Influence of temperature and CO 2 on the strontium and magnesium composition of coccolithophore calcite

Marine calcareous sediments provide a fundamental basis for palaeoceanographic studies aiming to reconstruct past oceanic conditions and understand key biogeochemical element cycles. Calcifying unicellular phytoplankton (coccolithophores) are a major contributor to both carbon and calcium cycling by photosynthesis and the production of calcite (coccoliths) in the euphotic zone, and the subsequent long-term deposition and burial into marine sediments. Here we present data from controlled laboratory experiments on four coccolithophore species and elucidate the relation between the divalent cation (Sr, Mg and Ca) partitioning in coccoliths and cellular physiology (growth, calcification and photosynthesis). Coccolithophores were cultured under different seawater temperature and carbonate chemistry conditions. The partition coefficient of strontium ( DSr) was positively correlated with both carbon dioxide ( pCO2) and temperature but displayed no coherent relation to particulate organic and inorganic carbon production rates. Furthermore, DSr correlated positively with cellular growth rates when driven by temperature but no correlation was present when changes in growth rates were pCO2-induced. Our results demonstrate the complex interaction between environmental forcing and physiological control on the strontium partitioning in coccolithophore calcite and challenge interpretations of the coccolith Sr / Ca ratio from highpCO2 environments (e.g. Palaeocene–Eocene thermal maximum). The partition coefficient of magnesium ( DMg) displayed speciesspecific differences and elevated values under nutrient limitation. No conclusive correlation between coccolith DMg and temperature was observed but pCO2 induced a rising trend in coccolithDMg. Interestingly, the best correlation was found between coccolithDMg and chlorophylla production, suggesting that chlorophyll a and calcite associated Mg originate from the same intracellular pool. These and previous findings indicate that Mg is transported into the cell and to the site of calcification via different pathways than Ca and Sr. Consequently, the coccolith Mg / Ca ratio should be decoupled from the seawater Mg / Ca ratio. This study gives an extended insight into the driving factors influencing the coccolith Mg / Ca ratio and should be considered for future palaeoproxy calibrations. Published by Copernicus Publications on behalf of the European Geosciences Union. 1066 M. N. Müller et al.: Coccolith Sr / Ca and Mg / Ca ratios


