Introduction

The anthropogenic activities can have a drastic impact on marine environments (Cherchi et al. 2012; Frau et al. 2015; Romano et al. 2017), as they can lead to the degradation of sea sediments (Salvi et al. 2015) and seafood quality (Bilgin and Uluturhan-Suzer 2017). In coastal areas, metals are derived from the natural erosion of the upstream basins and from anthropogenic sources (Apitz et al. 2009; Morelli et al. 2017). Human activities comprise (i) modern or past mining/industrial activities (De Giudici et al. 2017, 2018a; Arfaeinia et al. 2016; Romano et al. 2017; Atzori et al. 2018) and (ii) special sources such as ballast waters (Soleimani et al. 2016; Dobaradaran et al. 2018b) and cigarette butts (Dobaradaran et al. 2017, 2018a). Marine bivalves are sensitive organisms that register environmental changes (Carroll and Romanek 2008; Hahn et al. 2012; Jou et al. 2016 and reference therein), and they help in assessing the effect of metal dispersion caused by anthropogenic activities (Boening 1999; Zuykov et al. 2013; Amoozadeh et al. 2014; Sarmiento et al. 2016; Wong et al. 2017). Moreover, bivalve shells are well represented in fossil records (Huanxin et al. 2000), resulting useful proxies for investigating into the past (Kastner 1999).

The ability of bivalves to accumulate trace metals and organic chemicals (Faggio et al. 2018) from the environment has been extensively studied in recent years (Lopes-Lima et al. 2012), as it represents a valuable information in the field of environmental sciences revealing details about biogeochemical interactions potentially useful for bioremediation purposes. Most of the previous studies focused on pollutant accumulation in soft tissues of bivalves from native specimens and/or transplanted specimens (Kucuksezgin et al. 2013; Rzymski et al. 2014), and showed that the soft parts of the organism are useful bioindicators in aquatic ecosystems (Andral et al. 2011; Guo and Feng 2018). Indeed, bivalve mollusks are filter-feeding organisms, able to concentrate in their soft tissues various contaminants from the ambient water, reaching concentrations several orders of magnitude higher than those of the ambient environment (Zuykov et al. 2013 and reference therein). Bivalve soft tissues can be used to assess environmental conditions over short time scales, whereas shells appear to be more sensitive to environmental changes over the long term (Koide et al. 1982; Brown et al. 2005; Vaughn 2018). Incorporation of metals in bivalve shells (Carroll and Romanek 2008) is influenced by the biochemistry of the extrapallial fluid from which the shell is deposited (Wilbur and Saleuddin 1983), biological factors such as growth rate (Tevesz and Carter 1980; Stecher et al. 1996), and metabolic controls on shell formation (Rosenberg and Hughes 1991). In addition, water pH, temperature, salinity, availability of nutrients, metal contents (including contaminants), food availability, and population density affect the shell composition (Hahn et al. 2012; Zuykov et al. 2013).

To our knowledge, no previous investigations were performed on metal incorporation into bivalve shells collected along the Sardinian costs. Past literature mainly focused on metal accumulation in the soft tissues (Lafabrie et al. 2007; Schintu et al. 2008; Andral et al. 2011; Moschino et al. 2017; Sforzini et al. 2018), demonstrating an alteration of the health status of transplanted mussels in the contaminated sites. In this study, we explored bivalve shells collected along Buggerru and San Nicolò beaches (SW Sardinia, Italy), located near the Malfidano mining district (Zn and Pb mine of the Sulcis-Iglesiente area). This abandoned mining district is characterized by the presence of 2.6 million m3 of open-pit excavations and 1.3 million m3 of dump and tailings (RAS 2008), which have drastic effects on freshwater and marine environments. During the mining period, the residues were disposed in a tailing pond located on the coast. Periodically, the pond was opened to allow the disposal of new residues. In the late 1970s, the pond was removed with the opening of the dam, and the residues were dispersed into the sea through water jets (RAS 2008). Harmful metal contamination on the coastal area of Buggerru is a particularly important issue because of the potential threats of metal toxicity to the health of human residents. Furthermore, it can compromise tourism development. Romano et al. (2017) demonstrated the environmental impact of mining activity in marine sediments of the Sulcis-Iglesiente district, investigating marine sediment cores, and showed that Zn concentration reaches 14 g/kg. Although Zn was naturally enriched because of the outcropping of metal-based ores in the area, its anthropogenic enrichment was recognized, and it was mainly attributed to dispersion along the coast of mine waste material from the inland. The most abundant genera we found in the coastal area of the Malfidano mining district are Donax, Glycymeris, Lentidium, and Chamelea, both at Buggerru and San Nicolò.

