Mytilus edulis were cultured for 3 months under six different seawater pCO2 levels ranging from 380 to 4000 μatm. Specimen were taken from Kiel Fjord (Western Baltic Sea, Germany) which is a habitat with high and variable seawater pCO2 and related shifts in carbonate system speciation (e.g., low pH and low CaCO3 saturation state). Hemolymph (HL) and extrapallial fluid (EPF) samples were analyzed for pH and total dissolved inorganic carbon (CT) to calculate pCO2 and [HCO3−]. A second experiment was conducted for 2 months with three different pCO2 levels (380, 1400 and 4000 μatm). Boron isotopes (δ11B) were investigated by LA-MC-ICP-MS (Laser Ablation–Multicollector–Inductively Coupled Plasma–Mass Spectrometry) in shell portions precipitated during experimental treatment time. Additionally, elemental ratios (B/Ca, Mg/Ca and Sr/Ca) in the EPF of specimen from the second experiment were measured via ICP-OES (Inductively Coupled Plasma–Optical Emission Spectrometry). Extracellular pH was not significantly different in HL and EPF but systematically lower than ambient water pH. This is due to high extracellular pCO2 values, a prerequisite for metabolic CO2 excretion. No accumulation of extracellular [HCO3−] was measured. Elemental ratios (B/Ca, Mg/Ca and Sr/Ca) in the EPF increased slightly with pH which is in accordance with increasing growth and calcification rates at higher seawater pH values. Boron isotope ratios were highly variable between different individuals but also within single shells. This corresponds to a high individual variability in fluid B/Ca ratios and may be due to high boron concentrations in the organic parts of the shell. The mean δ11B value shows no trend with pH but appears to represent internal pH (EPF) rather than ambient water pH.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 About 50% of anthropogenic carbon dioxide (CO2) released to the atmosphere is absorbed by the global oceans. It is predicted that the oceans approach a pH of ∼7.3 within the next 300 years [e.g., Caldeira and Wickett, 2003; Sabine et al., 2004]. This leads to a shift in the inorganic carbon equilibrium towards higher CO2 and lower CO32− concentrations. Therefore, the calcium carbonate (CaCO3) saturation state (Ω) will decrease by about 50% (for pCO2 of 840 μatm) [e.g., Feely et al., 2004; Fabry et al., 2008] and higher-latitude oceans are predicted to become undersaturated with respect to aragonite by the year 2050 [Orr et al., 2009; Cao and Caldeira, 2008] which may have considerable consequences for marine calcifying organisms [Orr et al., 2005]. Many (mostly short-term studies, days to week duration) ocean acidification experiments have been conducted and different responses to high pCO2 were found. The majority of the investigated species showed declining rates of calcification and reproduction [see Doney et al., 2009; Ries et al., 2009]. Most ocean acidification perturbation experiments are not able to properly account for the genetic adaptation potential of marine species, as time limitations usually prevent multi generation experiments. Thus, calcifying marine organisms from habitats with naturally high CO2 concentrations can serve as analogues for future ocean conditions [e.g., Hall-Spencer et al., 2008].
 Kiel Fjord (Western Baltic Sea, Germany) provides ideal conditions for ocean acidification studies due to summer hypoxia in bottom waters and upwelling of CO2 enriched waters [Hansen et al., 1999; Lehmann et al., 2002]. Kiel Fjord is already today frequently exposed to seawater pCO2 values that are predicted for the future global ocean of the next 100–300 years [Thomsen et. al., 2010; Caldeira and Wickett, 2003]. The habitat is characterized by low salinity (10–20), low alkalinity (1900–2150 μmol kg−1), low pH (minimum value <7.5) during summer and autumn, and high pCO2 (maximum value of 2340 μatm; 1 μatm = 0.101 Pa) events. The CaCO3 saturation state in the Kiel Fjord therefore can reach minimum values of Ωarag = 0.34 and Ωcalc = 0.58 [Thomsen et. al., 2010]. It can be assumed that calcifying communities in this habitat have already adapted to a fluctuating carbonate system speciation with frequent high pCO2 events for multiple generations.
 Predictions of future ocean pH scenarios can be improved by studies of climate shifts in the past. In this regard the elemental and isotopic composition of biogenic carbonates (e.g., bivalve shells, coral skeletons, foraminifera) serve as proxies for the reconstruction of past ocean chemistry. In particular, boron to calcium ratios (B/Ca) have been shown to decrease with declining pH in some biological and inorganic carbonates [Hemming and Hanson, 1992; Sanyal et al., 2000]. This results from the fact that in aqueous solutions, boric acid dissociates to B(OH)4− and H+ (equation (1)). Consequently, the proportion of boric acid to borate is pH dependent.