Introduction
Coccolithophores, a key functional phytoplankton group, evolved about 225 Ma (million years ago), and their calcitic plates (coccoliths) that cover the cell are found in the sediment record since then (Bown et al., 2004).Over their geological history, coccolithophores have experienced various changes in Earth's climate and the geochemical composition of their intracellularly produced coccoliths has been the focus of numerous studies reconstructing past oceanographic and environmental conditions (Stoll and Ziveri, 2004;Erba, 2006;Ziveri et al., 2012).Since the industrial revolution, however, coccolithophores have been exposed to rapidly changing oceanic conditions including an increase in seasurface temperature and shifting seawater carbonate chemistry speciation due to anthropogenically released carbon dioxide (Boyd and Hutchins, 2012).Understanding their physiological response to changing carbonate chemistry and temperature, as well as the underlying mechanisms responsible for the chemical signature of coccolith calcite, is essential to validate and apply coccolithophore-based palaeoproxies (Ziveri et al., 2012).The distribution and partitioning of divalent cations, replacing the calcium ion in the calcite lattice, are driven by a combination of thermodynamic precipitation kinetics and cellular physiology (Stoll et al., 2012;Langer et al., 2006;Müller et al., 2011).The physiological response of coccolithophores to changing carbonate chemistry speciation and temperature is relatively well described (Riebesell and Tortell, 2011;Raven and Crawfurd, 2012); however, associated changes in coccolith elemental partitioning and composition (e.g.Sr / Ca and Mg / Ca) are less well understood (Stoll and Ziveri, 2004;Stoll et al., 2012).
Coccolithophore productivity and growth has been hypothesised to primarily control coccolith Sr / Ca ratios with a secondary influence of temperature and the seawater Sr / Ca ratio (Stoll and Schrag, 2000;Stoll and Ziveri, 2004;Stoll et al., 2007a).Cellular concepts were developed to model and describe the Sr partitioning in coccolithophore calcite (Langer et al., 2006(Langer et al., , 2009)), based on the assumption that Ca and Sr are transported via the same pathways which is supported by experimental evidence (Müller et al., 2011).The effect of carbonate chemistry speciation on the coccolith Sr / Ca ratio is unknown as well as the associated influence of coccolithophore growth affected by seawater pH and CO 2 availability.
Coccolithophores precipitate low-Mg calcite (< 1 % mol) despite that Mg is the most abundant divalent cation in plant cells and its concentration in the cytosol is about 20 000 times higher than cytosolic-free Ca concentration (Brownlee et al., 1995;Bose et al., 2011).The transport of Mg in coccolithophores presumably differs from Sr and Ca and it is likely associated with specific ion transporters and channels (Mackinder et al., 2010;Holtz et al., 2013) as in bacteria and higher plants (Legong et al., 2001).A relation between the coccolithophore Mg / Ca ratio and temperature is still under discussion and complicated by the nontrivial removal of organically bound Mg (Blanco-Ameijeiras et al., 2012).
Potential changes in the partitioning of Sr and Mg, and thus in the Sr / Ca and Mg / Ca ratios of calcareous sediments, may alter their biogeochemical cycling over geological timescales because biogenic calcification and the subsequent burial of calcium carbonate represents a major elemental export process from the ocean to the lithosphere.Furthermore, the dissolution dynamics of carbonate-rich sediments are, to a certain degree, controlled by their Mg / Ca ratio (Andersson et al., 2008;Morse et al., 2006).
Here we present the partitioning coefficients for Sr and Mg and the related coccolith Sr / Ca and Mg / Ca ratios from laboratory culture experiments in which temperature and carbonate chemistry of the growth medium were tightly controlled and monitored.Calcite samples were cleaned with a newly developed organic matter removal procedure (Blanco-Ameijeiras et al., 2012).Our findings should be considered when assessing possible changes in the biogeochemical cycling of Ca, Sr and Mg due to ocean warming and acidification and, furthermore, when applying palaeoproxies based on Sr / Ca and Mg / Ca coccolith ratios.

Methods
The present study combines results from four independent culture experiments conducted with Emiliania huxleyi (Exp. 1 and 2), Coccolithus braarudii (Exp. 1 and 3), Gephyrocapsa oceanica (Exp.4) and Calcidiscus quadriperforatus (Exp.1).All experiments were conducted within a temperature range from 10 to 25 • C. The carbonate system was manipulated to either keeping pCO 2 constant (Exp. 1) or establishing a pCO 2 gradient ranging from ≈ 25 to 3520 µatm (Exp.2, 3 and 4).A summary of the seawater conditions (temperature and carbonate system) and abiotic parameters (light intensity and day length) are given in the Supplementary material.Additional information on Exp. 3 and 4 is provided in Krug et al. (2011) and Sett et al. (2014), respectively.

Culture conditions
Coccolithus braarudii RCC-1200, Calcidiscus quadriperforatus RCC-1135 and Gephyrocapsa oceanica RCC-1303 were obtained from the Roscoff Culture Collection.Emiliania huxleyi was isolated in 2005 during the PeECE III mesocosm study in the Raune Fjord (Norway) by M. N. Müller.Cultures were maintained in 0.20 µm filtered North Sea water (NSW) with a salinity of 34 except G. oceanica which was cultured in artificial seawater (ASW) with a salinity of 35.Macro-and micronutrients were added according to a modified f/2 media (Guillard, 1975) with final concentrations assuring nutrient replete conditions in all experiments (nitrate ≥ 64 µmol L −1 , phosphate concentration).Cultures received a daily illumination between 130 and 180 µmol photons m −2 s −1 under a light : dark cycle of 14:10 h (Exp. 1 and 2) or 16:8 h (Exp.3 and 4).

