Size effect on the mineralogy and chemistry of Mytilus trossulus shells from the southern Baltic Sea: implications for environmental monitoring

Mussels have the ability to control biomineral production and chemical composition, producing shells with a range of functions. In addition to biological control, the environment also seems to influence the process of biomineralization; thus, shells can be used as archives of ambient water parameters during the calcium carbonate deposition. Past and present environmental conditions are recorded in the shells in the form of various proxies including Mg/Ca or Sr/Ca ratios. For such proxies to be accurate and robust, the influence of biological effects including the size of studied organism must be examined and eliminated or minimized, so that the environmental signal can be efficiently extracted. This study considers mineralogy and elemental composition of shells representing four size classes of Mytilus trossulus from the Baltic Sea. Obtained results suggest that mineralogy and chemical composition change throughout the shell development due to most likely a combination of environmental and biological factors. The content of aragonite increases with increasing shell size, while the bulk concentrations of Na, Cd, Cu, U, V, Zn and Pb were found to decrease with increasing height of the shells. Therefore, using mussels for environmental monitoring requires analysis of individuals in the same size range.


Introduction
Many marine organisms produce exoskeletons in the form of a shell with specific properties. They perform a range of functions from providing primary defence against predators to delivering protection from external environmental stressors resulting from water parameters (Lowenstam 1981;Lowenstam and Weiner 1989;Mann 2001). Although the constitution of mollusc shell is mainly biologically and genetically controlled (Watabe and Wilbur 1960;Addadi and Weiner 1992;Belcher et al. 1996), environmental factors (e.g. temperature, salinity, water chemistry) also affect its elemental and mineral composition (Dodd 1965;Lorens and Bender 1980;Bourgoin 1990;Pitts and Wallace 1994;Klein et al. 1996a, b). Thus, sequentially formed calcium carbonate layers of shell record the growth histories, metabolism and environmental conditions in which the organism calcified (Fuge et al. 1993;Klein et al. 1996a, b;Stecher et al. 1996;Swart and Grottoli 2003;Strasser et al. 2008). Furthermore, comparison between the compositions of recent and fossilized shells gives an opportunity to estimate natural (pre-civilization) levels of trace element concentrations and to biomonitor the evolution of ecological parameters (Bourgoin 1990;Puente et al. 1996;Carroll et al. 2009). Such a comparison is especially important when instrumental records of oceanographic and climatic parameters are not available.
Bivalves, especially of the genus Mytilus, have been employed for decades as bioindicators in the marine environment (Smith et al. 1979;Bourgoin 1990;Puente et al. 1996;Stecher et al. 1996;Putten et al. 2000;Richardson 2001;Markich et al. 2002;Andral et al. 2004). They occupy widely distributed habitats in the modern oceans, from coastal, often brackish waters to more pelagic environments, as well as being relatively common throughout the fossil record since the Cretaceous (Freitas et al. 2006). Bivalve molluscs, as filterfeeding organisms, are able to concentrate various contaminants from ambient water due to the bioaccumulation process. Components present at often undetectable levels in water can be detected in bivalves due to their high bioaccumulation ability (Zuykov et al. 2013). Many studies suggest that chemical compositions of growth-layered mollusc shells were related to external environmental characteristics. Magnesium to calcium (Mg/Ca) and strontium to calcium (Sr/Ca) ratios in calcite and aragonite biominerals have been proposed as proxies of seawater temperature (Elderfield and Ganssen 2000;Lear et al. 2002;Schöne et al. 2011). Non-biogenic incorporation of manganese (Mn 2+ ) into calcite has been found to correlate with dissolved Mn 2+ concentration in seawater (Franklin and Morse 1983). Thus, Mn/Ca could potentially provide a proxy for dissolved Mn 2+ , whereby reflect the redox processes, primary production (Putten et al. 2000) and associated phytoplankton blooms (Lazareth et al. 2003). Barium to calcium (Ba/Ca) ratio has been linked with the timing and magnitude of diatom blooms (Thébault et al. 2009) or, alternatively, has been suggested to be used as an estuary-specific indicator of salinity (Gillikin et al. 2006;Poulain et al. 2015). Sodium to calcium (Na/Ca) ratio has also been proposed as a salinity indicator; however, the association of Na + with organic matter and its mobility in the crystalline phase of a shell must be considered (Dalbeck 2008). Furthermore, metals such as copper (Cu), cobalt (Co), zinc (Zn), lead (Pb), cadmium (Cd), nickel (Ni), iron (Fe) and vanadium (V) are detected during pollution monitoring (Protasowicki et al. 2008;Youssef et al. 2016).