Zn detoxification in bivalves is a well-known mechanism that can specifically occur via (i) Zn immobilization in Ca phosphate granules associated with digestive and excretory tissues (Coombs and George 1978; Brown 1982; Simkiss and Mason 1983; Deb and Fukushima 1999); (ii) Zn accumulation in membrane-limited vesicles associated with P (George et al. 1978); (iii) Zn deposition in the mineral structures (shell and/or microspherules; Pietrzak et al. 1976) and within the nacreous layer of the shell (Moura et al. 1999; Lopes-Lima et al. 2012). Previous studies on bivalve shells mainly focused on the mean metal content in whole shells (Steinhardt et al. 2016) and the shell pattern distribution of metals and its inter-shell variability (Tynan et al. 2005). A relationship was established between the metal content in the shells and the metal concentration in sediments and waters (Karbasdehi et al. 2016a, b; Cariou et al. 2017). To date, the coordination environment of Zn incorporated into the mineralized shell has not yet been investigated, though it represents a fundamental knowledge to assess the nature of Zn biomineral phases and the biological mechanism involved in the detoxification. Improvement in analytical techniques for accurate mineralogical and geochemical analyses, such as synchrotron x-ray techniques (De Giudici et al. 2014a; Luo and Zhang 2010; Medas et al. 2014b; Castillo-Michel et al. 2017) and the possibility to combine several complementary state of art probes promises to document chemical distribution and speciation of trace elements in biogenic minerals (Medas et al. 2017a, 2018).

In this work, we specifically addressed the question of S and trace element incorporation, in particular Zn and Fe, in the carbonate shells of Malfidano area combining laboratory and SR techniques, namely soft x-ray microscopy combined with low-energy x-ray fluorescence (XRF) mapping analysis, analytical transmission electron microscopy (TEM), and x-ray absorption spectroscopy (XAS) to achieve an effective multi-probe characterization of the same samples, obtaining an accurate and reliable understanding about the biomineralization process. Particularly, the aims of this work were (i) understanding if Zn is homogeneously incorporated and distributed across the shell, (ii) inferring if Zn correlates with other major and minor elements, and (iii) finding if the Zn speciation is unique or Zn can occur in multiple coordination environments. The results from this study will be also useful to develop biomonitoring techniques and to elucidate the impact of metals along the trophic chain in mine-polluted shoreface sediments (De Giudici et al. 2018b). Further studies will be devoted to investigate speciation of other metals.

Study area

The study was conducted in the Malfidano area, located in the Iglesiente mining district (SW Sardinia, Fig. 1). The deposits are mainly Paleozoic formations, which are represented by Cambrian–Ordovician rocks. The Lower Cambrian successions comprise siliciclastic sedimentary rocks and carbonate intercalations (Nebida group) as well as shallow water platform carbonate rocks (Gonnesa Gropus; Bechstädt and Boni 1994). The Middle and Upper Cambrian–Lower Ordovician successions are represented by nodular limestones and slates of Campo Pisano Formation and Cabitza Formation (Iglesias Group), respectively. The Quaternary deposits consist of eolian dunes along the coast and fluvial deposits in the internal area.

