight drying at 50 °C. Before analyses, samples were embedded in epoxy sections, polished using a diamond paste down to 1 μm and then coated with gold. Samples were put horizontal and were polished until the appearance of the top part of the knob (Rollion-Bard et al., 2008). Thus, only the sample part grown under laboratory conditions was analyzed with the ion probe.

Li isotope analyses and Li/Ca ratios were performed in situ with the ims 1270 ion probe at CRPG (Nancy, France). The analytical settings in this study are the same as those described in Vigier et al. (2007) and in Rollion-Bard et al. (2009). Briefly, the lithium isotopic compositions and Li/Ca ratios were measured using a 20- to 30-μm ¹⁶O primary beam of ∼50 nA. The mass resolving power (MRP) was set at ∼3000, with the energy slit well centered and fully opened. A single Li isotope analysis consists of peak switching on masses 6 and 7, which are successively counted on an electron multiplier. After 2 min of presputtering, each run lasts about 20 min (40 to 60 cycles of 12 s and 10 s, respectively). The background is measured every cycle for 2 s on mass 5.5. The instrumental mass fractionation (IMF) and the reproducibility of the isotopic analyses were calculated from repeated measurements of the CAL-HTP calcite reference material that was also analyzed by MC–ICPMS for Li isotopes (δ⁷Li = 13 ± 0.5‰, 2σ). The isotopic homogeneity of this sample was confirmed by multiple ion microprobe measurements along profiles (Vigier et al., 2007), and the reproducibility is better than 1‰ (1σ, n = 10).

For Li/Ca ratios, the analyses consist of peak switching on masses 7 (⁷Li⁺) and 44 (⁴⁴Ca⁺), which are successively measured on an electron multiplier and on a Faraday cup. The Li content was also measured in a glass standard (384 ppm Li, and 0.57 W.% CaO). No difference in the yield was observed between carbonate and glass standards. Backgrounds are measured every cycle for 2 s on masses 5.5 and 43.6 to ensure magnet stability. Li/Ca was analyzed in the same spot as δ⁷Li. External (spot-to-spot) reproducibility for Li/Ca ratios was determined from repeated measurements of the glass reference material and was better than 1% (1σ). The total precision for each Li/Ca analysis includes external reproducibility and internal precision.

Li/Ca and δ⁷Li data are reported in Tables 1, 2 and 3. δ⁷Li measured in the Amphistegina shells display no significant variation as a function of temperature and pH (Fig. 1a and b). For both sets of experiments, mean δ⁷Li values are close to the δ⁷Li of seawater (30.0 ± 0.9‰ and 29.8 ± 0.6‰ respectively, Fig. 1). This strongly suggests little isotope fractionation relative to the seawater – within uncertainty – for the given set of conditions. However, this is significantly different from the fractionation of Li isotopes observed during experimental growth of inorganic calcite, and strongly suggests a vital effect in favor of the heavy ⁷Li isotope (Δ⁷Li calcite–solution = –3‰ to –9‰, see Marriott et al., 2004a, 2004b). These results are also consistent with part of the published core top foraminifera data, which exhibit the same isotope compositions as seawater. However, pH and T cannot explain the significant range of δ⁷Li values measured in core top planktonic species (e.g., Hathorne and James, 2006; Misra and Froelich, 2012; Vigier et al., 2007).

Similarly, Li/Ca measured in the same tests exhibit no dependency on T and pH, with mean values of 16.4 ± 1.3 μmol·mol⁻¹ and 13.8 ± 2 μmol·mol⁻¹, respectively (Tables 1 and 2, Fig. 2). For comparison, the Li/Ca ratio of seawater is currently 2365 μmol·mol⁻¹. The lack of correlation with temperature is in agreement with studies of core top planktonic species (Hathorne and James, 2006). However, Lear and Rosenthal (2006), Lear et al. (2010), and Bryan and Marchitto (2008) have suggested that concentration in CO₃^2– could control the Li/Ca ratio of benthic foraminifera. Here, no significant influence of the pH is seen on both Li/Ca and δ⁷Li values. In this experiment performed at various pHs, but constant DIC and T, the CO₃^2– ion concentration changed by 0.4 mmol·kg⁻¹ and this did not affect the Li/Ca ratio nor the δ⁷Li. We can therefore reasonably rule out a key role of this parameter, at least for Amphistegina species. Similarly, Hathorne et al. (2009) show no significant influence of the carbonate ion concentration on Li/Ca measured in situ by LA ICPMS in the calcitic tests of two different planktonic species.