A robust marine geochemical proxy should depend on a single oceanographic parameter. However, recent studies have increasingly indicated that the use of the mineral and elemental composition of marine bivalve shells as proxies is not a simple task and factors controlling shell variability are not well enough understood (Putten et al. 2000;Freitas et al. 2006;Gillikin et al. 2005a;Lorrain et al. 2005). Multiple parameters, including environmental conditions, e.g. temperature or salinity, and physiological parameters, such as size, growth rate or metabolic activity, can influence these proxies, thereby confounding their use. Consequently, each potential proxy needs to be rigorously studied.
In this study, we focused on the effect of the size of the organism on the calcium carbonate polymorph produced and on the incorporation of trace elements into the shell. Mytilus trossulus has been selected as a model organism and the Gulf of Gdansk in the Baltic Sea, as a model locality. Generally, within a superfamily of Bivalvia, shell mineralogy is rather constant. Bivalves lay down two forms of calcium carbonate in their shells: aragonite and calcite. Shells may be wholly aragonitic, or may contain both aragonite and calcite, in separate monomineralic layers (Kennedy et al. 1969). Different crystal lattice of polymorphs affect the physical properties of the shells and trace element incorporation into them. It is already known that active incorporation of minor and trace elements during shell formation varies with the growth of the organisms (Rosenberg 1980). Therefore, it is important to examine the relationship between mineralogy and elemental accumulation in shells taking into account the biological processes. The amount of Sr and Mg incorporated into the shells of bivalve Mytilus sp. decreases with age (Dodd 1965). A similar trend was observed for Pb incorporation into the structure of shell-forming minerals in the abalone Haliotis spp. (Hirao et al. 1994;Arai et al. 2003). Changes in Sr concentrations related to the ontogenetic stage of the organisms have also been observed in the shell of Mya arenaria, for which Sr/Ca are correlated positively with the age of the individuals (Palacios et al. 1994). Similarly, Schöne et al. (2011) found that Sr/Ca and Mg/Ca ratios in Arctica islandica shells increased with age. According to Dodd (1966), shell age also affects the precipitation of aragonite, which seems to be strongly correlated with shell size in Mytilus edulis. These results confirm that before biomonitoring can be done reliably, the extent to which the composition of shells depends on the size of the organism must be thoroughly studied.
Therefore, the aim of this study is an analysis of the behavour of the 16 elements important either physiologically or in environmental monitoring: Ca, Fe, Mg, Na, Sr, Ba, Mn, Cd, Co, Cu, Ni, Pb, uranium (U), V, yttrium (Y) and Zn. Finding the variability of these elements within four size classes of a single bimineralic species M. trossulus will likely increase the validity of using this species as a tool for environmental monitoring in the Baltic Sea and other regions within its extent of occurrence.

Study area
Samples were obtained from the Gulf of Gdansk located in the southern region of the Baltic Sea (Fig. 1). The gulf is open, but partially sheltered, being bordered to the north-west by the Hel Peninsula and to the west and south by Polish coastline (Rainbow et al. 2004). The location is under the influence of north-east winds, which can cause storm waves reaching a height of 4-5 m. The sea is the most turbulent in January, and the quietest in June. The wind, in conjunction with the bottom topography and shore morphology, is the main reason for the formation of currents in the Baltic Sea. In the Gulf of Gdansk, currents are created mainly due to the north-east exposition and the outflow of water from the west side of the bay, therefore dominating south-west and south-east currents with small average orbital speed near the bottom amounting 0.03 m s −1 with maximum up to 0.1 m s −1 . Small tidal amplitudes, reaching a few centimetres, have only a minor importance for the induction of water currents (State Environmental Monitoring 2014).
An important feature of the southern Baltic Sea waters is seasonal variability of hydrophysical environmental parameters. The Baltic is characterized by three cyclic thermal systems: a summer system with the surface layer of warmest water (about 17-22°C) of up to about 50 m, below which there is cold water from the previous winter; an isothermal system that forms in the autumn and spring, when in the whole water profile the temperature is almost equal (4-6°C); and the winter system, when the surface temperature is the lowest (falls to −0.5°C) and rises at the bottom to 4-6°C (Uścinowicz 2011). The Gulf of Gdansk is a system of estuaries, where seasonal variability also affects the distribution of brackish water and seawater. An increase of melting, rains and river runoff during March and April leads to the salinity minimum. Then, the seawater inflow during autumn, the development of ice and vertical mixing in winter causes an increase in salinity. Consequently, the average salinity of waters in the Gulf of Gdansk varies from 5.5 in summer to 8.4 in winter (Bulnheim and Gosling 1988;Szefer et al. 2002).
