The seawater temperature experiment started on 27 March and ended on 21 July 2011. Prior to the start of the experiment the animals were marked with calcein. The staining leaves a clear fluorescent marker in the shells that can be used to identify which shell material has formed prior to and during the experiment. Initially, the animals were kept at 10.3 ± 0.2 ∘C for 48 days. Then, they were briefly removed from the tanks and marked again. Subsequently, the clams were cultured for 69 more days at 15.0 ± 0.3 ∘C. Ambient seawater was pumped from the adjacent Damariscotta River estuary and adjusted to the desired temperature. The salinity was measured with a Hydrolab^® MiniSonde. It ranged between 30.2 ± 0.7 and 30.7 ± 0.7, in the two experimental phases, respectively. During the entire culture period, all clams were exposed to ambient food conditions. At the end of the experiment the soft tissues were removed.

The food experiment was carried out from 9 February to 29 April 2015. The animals were placed in aquaria inside a climate room at 9 ∘C. Water temperature in the tanks ranged between 8 and 10 ∘C. Water salinity was measured by using an ENDECO 102 refractometer and ranged between 29.6 and 29.9 ± 0.1 in each aquarium. The 15 L tanks were constantly supplied with aerated water from the Wadden Sea. The clams were acclimated for 3 weeks before the start of the experiment. Three dietary regimes were chosen. One treatment consisted of feeding the animals with Microalgae Mix (food type 1), a ready-made solution of four marine microalgae (25 % Isochrysis, 25 % Tetraselmis, 25 % Thalassiosira and 25 % Nannochloropsis) with a particle size range of 2–30 µm. A second treatment was based on PhytoMaxx (food type 2), a solution of living Nannochloropsis algae with a 2–5 µm size range. A third treatment served as control; i.e., the animals were not fed with any additional food. In treatments with food type 1 and 2, microalgae were provided at the constant optimum concentration of 20 × 106 cells L-1 (Winter, 1969). A dispenser equipped with a timer was used to distribute the food 5 times per day. At the end of the experiment the soft tissues were removed. A distinct dark line in the shells indicated the transposition to the NIOZ aquaria and the associated stress. This line marks the beginning of the tank experiment.

The right valve of each specimen was cut into two 1.5 mm thick sections along the axis of maximum growth. For this purpose, a low-speed precision saw (Buehler Isomet 1000) was used. Given the small size and fragility of the juvenile shells used in the food experiment, the valves were fully embedded in a block of Struers EpoFix (epoxy) and air-dried overnight prior the sectioning. Sections of the clams used in the temperature experiment were embedded in epoxy after the cutting. All samples were ground using a Buehler Metaserv 2000 machine equipped with Buehler silicon carbide papers of different grit sizes (P320, P600, P1200, P2500). In addition, the samples were manually ground with Buehler P4000 grit paper and polished with a Buehler diamond polycrystalline suspension (3 µm). Sample surfaces were rinsed in demineralized water and air-dried. In the samples of the temperature experiment, the calcein marks were located under a fluorescence light microscope (Zeiss Axio Imager.A1m microscope equipped with a Zeiss HBO100 mercury lamp and filter set 38: excitation wavelength, ca. 450–500 nm; emission wavelength, ca. 500–550 nm).

The shell of A. islandica consists of pure aragonite and it is divided in two major layers, an outer shell layer (OSL) and the inner shell layer (ISL). The OSL is further subdivided in outer portion (oOSL) and inner portion (iOSL) (Schöne, 2013). These layers are characterized by specific microstructures (Ropes et al., 1984; Fig. 2). The oOSL largely consists of homogenous microstructures with a granular aspect (Schöne et al., 2013). This type of architecture is characterized by approximately equidimensional units of about 5 µm in width. The unit shape tends to be irregular with a bulky aspect. The organization lacks of specific structural arrangement typical of other types of microstructures such as the crossed-lamellar and crossed-acicular microstructures. The latter are the main component characterizing the iOSL and ISL (Dunca et al., 2009). Here, elongated units are arranged with two main dip directions, resulting in a relative oblique alignment. As shown in Fig. 2, the elongation of the structures becomes more evident in the ISL.

The present study focuses on ventral margin of the shells. Analyses were carried out exclusively in the OSL.