In contrast to temperature and pH (or CO₃^2–), both δ⁷Li and Li/Ca ratio exhibit large variability and good linear correlations with the solution DIC (Figs. 1 and 2). Amphistegina δ⁷Li values positively and linearly correlate with DIC. There is a strong dependency on DIC, with 20‰ differences between foraminifera tests cultured at 1.0 mmol·kg⁻¹ DIC and at 2.5 mmol·kg⁻¹ DIC. In parallel, Li/Ca ratios increase from ∼4 to ∼20 μmol·mol⁻¹ as DIC decreases from 2.5 to 1.0 mmol·kg⁻¹ (Fig. 2, Table 3). For both δ⁷Li and Li/Ca the experimental data corresponding to the highest DIC value (3.0 mmol·kg⁻¹) falls off of these linear trends. We can rule out a potential contamination from the reagents used in the DIC experiments, for several reasons. First, we used the same reagents as for pH experiments, except HCO₃^–. We added HCO₃^– in order to obtain the required DIC value, but if this procedure would have resulted in an increase of solution Li content, then the Li/Ca ratio in the shells should be positively correlated with DIC. The actual trend is the opposite (see Fig. 3). Moreover, since the same HCO₃^– reagent was added for both low- and high-DIC experiments, the variation of δ⁷Li going from values below the seawater Li isotope composition to values higher than seawater δ⁷Li could not be explained by the addition of this reagent. Thus, the increase in the δ⁷Li and decrease in the Li/Ca ratios must be caused by the DIC increase and related mechanism of calcite precipitation.

Our cultures indicate the key role of the concentration of dissolved inorganic carbon on the Li isotope composition and on the Li/Ca ratio of Amphistegina tests. Figs.1c and 2 show that, for given T and pH values, at low DIC, the calcite Li/Ca ratio is high and the δ⁷Li value is low. As described previously, we first rule out a potential influence of the carbonate ion (CO₃^2–) concentration. The foraminifera grown at various pH with constant DIC experienced significant CO₃^2– variability, yet their Li/Ca and Li isotope compositions remained constant (Fig. 1a). The same argument can be made with respect to alkalinity: the range of alkalinities in the pH experiments was significant (0.4 mmol·kg⁻¹) and again the δ⁷Li and Li/Ca were similar, within uncertainties. One possibility to explain the dependence of Amphistegina Li/Ca and δ⁷Li on DIC concentrations is that it may be related to the physiology of calcification in perforated foraminifera (Bentov et al., 2009, Erez, 2003). In previous studies, it has been shown that seawater vacuoles provide the fluid from which the shell is precipitated (e.g., Erez, 2003). In order to precipitate CaCO₃ from these seawater vacuoles, the foraminifera have to concentrate DIC until the CO₃^2– content matches (at least partially) the Ca²⁺ concentration in the seawater vacuoles, which is between 10 and 11 mM. This internal carbon pool has been thoroughly described using ¹⁴C pulse-chase experiments with A. lobifera (ter Kuile and Erez, 1987, 1988; ter Kuile et al., 1989). In order to concentrate DIC in the vacuoles, the foraminifera elevate their pH and the DIC is increased by diffusion of CO_2(aq) from the cytosol into the vacuolated seawater (Bentov et al., 2009, Erez, 2003). Elevation of pH in the vacuoles and the calcifying fluid may be achieved through several ways. For example, Ca-ATPase has often been suggested as one potential candidate in corals (e.g., Al-Horani et al., 2003; McConnaughey, 1989; Venn et al., 2013). However, there are other possibilities for such proton transporters, which are protein complexes that transport protons against the concentration gradient (note that the cytosol in which the vacuoles are residing has a low pH of about 7.2–7.4). Usually such processes require ATP, as an energy source, or utilize the cellular electrochemical gradient to achieve the needed result. We suggest that in foraminifera, this proton transport is achieved by an involvement of a Na/proton exchanger (antiporter) or by another pathway in which Na (and Li) ions are transported into the vacuole while protons are transported out into the cytosol. We propose a coupling between Na and Li because Li can often replace Na during such trans-membrane processes, due to similar chemical behavior and charge (Parker, 1986). It is interesting to note the non-dependency of Li and δ⁷Li on the experimental temperature, suggesting that Li and Mg behaviors are decoupled, despite similar ionic radii (assuming that Mg/Ca and T are correlated). This indicates that the ion charge is an important parameter in biological activity (in contrast to abiotic reactions where Li⁺ and Mg²⁺ are often coupled; e.g., Huh et al., 2001; Vigier et al., 2008). Our observations may therefore be explained as follows: when the solution DIC is low, the activity of the Na/proton exchanger is intensive in order to elevate the pH, the alkalinity, and therefore to accumulate DIC in the vacuoles. While this is happening, Li follows Na and accumulates in the vacuoles. Under these conditions, the lighter isotope (⁶Li) may be preferentially transported into the vacuoles, which would explain the high Li/Ca ratio associated with the low δ⁷Li values (Fig. 3). When CaCO₃ is precipitated from modified seawater, it records its Li concentration and its δ⁷Li value. Further speculations on this mechanism are perhaps too early, but it explains well our experimental observations and is in good agreement with the models suggested for foraminiferal calcification (Bentov and Erez, 2005; Bentov et al., 2009; Erez, 2003).