Experimental set-up
Temperature was controlled in growth cabinets (Rumed, series 1000, deviation of ±0.5 • C) and pCO 2 was adjusted in the NSW experiments (Exp. 1, 2 and 3) by addition of HCl or NaOH, keeping the dissolved inorganic carbon concentration (DIC) constant (Exp.1: 2273 ± 42 µmol kg −1 , n = 6; Exp.2: 2157 ± 18 µmol kg −1 , n = 10; Exp.3: 2105 ± 20 µmol kg −1 , n = 15) while changing the total alkalinity (A T ).The carbonate system of the ASW in Exp. 4 was adjusted by keeping A T constant (2311 ± 56 µmol kg −1 , n = 24) and varying DIC (addition of HCl and Na 2 CO 3 ).Precultures were kept in exponential growth and acclimated to the experimental conditions over 7-20 generations.Acclimated experimental cultures were incubated in triplicates (Exp. 1 and 3), duplicates (Exp.2) or without replicates (Exp.4) in acid cleaned (10 % HCl) and autoclaved polycarbonate bottles at target conditions.In all experiments, cultures were inoculated to low cell densities and allowed to grow exponentially for 7-10 generations, corresponding to a maximum DIC consumption of 12 % (Exp. 1 and 3) or 7 % (Exp. 2 and 4).Growth media (adjusted to the target condition) was sampled for DIC, A T and seawater elemental composition.At the termination of the experiment (2-3 h after the onset of light), samples were taken for DIC, A T , seawater ion composition, cell number, total particulate carbon and nitrogen (TPC and TPN), particulate organic carbon (POC), chlorophyll a (Chl a) and coccolith elemental composition.Sample pellets for coccolith calcite of G. oceanica (Exp.4) were sampled 2-4 days after samples for physiological parameters were taken.Thus, the coccolith geochemistry data of G. oceanica are not directly related to the measured seawater chemistry and physiological parameters.This issue will be considered and discussed later on (Sect.4.1).

Carbonate chemistry
The carbonate system was monitored by DIC and A T measurements at the start and the end of the experiments.DIC was analysed after Stoll et al. (2001), using an automated segmented-flow analyser (Quaatro) equipped with an autosampler (10 µmol kg −1 accuracy and 5 µmol kg −1 precision).Total alkalinity was measured by the potentiometric titration method after Dickson et al. (2003) with an accuracy and precision of 24 and 3.5 µmol kg −1 , respectively.DIC and A T were calibrated and measured against certified reference material supplied by A. Dickson (Scripps, La Jolla, USA).Carbonate system parameters were calculated from temperature, salinity, DIC and A T (mean values from the start and end of experiments) using CO2SYS (version 1.05 by E. Lewis and D. W. R. Wallace), with the stoichiometric equilibrium constants for carbonic acid given in Roy et al. (1993).

Cellular growth rates
Cell densities were determined with a Coulter Counter (Z Series).Samples were measured three times and the mean was used to calculate the growth rate µ (d −1 ) as where c 0 and c 1 denote the number of cells at the start (t 0 ) and end (t 1 ) of the incubation period (expressed in days).

Production rates of particulate organic and inorganic carbon, total nitrogen and chlorophyll a
Three sub-samples were taken for TPC/TPN, POC and Chl a from each treatment bottle (Chl a was not sampled in Exp.3), filtrated onto precombusted GF/F (glass fibre) filters (500 • C for 6 h) and stored frozen at −20 • C. For POC analysis, filters were fumed over 37 % HCl for 2 h to remove all inorganic carbon and dried overnight at 60 • C. Filters for TPC / TPN and POC were measured on an Euro EA Elemental Analyser (Sharp, 1974).Particulate inorganic carbon was calculated from the difference of TPC and POC.Until analysis, Chl a filters were kept in the dark to avoid photo-oxidation and measurements were performed using a fluorimetric method after Welschmeyer (1994).Production rates of particulate material (PM) were calculated as where PM can be POC, PIC, TPN or Chl a.

