4.1 Correlation Between DU and the Carbonate System
 We used two different methods to manipulate the seawater carbonate chemistry: an acid/base manipulation (treatments A1–A4, see Table 1) and a pH-stable manipulation (treatments B1–B4, see Table 1). Since the carbonate system parameters covary differently in the two experimental approaches, it is possible, by exclusion, to reject certain parameters of the carbonate system as causes for the observed changes in DU (Table 1). In the acid/base manipulation DU displays a positive correlation with pCO2, whereas the correlation of DU and pCO2 is negative in the pH-stable manipulation. Therefore, pCO2 cannot be the parameter of the carbonate system causing a change in DU. The change in DU under constant pH (pH-stable manipulation) was almost a factor of 10 larger than in the acid/base manipulation, where pH covaried. Consequently, pH can be excluded as a controlling factor and the negative correlation of DU with pH in the acid/base manipulation must be regarded as inherent to the carbonate system and not causal.
 In the pH-stable manipulation experiments, TA, DIC, , , and Ω correlate negatively with DU over a wide range of values (Table 1 and Figure 2 closed squares; note that the correlation of DU with Ω has not been plotted, since [Ca2+] was kept constant and consequently the distribution of Ω is essentially that of ; Figure 2a). By contrast, the ranges covered in the acid/base manipulation are, with the exception of and Ω, much smaller (Table 1). In the case of DIC (acid/base manipulation), most DU are similar. Nevertheless, when DU is plotted versus DIC, a conspicuous clustering of values can be seen, namely the DU of treatments A3 and A4 are similar, and so are the DU of treatments A1 and A2 (small inset in Figure 2c). The same clustering is obvious when DU is plotted versus and TA (insets in Figures 2b and 2d). Combining values of the acid/base and the pH-stable manipulation in one plot (Figures 2b–2d) might, especially in the case of TA (Figure 2d), suggest that TA causes the change in DU and that the curious clustering of treatments A3 and A4, and A1 and A2, respectively, simply reflects the transition of the curve from a steep to a shallow slope. If DIC or was the controlling factor, DU values should be more or less identical in the pH-stable manipulation, given the small range in DIC and . Since there is no reason why, given identical DIC/, there should be such a distinct cluster pattern, and DIC and cannot be the parameter affecting DU. The cluster pattern, however, is absent when plotting DU versus (Figure 2a). Hence, based on the correlations only, or Ω are likely candidates to be instrumental in changing DU, but leave open the possibility that TA might still be involved. It has to be noted, that Ca2+ was kept constant in all treatments and foraminifera do not respond to Ω as such, but to the concentrations of Ca2+ and (which also holds true for TA) [Dueñas-Bohórquez et al., 2011; Raitzsch et al., 2010]. Consequently, the correlation between DU and Ω is only caused by the concentration change in carbonate ions, leaving carbonate ions as the only candidate affecting foraminiferal U/Ca. Furthermore, we cannot exclude the possibility of parameters, such as TA and/or pH, exerting a modulating influence on the obtained correlations. While the modulating influence cannot be unambiguously identified using the data set presented here, we will point out that the correlation can be explained by a sole influence of carbonate ion concentration on DU. Hence, a modulating influence of other parameters, although possible, is not needed in order to interpret the changes in DU.
 The effect of on foraminiferal U incorporation could be explained in terms of uranium speciation in seawater. Uranium easily complexes with carbonate ions, and speciation thus strongly depends on of the seawater (Figure 3). With increasing , the percentage of the sum of the different carbonate complexes [UO2(CO3)(aq)], , and increases, whereas the percentage of the sum of the free forms and [UO2OH+] decreases. This change in speciation is not linear and particularly prominent below ~200 µmol/kg-sw (Figure 3). Interestingly, the correlation of foraminiferal DU with is also not linear, but exponential (see Figure 2a), with the largest change in DU µmol/kg-sw at the lower range of used here, i.e., below ~200 µmol/kg-sw (Table 1 and Figure 2). This matches the increase in and [UO2OH+] at low . We hypothesize that the free forms ( and [UO2OH+]) are more readily taken up by Ammonia sp. than the carbonate complexes. This speculation would explain the observed dependency of DU on . In support of this hypothesis, it was reported that the bioavailability of U (i.e., its ability to bind to or traverse the cell surface) in green algae increases with decreasing [Fortin et al., 2004; Markich, 2002]. The latter authors attribute this effect to the fact that primarily the free forms of U (especially ) are taken up by the cells. In analogy, we speculate that the free forms of U can cross the cell membrane of Ammonia sp. more easily than the carbonate complexes can. This would imply that U is taken up via transmembrane transport during chamber formation, which therewith would be a major pathway of ion transport for chamber formation in Ammonia sp. The latter assumption was also put forth in the context of proton transport [Glas et al., 2012]. We are aware that ion transport in foraminifera is usually assumed to be endocytosis mediated [Erez, 2003], but the latter hypothesis is based on experiments with a different species, and there might be species-specific differences in transport mechanism. We will point out that our explanation of the change in DU with seawater is consistent with a constancy of U fractionation during calcite precipitation. This is advantageous, because pH homeostasis in the calcifying fluid most likely leads to a constant U speciation which would be decoupled from seawater U speciation.
4.2 Paleoceanographic Implications
 Previous studies [e.g., Russell et al., 2004] reported a correlation between foraminiferal U/Ca and carbonate chemistry of seawater. While these studies attributed the effect to carbonate ion concentration/calcite saturation state, this inference remained conjectural, because in all available data sets, the parameters of the carbonate system covaried, rendering it impossible to tell, e.g., pH from carbonate ion effects. The reported U/Cacc values for benthic [Raitzsch et al., 2011] and planktic foraminifera [Russell et al., 2004] were 2–10 times lower than the ones determined by us for the same range of (80–110 µmol U/kg seawater). The difference in calcitic U/Cacc may be the result of species-specific fractionation against U during calcification and underscores the need for species-based calibrations when applying U/Cacc to reconstruct past . However, it needs to be stressed that the species used here, Ammonia sp., is not commonly used in paleoceanographic studies, due to its shallow-water benthic habitat. Nevertheless, its abundance, easy accessibility, the relatively common asexual reproduction, and the tolerance of a broad range of environmental parameters make it a suitable candidate when determining basic foraminiferal responses. Applying the here-introduced experimental protocol of decoupling C-system parameters to more relevant species in terms of paleoceanography is a step that should be undertaken in the future to analyze the DU- relationship further. While the slope of this relationship apparently is species specific, it is likely that the causal basis for this relationship is not. Our results therefore put the application of U/Cacc as a proxy on a firm footing.
 Even if not primarily of interest for paleoceanographic studies, a few properties of the correlation found here shall be given to facilitate comparability between different studies. The large range of applied in our culture study supports an exponential relation between carbonate ion concentration and calcitic U/Cacc as previously proposed by Russell et al. :
 Based on this calibration, we can infer that a decrease of 100 µmol/kg-sw in carbonate ion concentration from 300 to 200 µmol/kg-sw, as anticipated for a transition from full glacial to interglacial conditions, would be expected to result in an increase of 54% in foraminiferal U/Cacc. With our analytical approach, those changes can be quantified within the 95% confidence intervals. This sensitivity is approximately twice as high as that reported for two planktic species, Orbulina universa and Globigerinoides sacculifer by Russell et al. .