 Boron isotopes (δ11B) have been measured in skeletons of different organisms like corals [Reynaud et al., 2004; Hönisch et al., 2004] and foraminifera [Hönisch et al., 2003; Foster, 2008; Rollion-Bard and Erez, 2010] grown in laboratory experiments as a pH proxy. Therefore, the combination of these two proxies (B/Ca for [CO32−] and δ11B for pH reconstruction) might offer the opportunity to reconstruct the marine paleo-carbonate system as it is necessary to know two of the six parameters (pH, [CO2], [HCO3−], [CO32−], total alkalinity (AT) and dissolved inorganic carbon (CT)) to constrain the system and to establish how CaCO3 saturation state of the oceans changed in the past [Foster, 2008].
 Several studies showed marine calcifying organisms being sensitive to elevated pCO2 and lower pH [e.g., see Doney et al., 2009]. However, blue mussels from Kiel Fjord showed the ability to settle, survive and even calcify under similar low pH conditions, when food concentrations are high [Thomsen et. al., 2010]. Therefore mussels may be good candidate species to provide high resolution records of environmental conditions (like temperature, salinity, pH) and thus contribute to the reconstruction of past climate. Additionally, the wide distribution and adaptation to a broad range of environments [Gosling, 1992] suggests M. edulis as a model organism for such studies.
 Thus, to contribute to the understanding of the ability of calcifying organisms to live under acidified conditions and of biomineralisation mechanisms of M. edulis we sampled hemolymph (HL) and extrapallial fluid (EPF) of mussels at the end of two CO2 perturbation experiments. Fluids were analyzed for pH, CT and elemental ratios. To consider the suitability of M. edulis shell as a proxy archive for ocean pH, boron isotopes (δ11B) were investigated by LA-MC-ICP-MS (method see Fietzke et al. ) in shell portions precipitated during the experimental incubation (shells from the long-term growth study described by Thomsen et al. ).
2. Material and Methods
2.1. Culture and Samples
2.1.1. General Setup
 Two experiments were conducted to investigate how increased water pCO2 influences different parameters (pH, pCO2, HCO3−, CO32− and B/Ca, Mg/Ca and Sr/Ca ratios) of Mytilus edulis extracellular fluids and to test whether shell boron isotopes can be used as a pH proxy of the surrounding seawater.
 Atmospheric pCO2 averaged 386 μatm in 2009 and concentrations are predicted to reach values between 700 and 1000 μatm by the year 2100 [Intergovernmental Panel on Climate Change, 2007]. Particular marine habitats like Kiel Fjord (Western Baltic Sea, Germany) are already exposed to values above 2000 μatm and might occasionally encounter pCO2 values of >4000 μatm if water surface pCO2 doubles [see Thomsen et al., 2010]. Thus, pCO2 levels of 380–4000 μatm were chosen for the experimental incubations. All specimen used in the experiments were collected from subtidal populations in Kiel Fjord (54° 19.8′ N; 10° 9.0′ E) and cultured at the culturing facilities of the Leibniz Institute of Marine Sciences (IFM-GEOMAR) in Kiel. Animals were acclimated under control conditions for 2 weeks prior to experimentation.
 The experiments were conducted in a flow through system (200 ml min−1 per experimental unit) using water from Kiel Fjord which was cleaned (50–5 μm filters, UV-sterilized) and pumped into a storage tank, where it was aerated. The air saturated water was pumped to a header tank and then supplied to the aquaria via gravity feed. Different CO2-air-mixtures were used to equilibrate the experimental seawater (for a more detailed description see the auxiliary material).
 Water pH was measured daily between Mondays and Fridays with a WTW pH meter (pH 340i, electrode: Sen Tix 81 (calibrated with Radiometric IUPAC precision pH buffer 7 and 10), WTW GmbH, Weilheim, Germany; ΔpH ± 0.01). All pH values measured with the pH meter were corrected with respect to the pH values calculated from weekly AT and CT measurements (n = 223, R2 = 0.949) as the latter provide more accurate results:
For all pH values reported in this paper, the NBS scale was used.
 Hemolymph (HL) and extrapallial fluid (EPF) samples were taken directly after removing the bivalves from the aquaria. Valves were opened carefully to not injure the mantle and held open with a 1000 μl pipette tip. Hemolymph was drawn anaerobically with a syringe from the posterior adductor muscle and EPF was taken from the extrapallial space by inserting a flexible syringe needle (Ø 0.6 × 80 mm) between shell and the pallial line.
 From all aquaria water temperature, salinity (with WTW cond 315i salinometer and WTW TETRACON 325 probe) and pH were measured daily. AT and CT samples were taken for carbonate system calculations (Table 1; for more details of exp. 2 see Thomsen et al. ).
Table 1. Water Conditions During Experimental Trialsa
AT (μmol kg−1)
CT (μmol kg−1)
Data for experiment 2 are from Thomsen et al. . Errors are given as SD. Values represent mean over whole experiment.
Temperature in °C, 12.1 ± 0.6; salinity,18.8 ± 1.59.