Due to the costal location, the gulf is exposed to urban hazards. Variations in yearly temperature and salinity are minimally subjected to anthropogenic changes, which in turn, strongly modify the hydrochemical conditions of the reservoir. Emissions from coastal towns, ports, ships and atmosphere (Uścinowicz 2011) are discharged into the Gulf of Gdansk, making it a highly polluted sea area. The main freshwater inflow into the gulf comes from the Vistula River, which is the second largest input of river flowing into the Baltic Sea. The Vistula and its tributaries traverse highly industrialized and agricultural areas, featuring iron, steel, electrochemical, chemical, petroleum and refining industries, as well as mining industry (Szefer et al. 2002). HELCOM (2004) reported that the mean annual flow of Vistula is 1081 m 3 s −1 and its plume might extend up to 9-27 km from the river mouth. According to the report of State Environmental Monitoring ( The cyclical nature of the environmental conditions in the gulf is closely linked with plankton blooms. The feeding behaviour of M. trossulus as a suspension feeder depends on the presence of particulate matter, mainly phytoplankton, reaching its biomass peak during spring and summer. During autumn and winter, phytoplankton growth is limited or stopped due to the decreased temperature, limited availability of light and even surface ice cover (Pierścieniak et al. 2010;Staniszewska et al. 2015).
The seasonality of temperature, salinity, water chemistry and changing dominance of plankton has a huge impact on the saturation state (Ω) of calcite and aragonite in the Baltic Sea. The average seasonal amplitude of Ω is unusually high in the Baltic Sea (range of about 4) in comparison to open ocean locations (range of between 1 and 2), and varies from about 1 to 5 for calcite and from 0.5 to 2.5 for aragonite with minimum in winter and maximum in summer (Findlay et al. 2007).
Such low winter values do not occur in oceanic surface waters at similar latitudes, which are significantly oversaturated (Ω > 1) with regard to both calcite and aragonite throughout the year ).

Species
Populations of the common mussel Mytilus sp. are key elements of intertidal and subtidal hard-bottom communities from the temperate to subarctic coastal zones of the Northern Hemisphere (Väinölä and Strelkov 2011). For the purpose of this investigation, we focus on M. trossulus Gould (Mollusca, Bivalvia). This mollusc occurs on various types of substrates including rocky surfaces, wrecks, pier pilings and harbour walls (Rainbow et al. 1999); hence, it is easy to collect. M. trossulus can be the dominant member of a diverse community of invertebrates such as barnacles, nematodes, polychaetes, bryozoans, gastropods or hydrozoans. It is a filter feeder preying mainly on phytoplankton (Rainbow et al. 2004;Lauringson et al. 2013). At the same time it is an important component in the diet of many predators in the Gulf of Gdansk, especially for flatfish, cod and crab Eriocheir sinensis (Wojcik et al. 2015). It is a temperate and coldwater species, becoming more abundant in the northern reaches of its range  (Ricketts and Calvin 1971), where it encounters optimum growth temperature of 10-20°C (Haderlie and Abbott 1980). The physiology of M. trossulus in the Gulf of Gdansk is no doubt under the control of low salinity and seasonal variability. Generally, M. trossulus grows up to 10 cm in height (Gofas 2004), yet in the estuarine environment of the Gulf of Gdansk, it reaches a maximum length of up to about 50 mm (Abramowicz personal comm.). M. trossulus is better adapted to variable, particularly lower, salinity than other Mytilus species. Riisgård et al. (2013) revealed that the growth rates of M. edulis and M. trossulus under the same conditions with salinity around 20 are nearly identical. Both species have the ability to grow fast under optimal conditions and to reduce growth rate due to decreasing salinity, but only M. trossulus is able to completely acclimatise its filtration rate to salinities below 7 over extended time periods (Riisgård et al. 2013(Riisgård et al. , 2014. The annual thermal amplitude in the gulf makes the environmental conditions leading to mussels highest biomass production increase from spring which reaches a maximum in June, then decreases through autumn and winter. Such patterns could be linked to reduction or lack of food availability in colder seasons or to gonad development after spawning. Overlapping changes in metabolic rate result in winter hold up of absorbed food at the expense of reduction in vital processes (Pierścieniak et al. 2010). The shell of M. trossulus is bimineralic and consists of two calcium carbonate layers: outer calcite and inner aragonite in variable proportions between individuals (Eisma et al. 1975;Dalbeck 2008;Zhang and Zhang 2006).

