Calibration of δ 18O of cultured benthic foraminiferal calcite as a function of temperature

. The geochemical composition of deep-sea benthic foraminiferal calcite is widely used to reconstruct sea ﬂoor paleoenvironments. The calibration of the applied proxy methods has until now been based on ﬁeld observations in complex natural ecosystems where multiple factors are interfering. However, laboratory experiments with stable physico-chemical conditions appear to be the ideal way to evaluate the inﬂuence of a single parameter. In this paper, we present the oxygen isotopic composition of deep-sea benthic foraminiferal shells entirely calciﬁed under controlled experimental conditions over a large temperature range (4 to 19 ◦ C). The new laboratory protocols developed for this study allowed us to produce large quantities of shells in stable conditions, so that also the shell size effect could be investigated. It appears that when considering a narrow test size range, the curve describing the temperature dependency of δ 18 O in Bulimina marginata is parallel to the thermody-namically determined curve observed in inorganically pre-cipitated calcite ( − 0.22‰ ◦ C − 1 ) . This observation validates the use of δ 18 O of this benthic species in paleoceanographi-cal studies. Over the studied size range (50 to 300 µm), the effect of test size was 0.0014‰ µm − 1 , conﬁrming previous suggestions of a substantial test size effect on δ 18 O of benthic foraminifera. This study opens new perspectives for future proxy calibrations in laboratory set-ups with deep-sea benthic foraminifera (e.g. quantiﬁcation of the inﬂuence of the carbonate chemistry).


Introduction
Stable oxygen isotopes of carbonate microfossils are one of the most widely used tools in paleoceanography. The temperature dependency of oxygen isotope fractionation has previously been quantified on the basis of inorganically precipitated calcite (Urey, 1947;McCrea, 1950;O'Neil et al., 1969;Kim and O'Neil, 1997), and has been verified for living organisms in field and/or laboratory cultures of corals (Reynaud-Vaganay et al., 1999), molluscs (Epstein et al., 1953) and planktonic foraminifera (Erez and Luz, 1983;Bouvier-Soumagnac and Duplessy, 1985;Bouvier-Soumagnac et al., 1986;Bemis et al., 1998). For benthic foraminifera, until now, all existing temperature calibrations are based on core top material. On the sea floor, not only temperature and the isotopic composition of the seawater influence the 18 O/ 16 O composition of foraminiferal calcite, but also other factors, such as the carbonate ion effect (Spero et al., 1997;Zeebe, 1999;Rathmann and Kunhert, 2008), vital effects (Duplessy et al., 1970) and diagenetic processes may strongly influence the δ 18 O of carbonate microfossils. Since many of these factors co-vary in the natural environment, only culture experiments can precisely reveal the influence of a single parameter, such as temperature.
Several laboratory studies have been performed to study the oxygen isotopic fractionation in planktonic and shallow water benthic foraminifera (e.g. Erez and Luz, 1983;Bouvier-Soumagnac and Duplessy, 1986;Chandler et al., 1996;Spero and Lea, 1996;Spero et al., 1997;Bemis et al., 1998). However, experiments with deep-sea benthic foraminifera are very scarce (Wilson-Finelli et al., 1998;Mc-Corkle et al., 2008;Filipsson et al., 2010). Actually, the growth of deep-sea benthic foraminifera takes much longer Published by Copernicus Publications on behalf of the European Geosciences Union. than for planktonic foraminifera so that the experiments in stable conditions have to last for periods extending to several months. However, benthic foraminifera present the indisputable advantage that they can reproduce in the laboratory McCorkle et al., 2008;Barras et al., 2009;Filipsson et al., 2010). It is therefore possible to measure the isotopic composition of shells entirely calcified under controlled conditions.
In order to obtain the results presented in this paper, we developed new laboratory protocols to produce large quantities of Bulimina marginata shells under controlled and stable conditions and over a large range of temperatures (4-19 • C), making it possible to investigate the influence of temperature on the δ 18 O of deep-sea benthic foraminiferal calcite. The large amount of foraminiferal shells produced allowed us also to investigate the effect of test size on isotopic fractionation.