Similar to other mollusks, the shell of A. islandica contains pigment polyenes, which are obviously visible when using CRM (Hedegaard et al., 2006). Polyenes are organic compounds containing single (C-C) and double (C=C) carbon–carbon bonds forming a polyenic chain. Their distribution across the shell is not homogenous. The pigments are abundant in the oOSL, whereas they become scarce in the iOSL. Furthermore, an enrichment in polyenes has been observed in the growth lines, potentially indicating their involvement in the biomineralization process (Stemmer and Nehrke, 2014). However, the specific functions of these organic compounds have not been disclosed yet (Hedegaard et al., 2006; Karampelas et al., 2009). Given the high phenotypic variation in pigmentation among and within mollusk species, it has been proposed that coloration does not have a primary function as an adaptive tool (i.e., camouflage, warning signaling) as in other animals (Seilacher, 1972; Evans et al., 2009). This, in turn, can indicate a certain degree of influence of the environment on the pigments, in particular by diet (Hedegaard et al., 2006; Soldatov et al., 2013). In the current study, the effect of different dietary regimes was tested in order to explore the potential of polyenes as environmental proxy.

Shells were mapped with a WITec alpha 300 R (WITec GmbH, Germany) confocal Raman microscope. Scans of 50 × 50 µm, 100 × 50 µm and 150 × 50 µm were performed using a piezoelectric scanner table. All Raman measurements were carried out using a 488 nm diode laser. A spectrometer (UHTS 300, WITec, Germany) was used with a 600 mm-1 grating, a 500 nm blaze and an integration time of 0.03 s. On each sample two to six scans were made, depending of the thickness of the shell. For instance, in juvenile shells (food experiment), two scans of each sample were made. On larger shells used in the temperature experiment, six maps were completed, i.e., two maps in the oOSL, two in the middle of the iOSL and two in the inner portion of the iOSL. Each scan contained between 40 000 and 120 000 data points, depending on the map size. The spatial resolution equaled 250 nm. Half of the maps were performed on the shell portion formed before the experiments. The other half were made on the shell portion formed under experimental conditions. In order to avoid areas affected by transplantation or marking stress, the scans were located far off the calcein and stress lines. Raman maps on food experiment shells were performed 300 µm away from the stress line. In the shells from the temperature experiment, the scans were made 1 mm away from the calcein mark.

Polarized Raman microscopy is known to provide comprehensive information about the crystallographic properties of the materials (Hopkins and Farrow, 1985). The aragonite spectrum is characterized by two lattice modes (translation mode Ta, 152 cm-1 and librational mode La, 206 cm-1) and the two internal modes (in-plane band ν4, 705 cm-1 and symmetric stretch ν1, 1085 cm-1). The ratio (Rν1/Ta) between peak intensities belonging to ν1and Ta is caused by different crystallographic orientations of the aragonitic units (Hopkins and Farrow, 1985; Nehrke and Nouet, 2011). Within each scan, Rν1/Ta was calculated for each data point. New spectral images were generated using WITecProject software (version 4.1, WITec GmbH, Germany). These images were then binarized by replacing all values above 2.5 with 1 and the others with 0. The orientation was quantified by calculating the area formed by pixels of value 1 over the total scan area. The imaging software Gwyddion (http://gwyddion.net/; last access: June 2016) was used for this purpose. The results were expressed in percentage.

The Raman scans of the food experiment shells were analyzed to investigate the pigment composition. Polyene peaks have definite positions in the spectrum according to the number of the C-C and C=C bonds of the chain, which are specific for certain types of pigments. The two major polyene peaks at ∼ 1130 (R1) and 1520 cm-1 (R4) were identified by using the “multipeak fitting 2” routine of IGOR Pro (version 7.00, WaveMetrics, USA). Their exact position was determined adopting a Gaussian fitting function (Fig. 3). The number of single (N1) and double carbon bonds (N4) was calculated by applying the equations by Schaffer et al. (1991): N1=476(R1-1082),N4=830(R4-1438). Spectral images of the R4 band were used to locate the polyenes in the shell and measure the thickness of the pigmented layer. The images were analyzed using the software Panopea (© Schöne and Peinl). The thickness of the pigmented layer was calculated as distance between the outer shell margin and the point where the concentration of polyenes suddenly declined. The measurements were taken perpendicular to the shell outer margin. This analysis was conducted only on the shells of the food experiment. Given the larger size of the shells used in the temperature experiment, the spectral maps were not sufficient for a correct localization of the pigmented layer boundaries and estimation of its thickness.

To quantify changes of the orientation of individual biomineral units of the juvenile shells (food experiment), the spectral maps were subdivided into two portions. The outermost shell portion (oOSL) was enriched in pigments whereas the iOSL showed a decrease in polyene content.