Sample collection and preparation
Bivalves were collected from 36 m depth in the Gulf of Gdansk in May 2013 at GN station ( Fig. 1), which is the site used by researchers over many years to obtain samples of M. trossulus (Szefer and Szefer 1990;Rainbow et al. 2000;Rainbow et al. 2004). Each of the individuals was from a single population, thereby was subjected to the same environmental conditions during development, reducing the number of variables that may influence the mineralogy and chemical composition of a shell. The water temperature at the bottom during sample collection was 3.1°C and the salinity was 7.3. Sample was obtained using a Van Veen grab sampler from RV Oceanograf-2.
All mussels were collected alive and transported to the laboratory. One hundred thirty six individuals were selected for further analysis. Mussels from the natural environment are abundantly covered with overgrowing flora and fauna, mainly barnacles and bryozoans, which would disrupt the results. Therefore, after the removal of soft tissues, every individual was viewed under a stereoscopic microscope to check for the unwanted organisms which, if present, were removed. Each shell was cleaned by sonication for 30 min in deionized water and air-dried. The periostracum was removed using the scalpel. By measuring along the shell height, using the calliper with an accuracy of ±1 mm, individuals were divided into four size classes: I (6-15 mm), II (16-25 mm), III (26-35 mm) and IV (36-44 mm) (Fig. 2), reflecting the range in size from juvenile to adult for the southern Baltic Sea population.
From all individuals, one valve was ground into powder using an agate pestle and mortar and subjected to mineralogical analysis to evaluate the calcite and aragonite content. To estimate the chemical composition of M. trossulus shells, the second valve from six individuals at each size class was selected for elemental analysis, weighed and grounded into powder using an agate pestle and mortar. The aliquots (5-320 mg) of powdered samples were weighed using a 5-digit analytical balance into a 15-ml plastic tube (Sarstedt), dissolved in 1.5 ml concentrated nitric acid (HNO 3 , Fig. 2 The ranges of each size classes of M. trossulus shells from the Gulf of Gdansk. N is a number of analysed individuals Sigma-Aldrich, Trace SELECT ® for trace analysis), 1.5 ml ultra-pure water (18.2 mohm cm −1 ) and 0.3 ml 30% hydrogen peroxide (H 2 O 2 , Merck Suprapur ® grade) for 24 h at 70°C and made up to 15 ml by weight with ultra-pure water. After mixing the solution and allowing enough time for the undigested organic material to settle, the top 10 ml of the sample solution was collected for the ICP-AES and ICP-MS analysis.

Mineralogical and elemental analyses
Mineralogical analyses were conducted using a highprecision Enraf-Nonius X-ray diffractometer with a position-sensitive detector and a cobalt X-ray source (XRD-PSD) at the Natural History Museum, London. Powdered samples were placed on a circular glass, forming a flat, homogeneous surface and loaded into a deep-well circular rotary mount. This reduced the error resulting from preferential orientation of crystallites, being the cause of randomly generated XRD patterns. Tube operation conditions were 40 kV and 40 mA. A primary germanium 111 monochromator in combination with slit settings of 0.14 × 5 mm was used to limit the X-ray beam to pure Co Ka 1 radiation. Diffracted intensities were collected simultaneously over a 2-Theta range of 120°. Time adopted for each sample analysis was 15 min. The amount of each polymorph in M. trossulus shells was determined by fitting peak intensities compared to standard patterns generated from 100% aragonite (BM 53533) and 100% calcite (ground Iceland spar). The error associated with this method was obtained by testing sample with a known calcite to aragonite ratio and it was estimated to be within ±1%. Patterns were collected and performed using Difgrab™ software. The software has capabilities of not only mineral determination but also peak shifts and intensity analysis. Difgrab™ has been specially developed inhouse for data collection for Enraf-Nonius XRD at Natural History Museum in London. The mineralogical results were expressed in percentage, assuming that the content of calcite + the content of aragonite = 100%.
Major and trace element composition of M. trossulus shells were determined at the Natural History Museum, London, using a Thermo iCap 6500 Duo inductively coupled plasma -atomic emission spectrometer (ICP-AES) for Ca, Mg, Na, Sr and Mn, and an Agilent 7700x inductively coupled plasma mass spectrometer (ICP-MS) for Ba, Cd, Co, Cu, Fe, Ni, Pb, U, V, Y and Zn. Calibration of ICP-AES analysis was performed using solutions that were matrix-matched to the high calcium concentration in the samples. Multiple wavelengths for each element were recorded and line selection was made accounting for the suitability of the wavelength to the sample concentrations. The accuracy and reproducibility of the analyses were checked using two calcium carbonate-rich certified reference materials (CRMs): JLs-1 Limestone and JDo-1 Dolomite (both from the Geological Survey of Japan) prepared by total digestion method (using hydrofluoric acid). The reference materials were diluted to match the concentrations of Ca in sample solutions. Ca, Mg and Sr concentrations were found to be within the uncertainty (1 standard deviation) of the reported values (Imai et al. 1996).