Experimental protocols
For this study, adult specimens of B. marginata (nonsymbiont-baring benthic species) sampled in the Bay of Biscay at 450 and 650 m depth, were used in different experiments to obtain reproduction and subsequent growth of the juveniles (detailed protocol and data on reproduction and growth rates of B. marginata in Barras et al., 2009). Before their introduction in the experiments, adult specimens were labelled using a calcein-tagging method  in order to distinguish specimens that totally calcified their shells in our controlled experiments (not fluorescent) from the adults initially added (partly fluorescent). Two different laboratory setups were used to obtain reproduction and growth of B. marginata under stable physicochemical conditions: (1) a closed system (CS I and CS II ), with 25 l microfiltrated (0.45 µm) natural seawater circulating through a reservoir and different experiment bottles, and (2) a Petri dish system (PD) where half of the seawater was renewed twice per week. Between 30 and 190 adult specimens of B. marginata were introduced in each experiment, which were regularly fed with fresh Phaeodactylum tricornutum diatoms. In all experiments, which lasted from 43 to 108 days, we obtained production and growth of juveniles of Bulimina marginata. Therefore, the isotopic composition of foraminiferal calcite was measured on tests of Bulimina marginata entirely calcified under controlled laboratory conditions (not fluorescent specimens).
Temperature was recorded inside the incubators (standard deviations range from 0.1 to 1.1 • C depending on the incubator). Culture water samples were collected every 3 to 7 days to verify the stability of salinity (35.8 ± 0.1), δ 18 O seawater (0.6 ± 0.1‰ vs. SMOW), pH and alkalinity, and the absence of significant evaporation (details in Table 1).
The carbonate chemistry was stable, and similar in experiments CS I and PD (7.94 ± 0.05 for pH, NBS-scale, and 2453 ± 34 µmol l −1 for alkalinity; Table 1). However, an episodic peak of high alkalinity and pH was recorded during the first week of the PD experiments, which is probably irrelevant for the geochemical composition of the newly formed shells, since B. marginata only reproduces after several weeks of incubation (Barras et al., 2009). For CS II , a gradual decrease of pH by 0.3 units between the start and the end (average of 7.79 ± 0.09, NBS-scale) occurred in the six experiments, whereas alkalinity remained stable, and similar to the other systems (2523 ± 14 µmol l −1 ) ( Table 1). In the hypothetical case of linear growth of the shells during the experimental period, this gradual decrease of pH by 0.3 units could theoretically result in a positive δ 18 O shift of about 0.15‰ of the newly formed foraminifera, due to the carbonate ion effect (Zeebe, 1999). However, benthic foraminifera do not have a uniform growth, chamber addition being faster during early ontogenetic stages (Bradshaw, 1957;Stouff et al., 1999;Barras et al., 2009).

Analytical procedures
Oxygen isotopic analyses were performed on 10 to 150 entire specimens of B. marginata. In order to study the ontogenetic effect on the 18 O/ 16 O ratios of the shells of deep-sea benthic foraminifera, specimens were separated into different size fractions (length measurements with microscale). Observation of the shells under the stereomicroscope showed that they were transparent with no mineral adhesives visible. Therefore specimens were only rinsed with deionised water before analysis. All tests were then roasted at 380 • C during 45 min to remove all organic matter. The 18 O/ 16 O ratio of foraminiferal calcite was measured with Isoprime and VG-Optima mass-spectrometers. Results are expressed as δ=((R sample −R standard )/R standard )·1000, where R is the 18 O/ 16 O isotopic ratio. The analytical precision of the δ 18 O analyses is ± 0.05‰ relative to the VPDB (Vienna Pee Dee Belemnite) standard.
Seawater δ 18 O (δ 18 O w ) was measured by equilibrating water samples with pure CO 2 which was subsequently analysed with a Finnigan Mass spectrometer. The analytical precision of the δ 18 O analyses is ± 0.05‰ relative to the VSMOW (Vienna Standard Mean Ocean Water) standard.
In order to determine the relationship between temperature and δ 18 O of B. marginata shells, we calculated least square regressions of the isotopic difference between foraminiferal shell and seawater (δ 18 O f − δ 18 O w ) versus temperature. The δ 18 O w data were converted from VSMOW to VPDB by subtracting 0.27‰ (Hut, 1987). We applied linear regression to our data sets since this provided equally good fits as quadratic regression. The choice of linear or quadratic equations was discussed by Bemis et al. (1998). If we consider for example the paleotemperature equations of Kim and O'Neil (1969)  of our experiments (4-19 • C), we obtain a maximum temperature offset of 0.2 • C compared to the quadratic equation. This variation corresponds to a δ 18 O f bias of 0.05‰ which is equivalent to the precision of the mass-spectrometer. The coefficient of determination (R 2 ) and the standard errors on the slope and intercept are indicated for each equation.