After performing Raman measurements, the samples were prepared for SEM analysis. Each shell slab was ground with a Buehler Metaserv 2000 machine and Buehler silicon F2500 grit carbide paper. To reduce the impact of grinding on the sample surface of juvenile shells, extra grinding was done by hand. Then, the slabs were polished with a Buehler diamond polycrystalline suspension (3 µm). Afterward, shell surfaces were etched in 0.12 N HCl solution for 10 (food experiment samples) to 90 s (temperature experiment samples) and subsequently placed in 6 vol % NaClO solution for 30 min. After being rinsed in demineralized water, air-dried samples were sputter coated with a 2 nm thick platinum film by using a Low Vacuum Coater Leica EM ACE200.

A scanning electron microscope (LOT Quantum Design 2nd generation Phenom Pro desktop SEM) with backscattered electron detector and 10 kV accelerating voltage was used to analyze the shells. Images were taken at the same distances from the calcein and stress lines as in the case of the Raman measurements to assure comparability of the data.

In addition, stitched SEM images of the ventral margins were used to accurately determine the shell growth advance during the culturing experiments. Growth increment widths were measured with the software Panopea. Given the difference in duration of the two phases of the temperature experiment, the measurements were expressed as total growth and instantaneous growth rate (Fig. 4a, b). The latter was calculated using the following equation (Brey et al., 1990; Witbaard et al., 1997): Instantaneous growth rate=(ln⁡(yt/y0)/a), where y0 represents the initial shell height, yt is the final shell height and a is the duration of the experiment. In the case of the food experiment, only the total growth was calculated (Fig. 4c).

When exposed to a water temperature of 10 ∘C, the shells grew between 11.67 and 14.17 mm during a period of 48 days. During a period of 69 days at 15 ∘C, the growth ranged between 2.32 and 5.77 mm (Fig. 4a). The instantaneous growth rate showed a decrease between the two experimental phases. At 10 and 15 ∘C, the average instantaneous growth per day was 0.0091 and 0.0013, respectively (Fig. 4b). The decrease in total growth and growth rate between the two temperatures was statistically significant (t test, p<0.01).

During the food experiment, shells grew between 0.37 and 3.71 mm with large differences due to the different food types. Growth of specimens exposed to food type 1 ranged between 1.87 to 3.71 mm, whereas those cultured with food type 2 grew between 0.55 to 0.96 mm. Both control specimens added 0.37 mm of shell during the experimental phase (Fig. 4c). ANOVA and Tukey's HSD post hoc tests showed significant differences between specimens cultured with food type 1 and 2 (p<0.05) and between food type 1 and control shells (p<0.05).

At a water temperature of 10 ∘C, the area occupied by microstructural units oriented with Rν1/Ta higher than 2.5 a.u. (= arbitrary units) ranged between 31.3 and 50.6 % in the oOSL and between 21.3 and 33.5 % in the iOSL. When exposed to 15 ∘C, values ranged between 25.6 and 48.7 and between 45.7 and 55.9 % in the oOSL and iOSL, respectively (Fig. 5). Whereas the slight difference of area with Rν1/Ta>2.5 in the oOSL was not significant between the two water temperatures (t test, p=0.62), the area with Rν1/Ta >2.5 in the iOSL significantly increased at 15 ∘C (t test, p=0.02). Under the SEM, no difference was visible between units formed at 10 and 15 ∘C (Fig. 6).

In the shells cultured with food type 1, the area occupied by biomineral units oriented with Rν1/Ta higher than 2.5 a.u. during the experiment ranged between 24.8 % (oOSL) and 43.0 % (iOSL). In the shell portion deposited during the acclimation phase, the ratio varied between 19.4 % (oOSL) and 36.2 % (iOSL). Although a trend was recognized, these variations were not statistically different (t tests OSL: p=0.43; ISL: p=0.57; Fig. 7a). On the contrary, in the clams exposed to food type 2, the area occupied by units oriented with Rν1/Ta>2.5 ranged between 11.7 % (oOSL) and 20.4 % (iOSL). Before the experiment, the proportions were higher, i.e., 18.1 % (oOSL) and 26.3 % (iOSL) (Fig. 7b). As for the other treatment, the difference was not significant (t tests oOSL: p=0.34; iOSL: p=0.28). In the control shells grown with no extra food supply, the area with Rν1/Ta>2.5 ranged between 24.6 % (oOSL) and 44.8 % (iOSL) during the experiment and 21.2 % (oOSL) and 44.5 % (iOSL) before the experiment (Fig. 7c). Hence, no trend was visible and the two portions did not show significant differences (t tests oOSL: p=0.59; iOSL: p=0.99). As for the temperature experiment, under the SEM, the microstructure of the shells from the food experiment did not show any change (Fig. 8).