Limits of quantification (LOQ) in solution for ICP-MS were generally determined as a concentration corresponding to ten times standard deviation of the signal obtained by analysing 5% HNO 3 solution (6-7 times) in each individual run. If a preparation blank for a particular element (e.g. Zn) was higher than LOQ, the operational LOQ in the sample was determined either five times the concentration in the preparation blank or ten times standard deviation of the preparation blank (where at least three individual blanks were available in a single digestion batch), whichever is greater. ICP-MS was run in helium (He) mode (5 ml min −1 He, 99.9995% purity) for most trace elements (V, Mn, Fe, Co, Ni, Cu and Cd) to minimize the molecular interferences from plasma and solution components and Ca from samples; however, 59 Co + and 60 Ni + signals still suffered from potential interferences from 43 Ca 16 O + and 44 Ca 16 O + , respectively. Pure Ca solutions at the same concentrations of Ca as the samples were run periodically to check on the level of potential interferences and based on the results the operational LOQ was set up for Co at 0.06 mg kg −1 and for Ni at 0.58 mg kg −1 in the samples containing 35% of Ca. The accuracy and reproducibility was checked by periodic analyses of JLs-1 and JDo-1. The results obtained for all elements (Table 1) were within the uncertainty (2.5 SD) of the recommended values (Imai et al. 1996). Accuracy of Pb determination cannot be checked using these CRMs because of the large spread of reference values probably due to insufficient homogeneity of Pb distribution in these samples. Based on the analyses of CRMs and matrix-matched solutions, the maximum analytical error for the typical range of concentrations in the shells can be estimated (in relative percentage) as 1.5% for Ca, Mg and Sr; 3% for Fe and Ba; 20% for Co, Ni, Cu, Zn and U; and 4-10% for all other elements. The results were expressed in weight percent and milligrams per kilogram (ppm) of the shell dry weight (Tables 2, 3 and 4). Mg and Sr contents were converted to molar concentrations, normalized to Ca concentration and expressed as element/Ca (mmol/mol) (Table 4).

Statistical analysis
During this study, two main hypotheses were tested. The null hypothesis (H 0 ) assumed that there are no statistically significant differences between the elemental and mineralogical composition of M. trossulus shells in four size classes. The alternative hypothesis (H 1 ) assumed statistically significant differences. In order to verify the assumptions, the concentration of 16 elements, Ca, Na, Mg, Sr, Fe, Mn, Zn, Cu, Ba, V, Ni, Pb, Y, Cd, Co and U, and the proportion of calcite to aragonite in shells were used. Statistics for size classes were made only for the elements with all the values above the LOQ (limit of quantification), that is for Ca, Mg, Na, Sr, Cd, Cu, U, V, Y and Zn. The data did not improve normality (Shapiro-Wilk test); therefore, variability of elements and the ratio of calcite to aragonite between size classes were identified using one-way nonparametric Kruskal-Wallis tests for multiple independent groups. In the case of Pb, two size classes were tested by Mann-Whitney U test for two independent groups. The significance level of the tests (p value) was set at 0.05. If the p value was smaller than the significance level, the null hypothesis was rejected and the alternative hypothesis was accepted as true. This allowed us to test whether the size of the shells affects the composition of the shells. All statistical procedures were run using STATISTICA ver. 10 (StatSoft Inc.).