Temperature in °C, 13.8 ± 0.6; salinity, 15.0 ± 0.6.
 The first experiment was conducted from September, 3rd to December, 18th 2008. We cultured Mytilus edulis (∼140 individuals per tank) ranging in size from 12.4 mm to 46.7 mm at six different pCO2 levels (380, 560, 840, 1120, 1400 and 4000 μatm). Shell length was measured at the longest axis (from umbo to edge at opposite site of the shell) using calliper with an accuracy of 0.1 mm. Fluids were sampled from the five largest bivalves (45.6 ± 0.6 mm) in each treatment. The experimental aquaria had a volume of 15 l and were exposed to a 14:10 hour light:dark cycle. Water pH ranged from 8.02 ± 0.08 to 7.28 ± 0.08 in the different treatments. Temperature was 12.2 ± 0.6 °C and salinity 18.8 ± 1.6.
 During the first 6 weeks mussels were fed with “DTs Live Marine Phytoplankton‒Premium Reef Blend” (DT′s Plankton Farm, Sycamore, IL, USA), a live phytoplankton mixture of three marine algae species (Phaeodactylum tricornotum (40%), Nannochloropsis oculata (40%) and Chlorella sp. (20%)). Since mussel growth was unsatisfactory during that period the animals were additionally fed with Artemia salina after that period as Wong and Levinton  reported best growth of M. edulis with a mixture of phyto- and zooplankton. Food was provided on 3 days a week (∼75 mg algal biomass and ∼45 mg Artemia salina (both dry weight) day−1 aquarium−1) by closing the flow through for 4 hours and adding food directly into the aquaria.
 Fluid samples were centrifuged (30 s) and pH was measured with a microelectrode (WTW Mic-D). Then total dissolved inorganic carbon (CT) was determined immediately after sampling via a Corning 965 CO2 analyzer, (Olympic Analytical Service Essex, UK; accuracy 0.1 mM) in two 100 μl subsamples. Carbonate parameters (pCO2, [HCO3−], and [CO32−]) of the fluids were calculated from measured pH and CT with using the Henderson-Hasselbalch equation:
where α is the CO2 solubility coefficient. pK′1 and pK′2 are the first and second apparent dissociation constants of carbonic acid. αco2 was calculated (0.045 mol/l*μatm) [Weiss, 1974] and pK′2 (9.321) [Roy et al., 1993] was chosen according to experimental temperature and salinity. pK′1 was calculated from pHNBS, CT and pCO2 (measured in Mytilus edulis body fluids by [Thomsen et. al., 2010]) using equation (6) [Albers and Pleschka, 1967]. A linear relationship was found for pK′1 in relation to pHNBS. After adjusting temperature and salinity according to this experiment the regression for pK′1 for M. edulis internal fluids was pK′1 = −0.1323pH + 7.2371 (R2 = 0.3027).
2.1.3. Experiment 2
 From 14th of May to 13th of July a second (long-term growth) experiment was conducted using three pCO2 levels (380, 1400 and 4000 μatm). The level of replication was four. Each 15 l aquarium contained eight small (mean 5.5 mm), eight medium sized (mean 13 mm) and two big (mean 40 mm) mussels. Shells from the medium specimen (13.3 ± 1.4 mm) and fluids from the big specimen were investigated in this study. The experiment was conducted under a mean temperature of 13.8 ± 0.6 °C and a salinity of 15 ± 0.6. The light/dark cycle was 14:10 hours. Food (Rhodomonas sp. suspension, 2900 ± 1200 cells ml−1) was provided continuously by adding it into the aquaria at a rate of 100 ml min−1. Body fluids of the big specimen were frozen at −20 °C directly after sampling and used for elemental analyses by ICP-OES (SPECTRO CirosCCDSOP) at the Institute of Geosciences, University Kiel, Germany. The tissue of the medium sized mussels was removed and shells were dried at 20 °C for 5 days.
 EPF pH could not be measured directly in experiment 2 specimen as there was not enough material for sub sampling and therefore no possibility to measure pH directly without contaminating the fluid samples with respect to boron (due to the high boron content of the glass pH electrode). Therefore boron isotopes and elemental ratios (experiment 2) were compared to EPF pH values estimated from experiment 1 fluid values (n = 12) and HL pH values measured in large M. edulis of an additional experiment presented by Thomsen et al. . The regression for the fluids and the corresponding water was pHEPF = 0.383(37) pHwater + 4.493(288) (R2 = 0.914). Additionally the mean offset between EPF and HL (0.037; experiment 1 this study; n = 6 each) was considered. Regression results for each treatment and the mean offset between HL and EPF pH have been subtracted from water pH of experiment 2.