Coccolith geochemistry
After samples for the physiological parameters were taken (Sect.2.2), the remaining culture medium (replicates of each experiment were pooled; Exp.1-3) was centrifuged at 14 000 rpm for 10 min.The supernatant was discarded and the sample pellet was dried at 60 Mg / Ca and Sr / Ca ratios of the coccoliths and of the ASW (Exp.4) were analysed using matrix-matched standards on a simultaneous dual ICP-AES (inductively coupled plasmaatomic emission spectrometry) (Thermo ICAP DUO 6300 at the University of Oviedo).Standards were prepared from single element ICP standards (CPI Corporation) diluted in the same acid matrix as the samples.The calibration standards covered the Mg / Ca and Sr / Ca range in the calcite samples.Elemental ratios are reported from measurements made in radial detection mode for Sr at 421 nm, Mg at 279.5 nm, Fe at 259 nm, Ca at 315 nm and P at 177 nm.Calibration was conducted offline using the intensity ratio method described by de Villiers et al. (2002).In this case, three standards were prepared with constant Ca concentrations and variable Sr / Ca and Mg / Ca ratios of 0.75-4 mmol mol −1 .Aliquots from this set of standards were uniformly diluted to provide curves for various Ca concentrations to match sample concentrations.Sr / Ca internal reproducibility was better than 0.02 mmol mol −1 , based on replicate analyses of the same sample dilutions.Mg / Ca internal reproducibility was better than 0.8 % of the measured ratio.
Coccolith P / Ca and Fe / Ca ratios are presented along with the Mg / Ca ratios as an indicator for Mg cleaning efficiency.Outlier detection, using the interquartile range (IQR), was performed on P / Ca and Fe / Ca ratios to determine non-sufficient Mg removal and the corresponding Mg / Ca ratios were removed from the data set.Sr / Ca ratio measurements were not excluded from the data set because Sr contamination from seawater and organic matter represents a minor issue for coccolith Sr / Ca ratio measurements (Blanco-Ameijeiras et al., 2012).Average values for P / Ca and Fe / Ca ratios (excluding outliers) were 1.36 ± 0.73 and 0.82 ± 1.19 mmol mol −1 (±1 SD; standard deviation), respectively.
Elemental composition of the NSW (Exp.1-3) was determined via Jobin Yvon JY 170 Ultratrace series ICP-OES (optical emission spectroscopy) at GEOMAR.The internal error for each analysis is typically < 1.2 % (±2 RSD; relative SD) and the external reproducibility determined by repeat analysis of IAPSO (International Association for the Physical Sciences of Oceans) seawater standard is ±1.5 %, ±0.7 % and ±0.3 % (±2 RSD, n = 3) for Ca, Mg and Sr, respectively.The partition coefficient of the trace metal (Sr and Mg), D Tr , between calcite (c) and seawater (s) was calculated as (3)

Results
At ambient pCO 2 levels, elevated temperature enhanced growth rates and PIC prod in all tested coccolithophore species.The POC prod was positively affected by temperature except for E. huxleyi, decreasing from 10 to 20 • C in Exp. 1. Cellular ratios of PIC / POC increased or did not change with temperature.
Under changing carbonate chemistry speciation, C. braarudii, G. oceanica and E. huxleyi were negatively affected in growth rate, PIC prod and the PIC / POC ratio when pCO 2 was increased from ambient to high levels with species-specific sensitivities depending on temperature.More detailed information regarding the physiological response of C. braarudii and G. oceanica can be found in Krug et al. (2011) and Sett et al. (2014), respectively.The POC prod , however, followed an optimum-curve behaviour in response to pCO 2 peaking between 400 and 1150 µatm.The C / N ratio in all tested coccolithophore species was insensitive to changes in pCO 2 .In general, Chl a prod followed similar trends as described for POC prod and the lowest Chl a prod rates were observed in E. huxleyi (Exp.2) at 10 • C.
Seawater carbonate chemistry speciation and physiological parameters of all experiments are listed in the Supplement Tables 1 and 2, respectively.
The coccolith Sr / Ca ratio and the strontium partition coefficient (D Sr ) ranged from 2.40 to 4.38 mmol mol −1 and from 0.28 to 0.45, respectively.The highest values were found in calcite produced at 25 • C by E. huxleyi and G. oceanica (Fig. 1, Table 1).Elevated seawater temperature at constant pCO 2 conditions (370 ± 70 µatm) resulted in a positive relationship between temperature and the coccolith strontium partition coefficient (when combining all species, Fig. 1a).Species-specific D Sr was positively correlated to pCO 2 with the steepest slopes detected for E. huxleyi (Fig. 1b, data points of G. oceanica at 25 • C were influenced by nutrient limitation which is discussed in Sect.4.1): The Mg / Ca ratio and magnesium partition coefficient (D Mg ) ranged from 0.066 to 98.6 mmol mol −1 and from 0.01×10 −3 to 18.7×10 −3 , respectively, depending on treatment and coccolithophore species (Fig. 2, Table 1).The highest D Mg was measured in E. huxleyi (Exp.2) at 10 • C and the lowest in G. oceanica (Exp. 4).No consistent correlation with temperature could be detected within the tested range  (when combining all species, Fig. 2a).Within the applied pCO 2 range, an increasing trend was apparent in D Mg of G. oceanica, C. braarudii and E. huxleyi (Fig. 2b), with a significant correlation present in G. oceanica:

Experimental approaches and procedures
The pCO 2 conditions in Exp. 1 varied from ≈ 280 to 500 µatm (average of 370 ± 70 µatm).Within this pCO 2 range, however, expected changes in coccolithophore physiology are below the detection limit of the applied measurements (Krug et al., 2011;Bach et al., 2011).Therefore, all changes in physiological rates of Exp. 1 can be attributed to variations in seawater temperature because all other experimental parameters were constant.In Exp. 2, 3 and 4 the seawater carbonate system was manipulated by applying two different methods.First, seawater A T was changed while keeping DIC concentration constant by the addition of HCl or NaOH (Exp. 2 and 3) and, second, adjusting DIC and A T by the addition of Na 2 CO 3 and HCl to artificial seawater (Exp.4).The advantages and disadvantages of these two methods have been presented and discussed in Riebesell et al. (2010) and Schulz et al. (2009).
The second method correctly mimics ocean acidification but both methods are comparable when applying pCO 2 levels within a range from ≈ 180 to 1100 µatm (Schulz et al., 2009).Therefore, physiological and geochemical measurements derived from pCO 2 conditions above 1100 µatm (Exp. 2 and 3) or below 180 µatm (Exp.4) should be considered with caution when relating to the process of ocean acidification.From a cellular perspective, however, the changes in the individual carbonate species (CO 2 , HCO 3 and CO 3 ) and H + concentrations, rather than DIC and A T , are driving changes in physiology and biomineralisation.Consequently, both methods are comparable and, indeed, result in similar physiological responses (Hoppe et al., 2011).
Calcite pellets of G. oceanica were obtained 2-4 days after samples were taken for physiological determinations and seawater carbonate chemistry.This approach was preferred to ensure sufficient material for elemental Mg analyses.However, the exponential growth of G. oceanica and increasing cell density induced a fast consumption of available nutrients (including DIC).Depending on growth rate and time of calcite sampling, this might have induced nutrient limitation, major changes in the seawater carbonate chemistry speciation and thus in cell physiology.In this regard, the theoretical total drawdown of DIC and dissolved inorganic nitrogen was calculated from growth rate, carbon and nitrogen cell quotas, and the time of calcite pellet sampling.For the experiment with G. oceanica conducted at 20 • C, this resulted in an average drawdown of DIC and nitrogen of 18 and 33 %, respectively, indicating no nutrient limitation at the time of calcite pellet sampling.At 25 • C, however, calcite pellets were sampled for when cells had already consumed all available nitrogen, probably being in limitation for more than 2 days.Nitrogen and phosphorus were initially added to the growth media in a Redfield ratio of 16 / 1 and thereby a combined limitation of nitrogen and phosphorus is plausible for G. oceanica at 25 • C. Consequently, the calcite geochemistry data of oceanica at 25 • C cannot be related to physiological parameters or the seawater carbonate chemistry speciation but rather indicates a response to nutrient limitation.