Results and discussion
Mineralogical and elemental description of M. trossulus shells Analysed shells of M. trossulus are built of two polymorphs of calcium carbonate: calcite and aragonite, with calcite being a dominant form (Fig. 3). Recent studies found CaCO 3 polymorph selection might be  n/a n/a n/a 189 b n/a n/a n/a n/a n/a The results with less-than sign (<) were not used in the calculations n/a not analysed a Sample contaminated with Zn in the laboratory and the result was not used in the calculations b Outliers temperature-dependent, with aragonite precipitation at high temperature and calcite precipitation at low temperature (Cohen and Branch 1992;Kuklinski and Taylor 2009;Ramajo et al. 2015;Krzeminska et al. 2016). Lowenstam (1954) . 3) with mean yearly water temperature about 8°C, supporting the temperature dependence. However, this is probably not the only factor shaping calcium carbonate polymorphism. Parameters such as salinity, pH or elemental composition of the environment indirectly determine the mineralogy of shells (Dissard et al. 2010). Dodd (1963) revealed an inverse relationship between salinity and shell aragonite content, what was confirmed by Zhong and Mucci (1989); Malone and Dodd (1967) or Salisbury et al.   Ramajo et al. (2015) observed that aragonite levels in shells of marine snail Concholepas concholepas along the Chilean coast with opposite temperature and salinity gradients are under temperature control, establishing that temperature is the major factor controlling aragonite to calcite ratio. In the modern world, including the Gulf of Gdansk (Fig. 1), changes in atmospheric pCO 2 also potentially contribute to shifts in skeletal mineralogy (Zhuravlev and Wood 2009;Lee and Morse 2010). With rising pCO 2 and following reductions in carbonate saturation state, the precipitation of calcite, a mineral with higher stability and lower solubility, seems to be favoured (Morse et al. 2006;Zhuravlev and Wood 2009) due to its higher resistance to corrosion. In the bimineralic system of M. trossulus shells from the Gulf of Gdansk, it is likely that outer calcitic layer serves as a protective shield for the inner aragonite layer, and the variability in their proportion on a global scale seems to be the ability of mussels to adjust to the environment. Ocean acidification studies have also reported that species exposed to short periods of reduced pH present different mechanisms of acid-base regulation (Hendriks et al. 2015), usually linked with a metabolic depression (Dupont et al. 2010;Stumpp et al. 2011) and reallocation of energy for essential biological processes such as reproduction and growth (Gattuso and Hansson 2011). As calcite is less energetically costly than aragonite for the organism to precipitate , any fluctuations in metabolic demand including those related to growth and reproduction might stimulate dominance of a given mineral in the shell structure. Palmer (1992) found that for a shell with only 1.5% organic matter, this organic material costs 22% of the total energy of the shell production. In a shell with 5% organic matter, this cost rises to 50%. Aragonite requires more organic material in construction; thus, it is more expensive to precipitate. Therefore, secretion of this polymorph might be regulated by metabolic demand of the organism. As the Gulf of Gdansk is a very seasonal system, the food availability with abundance peaks in spring and summer which could be also a potential factor responsible for observed mineral proportions dominated by calcite (Fig. 3). A reduced food base in colder periods might force M. trossulus to minimize all the vital processes (Pierścieniak et al. 2010) and to sustain on accumulated food energy. While warmer seasons favour aragonite layering, annually, domination of calcite appears to be an adaptation to the ongoing environmental conditions. The aragonite represents a carbonate polymorph that is mechanically stronger, denser and more elastic than calcite (Taylor and Layman 1972;Zuschin et al. 2003). As such, aragonite, due to its mechanical design, is considerably more resistant to crushing and boring by predators when compared with calcite and is also more resistant to hydrodynamic erosion including storms or waves (Jackson et al. 1988;Gray and Smith 2004). Yet, the site, where our specimens were collected, is rather deep (36 m), hence protected from wave action and, generally, from stormy events. Also, Mytilus species in the Gulf of Gdansk do not have many competitors for space or food as it is one of a few bivalves occurring on firm substrate in the area. Thus, all the adaptations which may lead to a change in mineral proportion as a result of hydrodynamics or protection against competition or predation can be considered insignificant. Different ions, replacing Ca 2+ in the calcium carbonate structure, are incorporated into the shell during its formation. The availability and bioaccumulation rate of ions from sea water seem to be a function of environmental and biological factors (Rainbow 1990;Wright 1995;Wang and Fisher 1997). Thus, different habitats, species or even individual specimens at a different stages of development, may present different patterns of metal uptake. This is the reason why each potential bioindicator should be examined from several angles. Generally, it is established that the activity of free ions rather than total metal concentration determines the biological availability of dissolved metals. Both low salinity and low pH increase free ion activity of several metals (Fritioff et al. 2005), which is especially important in brackish habitats. At salinities below 10, the decrease of Ca 2+ concentration in water promotes an influx of ions such as Cd 2+ , Co 2+ , Cu 2+ or Zn 2+ by reducing competition for the same sites in the crystal lattice (Wright and Zamuda 1987). This tendency is enhanced by the fact that the concentration and bioavailability of Sr 2+ and Mg 2+ , major substituents of Ca 2+ in biocarbonates, is usually much lower in fresh water than in seawater (Dettman et al. 2004;Poulain et al. 2015). Furthermore, lower salinity means lower level of chloride ions (Cl − ), forming complexes with metals; hence, their availability is higher in brackish than in marine environments (Fritioff et al. 2005). In the shells of M. trossulus from the southern Baltic Sea, 16 elements were measured: Ca,Fe,Mg,Na,Sr,Ba,Mn,Cd,Co,Cu,Ni,Pb,U,V,Y and Zn (Table 2) with mean concentrations ranging from 34.5 ± 3.6 wt% for Ca to 0.058 ± 0.039 mg kg −1 for U (Table 3). Due to large differences in the concentrations, the results for Ca, Na, Sr, Mg and Fe were expressed as weight percent, while for Mn, Zn, Cu, Ba, V, Y, Co, Cd and U as milligrams per kilogram of the shell dry weight (Table 3). The Baltic Sea region has been sporadically examined for elemental levels in shells of Mytilus species. Szefer et al. (2002) and Protasowicki et al. (2008) studied shells of Mytilus sp. from the Gulf of Gdansk for identification of coastal areas exposed to metallic contaminants. In both cases, levels of metals were lower for Zn, Cu and Fe, higher for Mn and similar for V, compared with this study (Table 3). The differences may be related to variations in environmental metal concentrations, which could be the result of intermittent terrestrial run off or sediment resuspension, caused by hurricanes or storms. Although those events dominate in the Gulf of Gdansk during winter and early spring, when the growth and precipitation of mussels is limited, yet they can alter the biogeochemistry of the water column for several months (Paerl et al. 2001;Gillikin et al. 2005b).