All treatments showed a slightly thicker pigmented layer formed during the experiment than during the acclimation phase (Fig. 9a). During the experiment, clams cultured with food type 1 showed, on average, a thickening by 6.4 %. In the food type 2 specimens, the layer thickness increased by 9.9 %. Control shells showed an increase of 10.4 % (Fig. 9b). However, none of these differences was statistically significant (t test. Food type 1: p=0.43; food type 2: p=0.39; control: p=0.10). According to the position of the polyene peaks, the number of single carbon bonds in the pigment chain did not change between the acclimation and experimental phase (N1=10.1±1.3 and N1=10.0±0.9, respectively). Likely, no significant variation was observed in the number of double carbon bonds (N4=10.5±0.2 and N4=10.4±0.3, respectively; Table 2).

The environmental conditions experienced by mollusks during the process of biomineralization appear to influence shell organization (Carter, 1980). Among the different environmental variables, water temperature is the most studied driving force of structural changes of the shell. For instance, shell mineralogy can vary depending on water temperature (Carter, 1980). According to the thermal potentiation hypothesis, nucleation and growth of calcitic structural units is favored at low temperatures by kinetic factors (Carter et al., 1998). As a consequence, bivalve species living in cold water environments exhibit additional or thicker calcitic layers compared to the corresponding species from warm waters (Lowenstam, 1954; Taylor and Kennedy, 1969). Changes in the calcium carbonate polymorph also affect the type of microstructures (Milano et al., 2016a). However, architectural variations often occur without mineralogical impact (Carter, 1980).

The present results indicate that temperature induces a change in the crystallographic orientation of the biomineral units of A. islandica. Although water temperature was previously shown to have an impact on microstructure formation, the attention has been mainly focused on the effects on the morphometric characteristics (e.g., size and shape) or on the type of microstructure. Milano et al. (2017) demonstrated that size and elongation of prismatic structural units of Cerastoderma edule were positively correlated to seawater temperature variation throughout the growing season. Likely, low temperatures induced the formation of small nacre tablets in Geukensia demissa (Lutz, 1984). Seasonal changes of the microstructural type were reported in the freshwater bivalve Corbicula fluminea (Prezant and Tan Tiu, 1986; Tan Tiu and Prezant, 1989). During the warm months, crossed-acicular structure was produced, whereas simple crossed-lamellar were formed during the winter period. So far, variations of the crystallographic properties of bivalve biominerals have been exclusively investigated as a response to hypercapnic (acidified) conditions. Mytilus galloprovincialis and Mytilus edulis showed a significant change in the orientation of the prisms forming shell calcitic layer when subjected to hypercapnia (Hahn et al., 2012; Fitzer et al., 2014a). Altered crystallographic organization may derive from the animal exposure to suboptimal conditions. These findings together with the present results suggest that thermal- and hypercapnic-induced stress are likely to affect the ability of the bivalves to preserve the orientation of their microstructural units (Fitzer et al., 2014b).

Different food sources do not significantly influence the orientation of the biomineral units or the composition and distribution of pigments in shells of A. islandica. In previous studies, the relationship between microstructure and diet was virtually overlooked resulting in a lack of data in the literature. As suggested by Hedegaard et al. (2006), however, the type of polyenes is influenced by food. The ingestion of pigment-enriched microalgae potentially leads to an accumulation of pigments in mollusk tissues and the shell (Soldatov et al., 2013). On the other hand, it has been argued that polyenes do not generate from food sources like other pigments (i.e., carotenoids), but they are locally synthesized (Karampelas et al., 2009). In accordance to Stemmer and Nehrke (2014), the results presented here support the view that the specific diets on which the animals rely on do not influence shell pigment composition. The chemical characteristics of the polyenes are likely to be species-specific and independent from the habitats.

From a methodological perspective, the present study represents an innovative approach in the investigation of shell microstructural organization. Electron backscatter diffraction (EBSD) has been previously used to determine the crystallographic orientation of gastropod (Fryda et al., 2009; Pérez-Huerta et al., 2011) and bivalve microstructural units (Checa et al., 2006; Frenzel et al., 2012; Karney et al., 2012). Whereas, CRM on mollusk shells is generally applied within studies on taphonomic mineralogical alteration and pigment identification (Stemmer and Nehrke, 2014; Beierlein et al., 2015). Both techniques provide considerably high spatially resolved analysis up to 250 nm, allowing for the identification of individual structural units at micrometer and nanometer scale (Cusack et al., 2008; Karney et al., 2012). CRM offers important advantages supporting a broader application of this methodology in the biomineralization research field. For instance, samples do not require any pre-treatment. Unlike EBSD, there is no need for preparing thin sections (∼ 150 µm thick) or etching the shell surface (Griesshaber et al., 2010; Hahn et al., 2012). Therefore, further structural and geochemical analyses can be easily performed on the same sections (Nehrke et al., 2012). In addition, the size of CRM scans can be remarkably large (∼ 7–8 mm2) without compromising the achievable resolution. By overlapping adjacent scans, it is possible to produce stitched scans allowing one to further increase the region of interest on the shell surface. On the other side, EBSD provides a relevant advantage to take into consideration. It allows for absolute measures of the crystallographic orientation of the carbonate structures. The CRM, instead, determines the relative change in the orientation between the single units without providing absolute values.