Influence of temperature on the δ 18 O of cultured foraminifera
Knowing that shell size may have an effect on isotope ratio in foraminifera (Spero and Lea, 1996;Bemis et al., 1998;Elderfield et al., 2002;Schmiedl et al., 2004) Table 1).
The 18 O/ 16 O composition of B. marginata appears similar for the 3 experimental protocols (CS I , CS II and PD) for a given temperature and given size fraction (Fig. 1). For the ≤150 and 150-200 µm size fractions, where sufficient data are available, we used Lin's test (Lin, 1989) to estimate the concordance of the regression lines for the three systems. For all cases, we obtained concordance correlation coefficients above 0.990, confirming the high degree of similarity of the data obtained with the three systems. Therefore, we conclude that the pH decrease in CS II did not cause a significant shift of the δ 18 O of foraminifera calcified in these experiments. Since the δ 18 O of B. marginata appears to be independent of the applied protocol, in the following text we will no longer distinguish the three experimental set-ups. The linear equations which best describe the relationship between temperature and δ 18 O of foraminiferal tests entirely calcified under controlled laboratory conditions are, for the four different size fractions (Fig. 1):  T Equations (1, 2 and 3) exhibit similar slopes considering the standard errors on the slope estimates. For these three size fractions, the relative influence of temperature on the oxygen isotopic composition of B. marginata is −0.22‰ • C −1 . For the > 250 µm size fraction, the linear regression between δ 18 O f − δ 18 O w and temperature presents a steeper slope (Eq. 4). However, the linear regression for this size fraction is less well defined than that obtained for the smaller size fractions, since data are available only for three different temperatures and only few individuals attained a size larger than 250 µm. Further experimental work is needed to refine this (Eq. 4), which we will not consider in the remaining part of this paper.