Method ICP-AES ICP-AES ICP-AES ICP-AES ICP-AES ICP-AES ICP-MS ICP-MS ICP-MS ICP-AES ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS
Size-dependent variations in mineralogy and elemental constitution of M. trossulus shells The size of the shells had a statistically significant impact on the content of the produced polymorphs. The content of aragonite increased with increasing body size (Kruskal-Wallis test: H 3,136 = 35.98, p < 0.0001), ranging from 23.8 ± 3.2% in the size class I to 35.0 ± 7.1% in the size class IV (Fig. 3). Generally within genus Mytilus, as suggested by Dodd (1966), the percentage of aragonite in the bimineralic shells seems to be correlated with seasonal temperature variability. Dodd (1963) also noted that the mineralogy of small (less than 15 mm long) shells of Mytilus californianus is not temperature-dependent while larger specimens show a positive correlation between the percentage of aragonite and temperature. Similarly, variations in aragonite content within shells of Mytilus sp. were shown to change with geographical location, decreasing towards higher latitudes (Dodd 1963). In our study, M. trossulus shells representing all size classes were collected in May, when the water temperature was 3.1°C. However, the Gulf of Gdansk is under the influence of seasonal variability (Uścinowicz 2011). Considering that M. trossulus reproduce from late spring to early autumn, and in the study area mainly in August (Dziubińska and Janas 2007), individuals from size class I briefly lived in the warm summer season in comparison with other size classes. This may explain the smallest content of aragonite in the youngest shells (Fig. 3). Seasonality determines not only temperature but also, inter alia, food availability, salinity, light and pH, and all these parameters control the growth rate and calcification of bivalves (Wong and Levinton 2004;Berge et al. 2006;Hiebenthal et al. 2012), so that the summer growth maximum results in aragonite accumulation. In the context of ontogeny, the key variable during the development of living organisms is metabolic rate (Hahn et al. 2014). As mentioned above, calcite precipitation is less energetically costly ; therefore, younger individuals with a higher metabolic rate, have shells built mainly with calcite (Fig. 3), placing energy in the physiological processes. Ramajo et al. (2015) suggested that juvenile individuals of snails C. concholepas may favour the precipitation of calcite, as less soluble and thermodynamically more stable form than aragonite, to make their small thin shells more resistant against corrosion and dissolution. Furthermore, it is possible that, if environmental factors promote erosion of the outer calcitic layer, older shells are left with a high aragonite content.