2.2. Analytical Methods
 The fluids from the second experiment (5–7 samples per treatment) were thawed and diluted 50fold with ultrapure 2% HNO3. They were analyzed for Mg/Ca, Sr/Ca and B/Ca elemental ratios by ICP-OES (Inductively Coupled Plasma–Optical Emission Spectrometry; SPECTRO CirosCCDSOP) at the Institute of Geosciences, University Kiel, Germany. An intensity-ratio calibration procedure was applied using matrix-matched calibration standards and IAPSO seawater as a consistency standard. External precision of the elemental ratios was ∼0.1% RSD for Mg/Ca and Sr/Ca, and 1–4% RSD for B/Ca.
 Ten shells from each treatment of the second experiment were broken to obtain fragments that were grown under treatment conditions (defined by length growth). Subsequently, the fragments were bleached in 10% NaOCl (1% active chlorine) for two days to remove organic components (periostracum and other shell organics). After 24 hours samples were placed into an ultra sonic bath for 10 min and the bleaching solution was replaced. Afterwards shells were rinsed three times with ultrapure water (18.2 MOhm) and then dried at 20 °C. To prevent dissolution of the carbonates the water was adjusted to a pH of ∼9 by adding NH4OH. Boron isotopes were measured by LA-MC-ICP-MS (Laser Ablation–Multicollector–Inductively Coupled Plasma–Mass Spectrometry; Thermo Fisher MC-ICP-MS AXIOM, originally designed and manufactured by VG) connected to an ESI New Wave Research UP193FX excimer laser ablation system equipped with an ESI New Wave Research LFC (large format cell) via a standard sample standard bracketing procedure described by Fietzke et al. . Soda lime glass SRM (NIST610) was used as external standard as Fietzke et al.  showed that there is no matrix effect between carbonate and glass. First, three lines (1000 μm x 150 μm) were measured on the NIST. Then, seven lines of sample and seven lines of standard were measured in alternation. Prior to each of these measuring procedures, a preablation was conducted to remove surface contamination. All instrumental parameters are given in Table 2. The location of the sample lines on M. edulis shells is shown in Figure 1.
Table 2. Instrumental Parameters Laser Ablation
0.9 l/min (Ar)
R.A. Chilton RAC19/RAC705
Ablation cell gas
0.6 l/min (He)
Line scan (1000 μm @ 12 μm/s)
Scan mode Preablation
Line scan (1000 μm @ 50 μm/s)
Pause between lines
3. Results and Discussion
3.1. Water Parameters
 Since the experimental incubation systems used in this study were flow through designs with continuous supply of water from Kiel Fjord, all treatments were influenced by changes in Kiel Fjord water chemistry. Although the water was vigorously aerated prior to introduction into the experimental aquaria, mean seawater pCO2 deviated from the nominal values. Mean seawater conditions during the two experiments are given in Table 1. Carbonate system speciation measured in the extracellular fluids are not compared to these mean values but to values from dates of sampling (Table 3).
Table 3. Acid-Base Status of Hemolymph and Extrapallial Fluid and of the Water at Date of Sampling (Experiment 1)
Treatment pCO2 (μatm) at Day of Sampling
EPF (SD, n)
HL (SD, n)
Water at Day of Sampling
7.52 (0.04; 3)
7.54 (0.05; 3)
7.50 (0.09; 4)
7.53 (0.05; 3)
7.46 (0.05; 3)
7.52 (0.04; 3)
7.45 (0.02; 3)
7.47 (0.05; 4)
7.41 (0.04; 4)
7.47 (0.07; 4)
7.34 (0.09; 4)
7.37 (0.02; 4)
1.79 (0.18; 3)
1.63 (0.31; 3)
1.41 (0.03; 4)
1.62 (0.24; 3)
1.71 (0.18; 3)
1.79 (0.21; 3)
1.55 (0.07; 3)
1.77 (0.14; 4)
1.85 (0.05; 3)
1.61 (0.34; 4)
1.91 (0.07; 4)
1.89 (0.31; 4)
1938 (93; 3)
1648 (183; 3)
1717 (427; 4)
1623 (190; 3)
2297 (284; 3)
1987 (383; 3)
2033 (169; 3)
2226 (348; 4)
2611 (181; 4)
2199 (744; 4)
3430 (696; 4)
3122 (578; 4)
1.70 (0.18; 3)
1.55 (0.30; 3)
1.33 (0.04; 4)
1.54 (0.23; 3)
1.61 (0.17; 3)
1.70 (0.19; 3)
1.46 (0.07; 3)
1.67 (0.13; 4)
1.73 (0.06; 4)
1.52 (0.31; 4)
1.77 (0.08: 4)
1.75 (0.28; 4)
0.028 (0.005; 3)
0.027 (0.009; 3)
0.020 (0.005; 4)
0.027 (0.005; 3)
0.022 (0.004; 3)
0.027 (0.003; 3)
0.020 (0.001; 3)
0.024 (0.003; 4)
0.023 (0.003; 4)
0.021 (0.003; 4)
0.019 (0.004; 4)
0.019 (0.003; 4)
3.2. Acid-Base Parameters of Fluid Samples (Experiment 1)
 Overviews of all mean extracellular acid-base parameters of the internal fluids (HL and EPF) are given in Table 3 and Figure 2 (for detailed data and regressions see the auxiliary material).