SEM has previously been demonstrated to provide a convenient approach for the identification of individual structural units and the quantification of potential changes occurring within them (Milano et al., 2017, 2016b). However, SEM exclusively provides information about the morphometric characteristics of the microstructural units. As highlighted by the present study, to achieve an exhaustive examination, it is suggested to combine SEM with techniques assessing crystallographic properties of the biomaterials. For instance, our results show that the effect of water temperature is detectable in crystallographic orientation but not in morphometric features of the biomineral units.

Numerous previous studies demonstrated that the growth rate of A. islandica is linked to environmental variables (e.g., Witbaard et al., 1997, 1999; Schöne et al., 2004; Butler et al., 2010; Mette et al., 2016). However, the relative importance of the main factors, temperature and food supply/quality driving shell formation are still not well understood. Positive correlations between shell growth and water temperature have been identified (i.e., Schöne et al., 2005; Wanamaker et al., 2009; Marali et al., 2015), but the relationship between shell growth and environment is more complex (Marchitto et al., 2010; Stott et al., 2010; Schöne et al., 2013) and likely dependent on the synergic effect of food availability and water temperature (Butler et al., 2013; Lohmann and Schöne, 2013; Mette et al., 2016). Tank experiments were run in order to precisely identify the role of these two parameters on the shell growth of A. islandica (Witbaard et al., 1997; Hiebenthal et al., 2012). A 10-fold increase in instantaneous growth rate was observed between 1 and 12 ∘C, with the greatest variation occurring below 6 ∘C (Witbaard et al., 1997). On the contrary, a temperature increase between 4 and 16 ∘C was shown to produce a slowdown of shell production (Hiebenthal et al., 2012). Our results are in agreement with the latter study and show a decrease in the instantaneous growth rate between 10 and 15 ∘C. High temperatures are often associated with an increase of free radical production (Abele et al., 2002). A large amount of energy then has to be allocated to limit oxidative cellular damage (Abele and Puntarulo, 2004). This translates into a higher accumulation of lipofuscin and slower shell production rate (Hiebenthal et al., 2013). The contrasting results of previous studies may be explained by individual differences in the tolerance toward temperature change (Marchitto et al., 2000).

Along with water temperature, food availability was also shown to influence A. islandica shell growth (Witbaard et al., 1997). At high algal cell densities, the siphon activity increased. This, in turn, was positively correlated to shell growth. Previous experiments used different combinations of algae, such as Isochrysis galbana and Dunaliella marina (Witbaard et al., 1997), or Nannochloropsis oculata, Phaeodactylum tricornutum and Chlorella sp. (Hiebenthal et al., 2012) to grow the clams. However, there are still uncertainties about the composition of the primary food source for this species (Butler et al., 2010). Even though it is challenging to determine the preferred algal species, our results show that the use of a mixture of different algal species results in significantly faster shell growth than the used of just one algal species. In the natural environment, suspension feeders such as A. islandica preferentially ingest certain particle sizes (Rubenstein and Koehl, 1977; Jorgensen, 1996; Baker et al., 1998). The exposure to a limited algal size range, as in the case of food type 2, may affect shell growth. Furthermore, multispecific solutions contain a higher variability of biochemical components that better meet the nutritional requirements of the animal (Widdows, 1991). Our results are in good agreement with previous findings. For instance, it has been shown by Strömgren and Cary (1984) that Mytilus edulis shell growth increased as a result of a diet based on three different algal species. Furthermore, Epifanio (1979) tested the differences on the growth of Crasssostrea virginica and Mercenaria mercenaria of a mixed diet composed by Isochrysis galbana and Thalassiosira pseudonana and diets consisting of the single species. Faster growth was measured in the mixed diet treatment, indicating a synergic effect of the relative food composition (Epifanio, 1979). Likely, Mytilus edulis grew faster when reared with different types of mixed diets as opposed to monospecific diets (Galley et al., 2010).