This study furthermore confirms that the given size of the individual is associated with a certain range of concentrations of elements in shells (Fig. 4). The nonparametric Kruskal-Wallis tests used in this study revealed a significant correlation (p < 0.05) between the element concentrations and shell height ranges for six out of ten selected elements: Na (H 3;21 = 8.99, p = 0.03), Cd (H 3;20 = 10.54, p = 0.01), Cu (H 3;20 = 13.86, p = 0.003), U (H 3 ;2 0 = 11.14, p = 0.01), V (H 3;20 = 12.16, p = 0.01) and Zn (H 3;19 = 10.42, p = 0.02). In all these cases, the element concentrations decrease with increasing shell size (Fig. 4). Eight out of 20 results for Pb were below the LOQ, but its concentrations appreciably decrease with shell growth (Table 2), which was confirmed by Mann-Whitney U test (p = 0.037). High variability among different individuals of the same species confirms the general assumption that bioaccumulation changes with organism size as well. The dependence of metal influx rates on size is commonly explained by size dependence of the metabolic rate (Newman and Heagler 1991;Wang and Fisher 1997), which, generally, is faster in the smaller individuals than in the larger ones. Therefore, the intensity of the metal uptake could be greater in smaller bivalves than in larger ones for the same species and habitat (Lee et al. 1998). The bimineralic nature of M. trossulus and the increasing content of aragonite throughout the size classes (Fig. 3) is probably one of the factors contributing to the elemental variation. Differences in the trace element concentrations between the two polymorphs of CaCO 3 are in agreement with the results obtained by other authors (England 2005;Dalbeck 2008;Iglikowska et al. 2017). Calcite and aragonite have the same chemical composition, but different crystal structures. Hence, trace elements with ionic radii less than Ca 2+ (e.g. Mg 2+ , Cu 2+ , Zn 2+ , U 4+ , Cd 2+ , V 3+ , Y 3+ ) generally accumulate preferably in calcite, while metal ions larger than Ca 2+ (e.g. Sr 2+ , Pb 2+ , Na + ) tend to become preferentially incorporated in aragonite (Langston et al. 1998). In our study, the concentration of Na, Cu, Zn, U, Cd and V decrease with increasing shell size (Fig. 4). The ontogenetic development of M. trossulus is accompanied by decreasing content of calcite in favour of aragonite (Fig. 3), which may be linked with the reduced accumulation of Cu, Zn, U, Cd and V. However, the concentration of Na decreases despite the growing aragonite crystal lattice, and the concentration of Mg and Sr do not seem to depend on the shell length (Tables 4 and 5). Both Mg and Sr are important elements in context of geochemical, paleooceanographic and environmental monitoring studies based on bivalves. Many authors suggested their potential usage as temperature proxy (Dodd 1965;Klein et al. 1996b). However, the majority of previous studies have concluded that the Sr/Ca and Mg/Ca ratios of bivalves depend on the growth rates or other physiological parameters instead of water temperatures. Some authors have assumed an indirect correlation between water temperatures and shell Sr/Ca and Mg/Ca values because the metabolism is influenced by the ambient environment (Gillikin et al. 2005a;Freitas et al. 2006;Elliot et al. 2009, Poulain et al. 2015). Yet, there is no certainty whether environmental variables can be reconstructed from metal proxy excluding vital effect processes (Schöne et al. 2011). Common heavy metal pollutants Cd, Cu, Zn, V and U are often used as indicators of human impact on marine environment (Lindström 2001;Reiss et al. 2004). The concentration of each element decreases throughout the size classes of M. trossulus (Fig. 4). Potential reasons for this may include changes in anthropogenic input of metals over the growth period, seasonal variability in water parameters of the Baltic Sea or physiological changes of mussels during development. The intensity of heavy metals discharge into the Gulf of Gdansk depends on a number of events such as the flow rate of rivers associated with rainfall or melting, storms and water mixing or current scale of pollution. Subsequently, the availability of free ions is controlled, inter alia, by temperature or salinity (Fritioff et al. 2005), and these change between seasons and even years. That is why the environmental bioavailability of elements may vary throughout the life of M. trosullus.

Conclusions
All examples considered in this study, although to some extent speculative, suggest that elemental concentrations in the shells are coupled with the mineralogical composition; thus, the mechanisms controlling their concentration, at least for some of the elements, might be similar. The crystal structure of the polymorph produced can play an important role in determining the uptake of certain elements, and disorders in the crystal lattice can facilitate incorporation of some other ions. However, this relationship seems to be very complex, and studies in other regions and of other species (Dodd 1967;Cronin et al. 2005;Gillikin et al. 2005a;Strasser et al. 2008;Izumida et al. 2011) suggest that it can be controlled by the external environmental parameters or by certain physiological mechanisms specific for the organism, which can lead to observed ontogenetic effects. Consequently, this study indicates that, when using mineralogical and chemical composition of the shells as environmental indicators, the size of the organisms should be taken into account. In order to minimize the bias in mineralogical or chemical composition related to the size of the shell, individuals of the same or very similar dimensions should be selected for the analysis.
Lack of statistically significant changes in the concentration of Ca, Sr, Mg and Y throughout size classes of N number of analysed individuals, SD standard deviation of the average M. trossulus determined in this study may suggest that these elements can be used as environmental proxies regardless of the shell size. Yet, in the reconstruction of the environmental parameters, comparison of chemical compositions of whole shells of the same size will most likely introduce the least bias into the interpretation.