Calcite geochemistry and physiological influence
The partition coefficient of strontium (D Sr ) for each tested coccolithophore species increased with seawater temperature (Fig. 1a) and was well within the range of reported values from other culture experiments (Rickaby et al., 2002;Stoll et al., 2002;Langer et al., 2006).Species-specific D Sr was significantly correlated to pCO 2 at 10, 17 and 20 • C (Fig. 1b) but a correlation was not present at 25 • C. The correlation between D Sr and pCO 2 in G. oceanica at 25 • C is presumably biased by the effect of nutrient limitation (discussed above) and therefore not comparable to the regressions found at 10, 17 and 20 • C.
No general relation was found between D Sr and a specific physiological parameter (growth, POC prod , PIC prod , TPN prod and Chl a prod ) when using combined data from temperature and carbonate chemistry manipulations.A significant correlation for all tested coccolithophore species was found between D Sr and temperature induced variations in growth rate at ambient pCO 2 (370 ± 70 µatm) levels (Fig. 3a).This correlation is likely induced by physiological control rather than temperature itself because Sr incorporation during inorganic calcite precipitation is reported to be negatively correlated with temperature (Tang et al., 2008).It seems that coccolith D Sr is an appropriate indicator for coccolithophore productivity (in terms of growth rate) and a similar relation between D Sr and growth rate would be expected when changes in growth rate are induced by pCO 2 (or carbonate chemistry speciation).Whereas pCO 2 was positively correlated with D Sr in all tested coccolithophore species (see Fig. 1b and equations in Sect.4), no species-specific correlation was found between pCO 2 induced variations in growth rate and D Sr (Fig. 3b).Coccolithophore growth rate and POC prod follow a species-specific optimum-curve behaviour peaking at pCO 2 values between ≈ 400 and 1100 µatm (Krug et al., 2011;Sett et al., 2014;Bach et al., 2011).Here, we demonstrate that over an extended range of pCO 2 from 25 to 3500 µatm the D Sr follows a linear regression depending on temperature and species (Fig. 1b).This suggests that seawater carbonate chemistry speciation has a significant influence on coccolithophore D Sr independent of cellular productivity (growth and POC prod ).The Palaeocene-Eocene thermal maximum (PETM), for example, was marked by a massive input of carbon which resulted in elevated oceanic and atmospheric CO 2 concentrations (Zachos et al., 2005).A theoretical increase in pCO 2 from 1000 to 1700 µatm during the PETM (Zeebe et al., 2009) would translate to an increase of 0.27 in the coccolith Sr / Ca ratio of G. oceanica at 20 • C.This CO 2 induced rise in coccolith Sr / Ca can partly explain the observed increase in fossil coccolith Sr / Ca ratios related to the PETM (Bolton et al., 2012;Stoll et al., 2007b).Therefore, fossil coccolith Sr / Ca ratios from times that experienced changing carbonate chemistry speciation have to receive corrections based on an independent pCO 2 proxy, similar to changes in temperature and sea level (Stoll and Schrag, 2001;Bolton et al., 2012).It remains open to debate which carbonate chemistry parameter is responsible for the observed change in coccolith D Sr , and thus the Sr / Ca ratio.This study was not set up to disentangle the effects of single carbonate parameters on coccolith elemental ratios but future laboratory experiments can accommodate this missing gap (see Bach et al., 2013).Besides temperature and pCO 2 , nutrient limitation has a considerable influence on coccolith D Sr and Sr / Ca ratios.The coccolith Sr / Ca ratio correlates well with nitrogen limited growth of coccolithophores in laboratory experiments (Rickaby et al., 2002;Stoll et al., 2007a), and therefore the application as a productivity proxy seems feasible.However, all studies to date focus solely on the effect of nitrogen as limiting factor, even though the diverging effects on coccolithophore growth and calcification of nitrogen, phosphorus and micronutrient limitation are known (Müller et al., 2008;Schulz et al., 2004).Controlled chemostat experiments, where it is possible to vary the source of limitation while keeping the carbonate system constant (Müller et al., 2012), provide a suitable but labour intensive tool to investigate coccolith elemental partitioning in regard to nutrient limitations (e.g.C, P, Fe and Zn).