When DIC is high, calcite δ⁷Li is also higher (higher than seawater value, 31‰) and Li/Ca is lower (Fig. 3). Published δ⁷Li values for core-top foraminifera tests higher than 31‰ (seawater) are rare, but exist (see, e.g., Hall et al., 2005; Kosler et al., 2001). Actually, these values are still consistent with a vital effect in favor of the heavy ⁷Li isotope, as underlined in section 3.1 and shown by T and pH experiments (Fig. 1a and b), compared to inorganic calcite (see section 3.1). Under high DIC conditions, we can speculate that there is no need to increase the DIC in order to calcify, and therefore, the high δ⁷Li observed in the Amphistegina tests cultured in these conditions likely reflects the isotope composition of the solution at the calcifying site. The reaction responsible for this increase may also be either kinetic or associated with a Rayleigh-type isotope fractionation occurring in the cytosol with preferential release of the light ⁶Li isotope, and decrease in Li/Ca ratio. A more refined biological study of the mechanisms occurring at high DIC is underway.

Amphistegina are symbiont-bearing foraminifera, belonging to the calcitic-radial perforate group, and thus are thought to precipitate their calcite using the same mechanisms as most planktonic foraminifera (Erez, 2003). This was illustrated by boron isotope ratios measured on cultured Amphistegina tests, using SIMS, which can be explained by a pH range similar to the one measured by pH microsensors in living planktonic Globigerinoides sacculifer (Jorgensen et al., 1985; Rollion-Bard and Erez, 2010). In order to test this for Li isotopes, we reported in Fig. 3 all the Li isotopes and concentration data published for various species of recent and core top planktonic foraminifera. The data plot along the trend defined by the Amphistegina cultured in this study. As a consequence, we consider that the trend between Amphistegina δ⁷Li and DIC may also be valid for the various planktonic foraminifera species studied thus far. It should therefore be possible to calculate the DIC at the time of the foraminifera test formation, using the linear regression of the experimental data points of Amphistegina (Fig. 1c). In addition, the DIC control evidenced in this study may partly explain the large published range of δ⁷Li measured for recent and sub-recent foraminifera (21.1–33.7‰, Hall et al., 2005; Hathorne and James, 2006; Marriott et al., 2004b; Misra and Froelich, 2012).

For short periods of time, such as for the last glacial/interglacial period, the ocean δ⁷Li has remained constant since the ocean Li residence time is larger than 1 Ma. However, foraminifera exhibit δ⁷Li variations during the last ∼40 ka (Hall et al., 2005). From 400 yr to 35.8 ka, the foraminifera δ⁷Li range is 30.4–33.7‰, which corresponds to a narrow DIC range (1.82–2.08 mmol·kg⁻¹), using the empirical relationship determined experimentally. For this period of time, the pCO₂ range is well known, since it is given by direct measurement of CO₂ trapped in air inclusions analyzed in the Vostok ice cores (Monnin et al., 2001). Using the calculated DIC and the following equation:

The 0–70 Ma record of δ⁷Li in foraminifera (Misra and Froelich, 2012) is also characterized by several periods of significant short-term variability (see Fig. 5b). This is particularly visible during the Miocene, between 11 Ma and 24 Ma (with, e.g., a 3.2‰ range of the δ⁷Li values between 14.25 Ma and 15 Ma). Since the Li residence time in the ocean is roughly 1 Ma, such rapid fluctuations cannot be explained by ocean variations due to changes in source or sink. Instead, our experimental results suggest that they can be related to changes of the Li isotope fractionation during foraminiferal growth, due to rapid variations of the DIC in seawater. We show that, for four critical periods, ranging between 16.5 Ma and 4.3 Ma, the foraminifera δ⁷Li and Li/Ca are linearly correlated and can be fitted by a similar slope as the one defined by the cultured Amphistegina and the core-top foraminifera (Fig. 3 and 5a). This strongly supports a DIC control on foraminiferal δ⁷Li at these times, which also correspond to critical periods of the Earth's history, i.e. the beginning of the Paleocene and the Mid-Miocene climatic optimum. The corresponding DIC variations range between 0.10 mmol·kg⁻¹ in the Pliocene up to 0.22 mmol·kg⁻¹ at the Miocene.