 High extracellular pCO2 values between 1000 and 4000 μatm are found in all aquatic ectothermic metazoans, as diffusive excretion of metabolic CO2 depends on a relatively steep gradient of CO2 from the body fluids to the seawater (see Melzner et al.  for a review). Additional increases in extracellular pCO2 under hypercapnia [Pörtner et al., 2004; Michaelidis et al., 2005; Spicer et al., 2007; Thomsen et al., 2010] are necessary in order to maintain metabolic CO2 flux. In this study a linear increase of fluid pCO2 with increasing water pCO2 was measured, with slightly lower values in the HL (1648-3122 μatm) than in the EPF (1938-3430 μatm). However, pCO2 values in HL and EPF were not significantly different from each other. At the highest pCO2 treatment a reduced pCO2 diffusion gradient could be observed (only low offset between water and extracellular body fluid pCO2), indicating reductions in metabolism in experiment 1. Similar reductions in metabolic rate have been observed in M. galloprovincialis under increased water pCO2 by Michaelidis et al. . This was indicated by a significant decrease in oxygen consumption. However, no reductions in metabolism were witnessed in another experiment using Baltic Sea M. edulis at 4000 μatm under higher food concentrations, suggesting that the observed reduction in the CO2 diffusion gradient in experiment 1 of this study might be due to food limitation [Thomsen and Melzner, 2010]. A recent study on the same mussel population could confirm the strong role of food supply on biomineralisation and shell corrosion intensity [Melzner et al., 2011].
 High HL and EPF pCO2 values are the primary driving force for comparatively low HL and EPF pH values: values range from 7.37 to 7.54 (HL) and 7.34 to 7.52 (EPF) respectively, which is significantly lower in all treatments than seawater pH with values of 8.10 to 7.39. Due to the progressive reductions in the pCO2 gradient between extracellular fluids and seawater, the pH offset between fluids and water decreased as well. At the highest seawater pCO2 level (2724 μatm) pH values of EPF/HL and water became nearly identical. For each treatment, pH values of HL were slightly lower than EPF pH with a mean offset of 0.04 units. However, these differences were not significantly different. Fluid pH values decreased by 0.17 units in the HL and by 0.18 units in the EPF compared to a drop of 0.71 units in the water. The change of hemolymph pH relative to the change in seawater (ΔpHHL/ΔpHSW) in this study was consistent with that found by Michaelidis et al.  (Table 4). The pH in M. galloprovincialis HL decreased from 7.55 to 7.36 (SW: 8.05 to 7.3). Although the control water pCO2 in their study was more than double compared to the study of Thomsen et al.  and this study (1079 μatm and 470/452 μatm, respectively) the water pH was nearly the same. This may be due to the difference in alkalinity but also to the observed accumulation of bicarbonate in that study. In general the Mediterranean Sea is characterized by a higher alkalinity (2500 μmol kg−1) in comparison to the Baltic Sea (∼1900 μmol kg−1). The resulting higher buffer capacity leads to a smaller drop in pH in the Mediterranean Sea at comparable seawater pCO2 values. This explains the different results with respect to hemolymph pH between the studies of Thomsen et al.  and of Michaelidis et al.  although the pCO2 was nearly similar. Thomsen et al.  measured a drop in M. edulis HL pH relative to ambient water pH twice as high (8.05 to 7.08 (SW) related to 7.59 to 7.16 (HL)) as measured by Michaelidis et al.  and in this study. Thomsen et al.  conducted experiments at pCO2 values ranging from 470 to 4250 μatm. However, total alkalinity in the present study was 130 to 150 μmol/kg higher than in that of Thomsen et al. , which is related to the lower salinity in the latter (11.8 compared to 18.8 in this study). This disparity in salinity/alkalinity was due to the high seasonal variability in Kiel Fjord and is likely to be responsible for the different drop in body fluid pH. These results show the high variability in extracellular pH that can be caused by the strong influence of alkalinity and of the reduced metabolism (after three months of experiment with low food conditions) on the pH decrease in response to increased pCO2.