The Mg / Ca ratio of biogenic carbonates of foraminifera and corals has been positively correlated with sea surface temperature (Mitsuguchi et al., 1996;Marr et al., 2011).This relationship, however, was not found in calcite samples in this study (Fig. 2a) and neither in the majority of coccolithophore field samples (Stoll et al., 2007a).Values for D Mg of G. oceanica were about 1-2 orders of magnitude lower than values measured of C. braarudii, C. quadriperforatus and E. huxleyi (Table 1).An offset due to the additional re-ductive cleaning step, used for calcite pellets of G. oceanica (see Sect. 3.5), can be ruled out because no similar offset was detected in the P / Ca and Fe / Ca ratios (used as indicators for the cleaning efficiency, Table 1).Gephyrocapsa oceanica was cultured in artificial seawater media whereas all other species were grown in natural seawater.An effect of the different culture media is unlikely as all other physiological and coccolith chemistry parameters are comparable (Table 1 and  Supplement Table 2).Therefore, we assume the observed offset in D Mg of G. oceanica is species-specific.Interspecies differences in the coccolith Mg / Ca signature might complicate the analysis of bulk fossil coccolith material due to an inherited species composition bias.The time intensive procedure to clean coccolith material from organic Mg aggravates a practical application of Mg / Ca ratios as palaeoproxy.A new and time saving method, based on the removal of labile organic bound Mg by active heavy metals, shows promising results and the potential to simplify this cleaning process (manuscript in preparation).
A consistent increase in D Mg with elevated pCO 2 was detected in C. braarudii, G. oceanica and E. huxleyi.This rising trend was statistically significant for G. oceanica at 20 • C but the small sample size for C. braarudii and E. huxleyi precluded a statistical interpretation for the latter two species.At 25 • C, values for D Mg of G. oceanica were up to 7 times higher and displayed a greater variability than at 20 • C which was presumably triggered by the experienced nutrient limitation before the calcite pellets were sampled.Nutrient limitation is commonly experienced by coccolithophores in the open ocean and during bloom events (Van Oostende et al., 2012) and our data indicate that this effect has the potential to greatly influence the Mg / Ca signature of fossil coccoliths.
The relatively high intraspecific variability in D Mg values from the presented culture experiments and from field samples (Stoll et al., 2007a) compared to D Sr lets us suggest that the control mechanisms and pathways for Mg and Sr partitioning in coccoliths are likely to be different.Magnesium plays a crucial role in all energy demanding processes by activating ATP in a number of enzymatic reactions as well as in DNA and RNA synthesis.In plants, Mg is furthermore essential for photosynthesis and the production of chlorophyll.The influx of Mg into the cell is presumably managed through unique membrane channels (Legong et al., 2001).Regarding the numerous processes involving Mg, the distribution of free Mg in the cytosol is likely to be under tight physiological control involving Mg-binding proteins.A disturbance or change in the physiological control/utilisation of the cytosolic-free Mg might have a secondary effect on the calcite magnesium partitioning.Indeed, a relationship between coccolith Mg chemistry and chlorophyll a was suggested earlier (Müller et al., 2011;Ra et al., 2010).This is supported by plotting the D Mg as a function of Chl a prod rate (Fig. 4a), resulting in a significant negative correlation for G. oceanica.range from 0.06 to 5.43 pg Chl a cell −1 day −1 (Supplement Table 2) and from 0.12 to 18.7×10 −3 (Table 1), respectively.A correlation, combining the latter three species, became visible when analysing log-transformed data whereby the correlation for G. oceanica remained significant (Fig. 4b).
The apparent relation between Chl a prod and the coccolith D Mg under both temperature and pCO 2 variations suggests a physiological control over the cytosolic-free Mg available for incorporation into the calcite lattice.We suggest that Mg bound to the calcite lattice originates from the same pool as organically associated Mg, resulting in the observed correlation of coccolith D Mg and Chl a prod .As Ca antagonist, Mg can alter the morphological structure and dissolution kinetics of calcium carbonate precipitates (Loste et al., 2003) and, indeed, laboratory studies indicate drastic malformation of coccolith at high Mg concentration (> 80 mmol L −1 ) in the external growth media (Herfort et al., 2004).The pathway of Mg to the site of calcification (coccolith vesicle) is unknown for coccolithophores but previous results (Langer et al., 2006;Müller et al., 2011) and the current study indicate that Mg is transported into the cell and to the site of calcification on a distinctly different pathway than Ca and Sr.Therefore, the Mg / Ca ratio of extant and fossil coccoliths should be influenced by the individual concentrations of Mg and Ca in seawater rather than the seawater Mg / Ca ratio.The Mg signature of foraminifera tests, however, can reflect changes in seawater Mg / Ca ratios (Segev and Erez, 20013).This obvious difference between foraminifera and coccolithophores is explained by their diverging calcification mechanisms.Coccolithophores presumably transport Mg and Ca over the cellular transmembrane system to the site of calcification whereas in foraminifera an additional source of Mg and Ca is derived from a passive transport, such as seawater vacuolisation or leaking gaps in the pseudopodial network (Nehrke et al., 2013).Nehrke et al. (2013) introduced a conceptual model to explain the trace-elemental composition of foraminifera tests, and our study on coccolithophore calcite gives supportive information to further develop this model.