Influence of shell size on the δ 18 O of cultured foraminifera
Interestingly, there is an increase in the intercept values with increasing size fraction (  presented as a function of test size for the three temperatures for which we had a sufficient amount of different size fractions to obtain a reliable regression equation (p < 0.01). Figure 2 shows that at 10.2, 12.7 and 14.7 • C, the δ 18 O of the foraminiferal tests increases by 0.0012-0.0022‰ µm −1 , with determination coefficients (R 2 ) between 0.4 and 0.7 (p < 0.01). The calcification rate plays an important role in the fractionation of the organisms since higher growth rates will result in a more depleted δ 18 O and δ 13 C (Mc-Connaughey, 1989a, 1989b. This is due to a "kinetic effect" which corresponds to the discrimination against heavy C and O isotopes during hydration (CO 2 + H 2 O → H 2 CO 3 ) and hydroxylation (CO 2 + OH − → HCO − 3 ) of CO 2 . Because younger foraminifera calcify faster (Berger et al., 1978), they may not attain equilibrium in the calcification reservoir before crystallisation, which would result in the production of more negative δ 18 O and δ 13 C values. Such a possibility was earlier proposed by Turner (1982). For benthic foraminifera, ecological experiments tend to prove that growth of specimens is not uniform and chambers addition is faster during the first ontogenetic stages (Bradshaw, 1957(Bradshaw, , 1961Hemleben and Kitazato, 1995;Stouff et al., 1999). This has also been observed for Bulimina marginata (Barras et al., 2009).
The influence of size on oxygen isotopic composition is well established for planktonic foraminifera (Spero and Lea, 1996;Bemis et al., 1998;Elderfield et al., 2002), whereas previous field-based studies of size-dependent trends in benthic foraminiferal isotopic values have been inconclusive (Vincent et al., 1981;Dunbar and Wefer, 1984;Grossman, 1987;Corliss et al., 2002). Generally, in these studies, benthic foraminifera do not show a significant change in δ 18 O with size. However, some authors observed an ontogenetic effect on the oxygen isotopic fractionation of Bulimina aculeata/marginata shells obtained in laboratory experiments (McCorkle et al., 2008;Filipsson et al., 2010) and living and dead Uvigerina mediterranea from the western Mediterranean Sea (Schmiedl et al., 2004). Schmiedl et al. (2004) found a 0.3-0.4‰ δ 18 O enrichment over a size range of 175 to 1250 µm. This enrichment was particularly important in the early growth stages (100-300 µm) and became weaker for adult forms, which might be explained by the decreasing metabolic rates towards more adult life stages. If we compare the slope of their logarithmic correlation equation for these younger stages (the size fraction we studied) with our data, their δ 18 O versus test size curve has an average slope of about 0.001‰ µm −1 which is similar to the size effect found in our experiments. Even if adult specimens of B. marginata are smaller than adult specimens of U. mediterranea, it is probable that the specimens measured in our experiments were not large enough to reach the stable isotopic composition typical of larger specimens, as observed for U. mediterranea (Schmiedl et al., 2004). Either our specimens were still growing when the experiments were stopped, or they died before attaining the "adult" stage. It would be useful in future experiments to grow living B. marginata during longer time than in our experiments and try to obtain larger size fractions.
On the basis of all our 83 δ 18 O measurements performed on specimens of B. marginata which totally calcified under controlled conditions (Table 1), we applied a multiple regression that takes into account δ 18 O of the shells, calcification temperature (4-19 • C) as well as test size (50-300 µm). According to this multiple regression, the averaged size effect on δ 18 O composition of B. marginata is 0.0014‰ µm −1 . It appears therefore that an ontogenetic effect on oxygen isotope fractionation exists also for benthic foraminifera and cannot be neglected in paleoceanographic studies. Since the regression lines of δ 18 O f −δ 18 O w versus test size are more or less parallel for the tested temperatures, we conclude that the mechanism responsible for this small ontogenetic effect is independent of calcification temperature. We recommend performing measurements in a size range not larger than 50 µm to fully exploit the 0.07‰ accuracy of mass-spectrometric analyses.

Comparison with equilibrium calcite as defined by
Kim and O'Neil (1997) Fig. 3. Comparison of our experimental calibration equation with the theoretical equation for equilibrium calcite of Kim and O'Neil (1997). The brown, blue and green lines represent the calibration equations of cultured B. marginata from ≤150, 150-200 and 200-250 µm size fractions, respectively. The quadratic equation derived from Kim and O'Neil (1997) relationship is represented by the red line.
our study, and measurements were performed on inorganic calcite, free of vital effects. The three experimental regression curves we determined for size fractions smaller than 250 µm exhibit similar slopes as the least square regression line applied to the quadratic relationship of Kim and O'Neil (1997) over the studied temperature range (Fig. 3). Therefore, the influence of temperature on the δ 18 O of calcite is similar, and independent of test size. Furthermore, the offsets of the foraminiferal curves with respect to the inorganic carbonate curve are very small. Regression lines (2) and (3) fit well with the Kim and O'Neil (1997)

Conclusions
The new protocols developed for this study allowed us to obtain reproduction and calcification of the deep-sea benthic foraminifer Bulimina marginata under controlled conditions at 12 different temperatures between 4 and 19 • C. In general, a 1 • C decrease in calcification temperature increases the δ 18 O of Bulimina marginata by +0.22‰, irrespective of the size fraction and culture setup considered. This effect is similar to the thermodynamical effect observed for inorganic calcite. However, our data show a small but conspicuous ontogenetic effect on δ 18 O values of about 0.0014‰ µm −1 that should be taken into account in order to produce accurate paleoclimatic reconstructions. Bulimina marginata specimens with a test length between 150 and 250 µm calcify very close to the equilibrium calcite as defined by Kim and O'Neil (1997). Finally, these experiments, leading to reliable data, proved that the foraminiferal treatment protocols developed for this study could be applied in future studies to investigate the impact of other physico-chemical parameters (salinity, carbonate chemistry. . . ) on benthic foraminiferal shell composition (isotopes, trace metals. . . ).