 However, no extracellular accumulation of bicarbonate (HCO3−) could be measured in HL or EPF, indicating that M. edulis does not actively buffer HL or EPF to avoid decreasing pH values. For both fluids, [HCO3−] was below that of the seawater and no significant differences (non-linear regressions) were found between HL and EPF. Accumulation of [HCO3−] (nearly equivalent to a net excretion of protons) is an efficient mechanism of extracellular pH stabilization that is primarily employed by active marine ectothermic organisms that are characterized by pH sensitive respiratory pigments (see Melzner et al.  for a review). The magnitude of this [HCO3−] accumulation response varies between taxa, but highest degrees of [HCO3−] related pH compensation are found in teleost fish, decapod crustaceans and cephalopod molluscs [Larsen et al., 1997; Pane and Barry, 2007; Gutowska et al., 2010]. Two studies also reported [HCO3−] accumulation in mussels to buffer extracellular pH [Lindinger et al., 1984; Michaelidis et al., 2005]. However, both of these were conducted in closed/recirculating systems and it is evident [Lindinger et al., 1984] or very possible [Michaelidis et al., 2005] that this increase in extracellular [HCO3−] was due to increases in ambient water [HCO3−] due to external/internal shell dissolution [see Thomsen et al., 2010]. Consequently, carbonate concentrations are lower in HL and EPF than in seawater, decreasing with decreasing pH (increasing pCO2). In the highest pCO2 treatment (2724 μatm), HL and EPF [CO32−] were almost equal to seawater [CO32−] within error bars.
 Generally, it is important to note that the acid-base conditions in the body fluids, especially in the EPF, which is in direct contact with the inner shell surface, may impact calcification and existing shell structures. Reduced shell growth under high pCO2 (4000 μatm) was reported by Thomsen et al.  and Thomsen and Melzner . Yet experiment 2 specimens increased their shell mass by >150% during the 8 week experiment, indicating great biological control and an ability to precipitate a shell in fluids that are highly undersaturated with regard to calcium carbonate. However, results from Thomsen et al.  demonstrate that external shell dissolution at the umbo region and thinner aragonite layers may be consequences of high seawater and EPF pCO2 values.
3.3. Elemental Ratios of Extrapallial Fluid and Water (Experiment 2)
 Elemental ratios (B/Ca, Mg/Ca and Sr/Ca) were measured in the EPF of M. edulis cultured at three different pCO2 levels (experiment 2, Table 5). Figure 3 shows the elemental ratios compared to EPF pH. The pH values were extrapolated using a regression of HL data on water pH (see methods). An increase of all elements (B, Mg and Sr) relative to Ca could be observed with rising pH. Although a high variability of the element concentrations was found in the fluid, especially B/Ca and Mg/Ca showed a linear trend (significant at 95% confidence level), while there is no significant trend for Sr/Ca. The variability between different individuals may be due to a strong individual physiological influence on element partitioning. The magnitude of increase with pH was variable for the different elements. For B/Ca ratios (19.8%) it was slightly more than twice as high as for Mg/Ca ratios (7.9%) and lowest for Sr/Ca ratios (1.2%).
Table 5. pH and Me/Ca Ratios of Treatment Water and EPFa
Treatment in μatm
Culture pH (SD; n) Experiment 2
EPF pH Calculated
Percent Shell Mass Growth
Water pH values represent the day of sampling. SD represents the error of different individuals (EPF) and different aquaria, respectively. The precision based on repeated standard (IAPSO) measurements was 0.20–0.25% for Mg, Ca and Sr and 0.88% for boron. EPF pH was calculated from experiment 1 and Thomsen et al.  (see methods).
 Other studies suggested different biological influences being responsible for the variations of Me/Ca ratios between individuals of one species but also within single shells. In addition to their results mentioned above, McCoy et al.  found organic-rich winter growth bands containing elevated B/Ca ratios. A comparable effect was reported by Schöne et al.  for trace elements in Arctica islandica shells. They found a strong influence of organic matrix on the determination of Mg, Sr and Ca. This confirms our results from qualitative laser ablation measurements that show boron concentrations in the periostracum being about one magnitude higher than in the carbonate. Carré et al.  observed increasing Me/Ca ratios (for Mg, Ba, Mn and especially Sr) in aragonitic bivalve shells (Mesodesma donacium, Chione subrugosa) when crystal growth rates increase. Their model predicts decreasing Ca2+-channel selectivity when rates increase. Klein et al. [1996b] found significantly higher Sr/Ca ratios in a young, rapidly grown Mytilus trossulus than in a slowly grown adult individual. They concluded that shell precipitation along lateral margins is dominantly controlled by mantle metabolic activity at the site of carbonate formation. This suggests that also the elemental composition is dependent on the mantle activity as it seems to be influenced by growth rates. Carré et al.  suggested the Sr/Ca measurements from different sections of M. trossulus shells (measured by Klein et al. [1996b]) can also be explained by differences in crystal growth. The results from Ford et al.  support that of this study and show Mg/Ca ratios in shells of Mytilus californianus to be a function of growth rate rather than temperature related. Takesue et al.  also found growth rate dependent alterations in Sr/Ca, B/Ca and Ba/Ca ratios in valves from Corbula amurensis. Sr/Ca seems to be influenced by both temperature and salinity [Dodd, 1965; Wanamaker et al., 2008] but also by precipitation rates [Lorens, 1981; Lorrain et al., 2005; Freitas et al., 2006].