Conclusions
The D Sr and Sr / Ca ratio of coccolithophore calcite has been subject to numerous palaeoceanographic studies and the application as proxy for coccolithophore productivity (in terms of growth rate) is widely applied.Our laboratory results indicate that this relationship is not valid when coccolithophore growth is influenced by variations in the seawater carbonate chemistry speciation as opposed to temperature.Furthermore, we found that carbonate chemistry speciation has a distinct effect on the D Sr independent of coccolithophore physiology.Strontium and calcium share similar transport pathways from seawater to the site of calcification but magnesium seems to follow distinctly different cellular control mechanisms, indicating that coccolith Mg partitioning is independent from the seawater Mg / Ca ratio.Whereas no overall correlation was present between D Sr and one of the physiological parameters measured, coccolith D Mg was correlated to chlorophyll a production (both under temperature and carbonate chemistry variations), suggesting a tight link between coccolith Mg partitioning and Mg-associated physiological processes.These findings provide a useful contribution for the calibration of palaeoproxies based on coccolith Sr / Ca and Mg / Ca ratios, and give further insights into the underlying cellular mechanisms that lead to coccolith Sr and Mg partitioning.

Supplementary material related to
Strontium partition coefficient (D Sr ) as a function of (A) temperature (pCO 2 = 370 ± 70 µatm) and (B) pCO 2 (temperature as indicated).Black line in (A) represents a linear regression through all data points.Please note that G. oceanica experienced nutrient limitation at 25 • C (see text for details).

Fig. 2 .
Fig. 2. Magnesium partition coefficient (D Mg ) as a function of (A) temperature (pCO 2 = 370 ± 70 µatm) and (B) pCO 2 (temperature as indicated).In (B) markers of E. huxleyi and C. braarudii are related to the y axis on the right.Please note that G. oceanica experienced nutrient limitation at 25 • C (see text for details).
Strontium partition coefficient (D Sr ) as a function of growth rate (excluding the data from G. oceanica at 25 • C). (A) Growth rate driven by changes in temperature ranging from 10 to 25 • C (pCO 2 = 370 ± 70 µatm); black line indicates linear regression through all points.(B) Growth rate driven by pCO 2 ranging from ≈ 25 to 3520 µatm (temperatures as indicated).

Fig. 4 .
Fig. 4. Magnesium partition coefficient (D Mg ) as a function of chlorophyll a production rates (A), black line indicates linear regression through data points of G. oceanica at 20 • C. (B) The logtransformed data of (A) to illustrate the linear correlation in the combined data of E. huxleyi, C. braarudii and C. quadriperforatus (dashed line).The solid line in (B) indicates the linear correlation through data points of G. oceanica at 20 • C.

Müller et al.: Coccolith Sr / Ca and Mg / Ca ratios
The values for Chl a prod and D Mg of E. huxleyi, C. quadriperforatus and C. braarudii displayed a wide