 Experiment 2 consisted of four replicate aquaria (two big individuals per aquarium) for each treatment, and variability in the elemental ratios of the water between different aquaria was observed (single values see auxiliary material). To eliminate this problem Figure 4 shows the distribution coefficients between EPF and water. When plotted against growth rates (shell mass growth during experiment in % from [Thomsen et. al., 2010]) elemental ratios are slightly lower in the EPF than in the ambient water at lower growth rates at 4000 μatm. For this treatment external shell dissolution at the umbo region as well as thinner aragonite layers were observed [Thomsen et. al., 2010]. At high growth rates (pCO2 of 385 and 1400 μatm), elemental ratios are higher in the EPF than in the ambient water. Thus, with increasing growth rates elements like B, Mg and Sr become enriched in the EPF with respect to Ca. This effect may be due to higher growth and calcification rates. Therefore the EPF is more depleted with respect to calcium when compared to other elements and the elemental ratios of the EPF are shifted. As elemental ratios in the EPF increased with growth rate it is also likely that the elemental ratios in the shells rose when precipitated from altered fluids. Therefore our results support findings from different studies showing increased Me/Ca ratios in shells grown at high rates [e.g., Klein et al., 1996b; Carré et al., 2006; Ford et al., 2010].
 Food availability, temperature, salinity and elevated pCO2 (decreased pH) are influencing growth rates (and accordingly the amount of organic matter in the shell) of M. edulis [e.g., Malone and Dodd, 1967; Kautsky, 1982; Kossak, 2006; Thomsen et al., 2010]. While the biological performance depends on these biotic and abiotic factors, seawater pCO2 or pH does not seem to be directly reflected in Me/Ca ratios of bivalve carbonates, rather, they might influence them indirectly via decreased rates of growth and calcification due to a repartitioning of the energy budget of the mussel [Thomsen and Melzner, 2010]. Our results show that the dependence between growth and Me/Ca ratios incorporation is represented in the extrapallial fluid. Thus, elemental ratios (B/Ca, Mg/Ca, and Sr/Ca) in the EPF of M. edulis may primarily serve as an indicator for growth rates/shell dissolution as they become enriched with increasing length growth and depleted by dissolution. This indicates that Mytilus edulis shells cannot be directly used as archive for experienced environmental conditions. Extensive calibrations are necessary to see how parameters such as temperature or salinity influence growth rate. With such calibrations and one for the influence of growth rate on Me/Ca ratios in the shell it may be possible to reconstruct environmental conditions. However, the encountered variability is high due to individual physiology and a high amount of replicates must be measured.
3.4. Boron Isotope (δ11B) Data of M. edulis Shells (Experiment 2)
 Shell portions precipitated during the experimental duration (additional length growth) were investigated for their δ11B composition (Table 6). While the analytical precision of the carbonate boron isotope ratio determination was in the order of 0.9 ‰ the variability between individuals from the same treatment was much higher (3.1–5.5 ‰). This analytical precision results from a very low boron content in the shells, still the inter-individual variability is the dominating factor compromising the use of δ11B. Using the empirical borate δ11B/pH dependence [Klochko et al., 2006] and the difference in the ambient water pH we expect a δ11B change of 4.3 ‰ for this pH range (7.28–8.11). Considering this we would expect to find a systematic change in the mean δ11B values depending on the pH. The results showed a high variability not only between different individuals but also within single shells. The high individual variability was already observed for the fluid elemental ratios. Additionally the considerable amount of boron in the periostracum (see section 3.3) suggests the organic matrix surrounding every single carbonate crystal could be a reason for this variability. Bivalve shells typically contain up to 5% of organic matter in the prismatic and nacre layer, respectively [e.g., Wilbur and Bernhardt, 1984]. The variability may be caused by individual and physiological influences as well as (crystal) growth rate. It is obvious that one factor inducing the variability in shell δ11B is the variability in body fluid pH (EPF), not only between individuals but also over time. Extracellular pH for example was shown to decrease when bivalves close their shell valves [e.g., Crenshaw and Neff, 1969; Crenshaw, 1972]. Crenshaw  observed a drop in EPF pH of Mercenaria mercenaria from 7.41 to 7.25 while valves were closed for 15 min or longer and a maximum drop in M. edulis to 7.2. The pattern how often and how long they close their valves for instance can be different between individuals. Average values of δ11B ranged from 17.4 in the lowest pCO2 (highest pH) to 17.9 ‰ in the highest pCO2 (lowest pH) treatment, respectively. However, the highest value (18.3 ‰) was found at 1400 μatm but within the high variability no significant difference could be observed. Moreover no clear trend could be observed with pH but values differed from those of other biominerals when plotted against ambient water pH (Figure 5). Inorganically precipitated calcite [Sanyal et al., 2000], foraminifera [Sanyal et al., 2001] and corals [Hönisch et al., 2004; Reynaud et al., 2004] are characterized by δ11B values that fit above the measured δ11B curve of B(OH)4- presented by Klochko et al.  (Figure 5) which correlates well with ambient water pH. For the curves plotted the following parameters were used: pKB = 8.597 [Dickson, 1990]; α = 1.0194 [Kakihana et al., 1977] and α = 1.0272 [Klochko et al., 2006], respectively. Our data, when plotted against ambient water pH, differ from this pattern. When depicted against EPF pH, the pH at the calcification site, they also fit into this range between theoretical and measured curve. As HL and EPF pH are relatively low (calculated 7.22-7.55 for this experiment), δ11B plots in a section of the curve where pH dependent changes in δ11B are less pronounced. It seems likely that δ11B in M. edulis shell reflects the internal fluid pH (EPF) and not that of ambient water. Because pH changes of internal fluids are minor in comparison to ambient water pH changes, only small changes in boron isotope fractionation of the shell can be expected. This fact and the high individual variability render boron isotopes in M. edulis shells a less sensitive pH-proxy for seawater pH variations.
Water pH values represent mean values over treatment time. 1SD represents the error between different individuals. Shell lengths reflect mean size at the end of the experiment. Initial mean length was 13.3 ± 1.4 mm over all treatments.
8.11 (0.008; 53)
17.4 (3.1; 10)
22.9 (2.4; 10)
7.73 (0.020; 53)
18.3 (4.8; 11)
24.0 (1.4; 11)
7.28 (0.035; 53)
17.9 (5.5; 11)
20.7 (1.9; 11)
 However, boron isotopes may be beneficial in providing more information on the processes of biomineralization in bivalves. The extrapallial space is described as the area where calcification takes place. Recent studies show how complex the underlying biological mechanisms probably are. Various organic components, such as a hydrophobic silk gel, acidic proteins, chitin and amorphous calcium carbonate as precursor phases are probably involved in the formation of bivalve shells [e.g., Levi-Kalisman et al., 2001; Weiss et al., 2009; Suzuki et al., 2009; Weiss, 2010]. Suzuki et al.  recently identified acidic matrix proteins (Pif97, Pif80) that bind to aragonite crystals and to chitin and might be important regulators of nacre formation. Suzuki et al.  and Weiss  proposed that crystals are formed in a microenvironment separated from the extrapallial fluid via a chitinous membrane. Nevertheless δ11B in the shell indicate that shell formation proceeds at pH values typical of the EPF which indicates that calcification is influenced by EPF conditions independent whether calcification occurs separated in a microenvironment close to the shell.
 Body fluids of Mytilus edulis from Kiel Fjord kept under different seawater pCO2 levels are characterized by low HL and EPF pH values. This is due to the requirements of metabolic CO2 excretion. HL and EPF pH decreases with 0.024 units per 0.1 units in seawater pH. No significant bicarbonate accumulation to buffer internal pH could be observed. Elemental ratios (Me/Ca) in the extrapallial fluid correlate with growth rates due to the higher consumption of [Ca2+] during calcification compared to other elements. More data on the mechanisms of biomineralisation and the general composition of shells not only related to ambient water but also to the body fluids are needed. Also the impact of the EPF on calcification independent of the exact precipitation area (directly the EPF or a microenvironment) and the involvement of components like hydrophobic silk gel, acidic proteins, chitin and also maybe amorphous calcium carbonate should be subject of future studies. It's necessary to better understand the underlying mechanisms of trace element partitioning and isotope fractionation in order to demonstrate a robust use of proxies in bivalve shells. Trace metal composition of the extrapallial fluid and hence of the shell seems to reflect abiotic conditions only indirectly via the impact on metabolism and growth rate. Boron isotopes in M. edulis shell do not reflect seawater pH, they seem to represent internal pH (EPF). In addition, boron isotopes show a high variability between individuals and also within single shells. This may reflect variable contributions of boron rich organic matrices present between single crystals. In the future, more attention should be paid to these organic shell parts also with respect to geochemistry as they often are not completely removed before measuring carbonates.
 The authors wish to kindly acknowledge Magdalena Gutowska for constantly supporting this study. We thank Ulrike Panknin for help with bivalve sampling and culturing algae, Karin Kiβling for supporting ICP-OES measurements and Susann Grobe for carbonate chemistry analysis. AH is grateful for statistical help provided by Claas Hiebenthal and Martin Wahl. AH and JF gratefully acknowledge Hauke Vollstaedt for supporting figure corrections. This study was funded by DFG Excellence Cluster “Future ocean” grants to FM. This work is a contribution to the German Ministry of Education and Research (BMBF)-funded project “Biological Impacts of Ocean ACIDification” (BIOACID). The authors like to thank two anonymous reviewers for their constructive and detailed reports significantly improving the manuscript.