Fig. 1
figure 1

Schematic geological map of the investigated area (Boni et al. 2003, modified). The red rectangular indicates the sampling area

Full size image

Ore deposits were exploited from 1870 to 1980 and they consisted both of primary sulfides, related to the metamorphism and magmatism that occurred during the Variscan Orogeny (Marcello et al. 2004) and calamine deposits. In the Iglesiente mining district, the primary sulfides occurred in massive, columnar, lens- and vein-like shapes in the carbonatic formations, and included sphalerite (ZnS) and galena (PbS) with a normal Zn grade of about 8–12 wt% (Marcello et al. 2004). The calamine deposits were derived from different weathering episodes that occurred within the Dolomia Gialla (hydrothermal Dolomia Geodica). Smithsonite and hemimorphite were the principal Zn-bearing minerals associated with nodules of remnant or supergene galena and sphalerite (Stara et al. 1996; Boni et al. 2003), with an overall Zn grade exceeding 20–22 wt% (Marcello et al. 2004).

Sample collection and methods

Bivalve samples of approximately similar shell sizes were collected handly from surface sediments along Buggerru and San Nicolò beaches (Fig. 2) on June 2016. The most abundant genera we found are Donax, Glycymeris, Lentidium, and Chamelea both at Buggerru and San Nicolò. At least 50 specimens for each genus were collected at each beach, immediately stored in bags, and transported to the laboratory. About 20 specimens per genus for each beach were selected, polled into one sample, and pulverized for mineralogical and chemical analysis.

Fig. 2
figure 2

Map of the sampling area. Image from Google Earth, modified (a). Photo of the sampling area (b)

Full size image

Mineralogical and chemical analysis

Shell samples were initially washed in a laboratory under a gentle jet of deionized water and were gently brushed in ultrapure water with a nylon toothbrush to eliminate algae, sand, and other impurities. Shells were then washed several times in Milli-Q water and dried at room temperature for 1 week. Samples were grounded in an agate mortar and analyzed by x-ray diffraction (XRD) using laboratory θ–2θ equipment (Panalytical) with Cu Kα radiation (λ = 1.54060 Å), operating at 40 kV and 40 mA, and an X′celerator detector.

About 20 of similar-sized individuals of different bivalve species and same genus from each sampling site were used to avoid differences in metal content because of size, and they were grounded together to form one composite sample. Acid digestions were performed by EPA method 3050 with slight modifications according to Conners et al. (1999) on 0.5 g of each ground sample by adding 5 ml of concentrated superpure nitric acid (65–69%, Carlo Erba) at 90 °C in a hot plate for 2 h. Then, 3 ml of hydrogen peroxide (30% w/w, Sigma-Aldrich) were added to remove any remaining organic material at 90 °C for 1 h. Samples were processed together with blanks prepared with the same mixture to evaluate contamination from the reagents and sample containers. Finally, samples were filtered (0.4 μm) and the solutions were diluted to a final volume of 25 ml using Milli-Q water. Sulfur and metal concentrations (Zn and Fe) were quantified by inductively coupled plasma optical emission spectrometry (ICP-OES, ARL Fisons ICP Analyzer 3520 B), and inductively coupled plasma mass spectrometry (ICP-MS, PerkinElmer, Elan 5000/DRC-e, USA; Zn, Fe, Pb, Mn, Ni, Co, Cr, Cu, Cd, Se, Mo, As, Sb, and U). Trace elements not shown are below the limit of quantification (Cu < 0.1 mg/kg, Cr < 0.7 mg/kg, Se < 0.1 mg/kg, Mo < 0.1 mg/kg, As < 0.5 mg/kg, Sb < 0.05 mg/kg). In this work, we focused on S, Zn, and Fe incorporation because (i) they are important mineral components involved in biological reactions but an exposure to high concentrations of Zn and Fe may lead to variation in the properties of the bivalve shell (Hahn et al. 2012; Zuykov et al. 2013) and (ii) Zn and Fe are among the most abundant contaminants in the investigated area (Cidu et al. 2007, 2009).

Two duplicate samples were analyzed to estimate method precision (expressed as standard deviation/mean concentration) that was in the range 3–7% for Fe, 0.1–0.3% for Pb, 0.3–2% for Mn, 8–10% for Ni, 1–4% for Co, 2–4% for Ag, 2–5% for Cd, and 0.3–1% for U. The limits of detection (LOD) and of quantification (LOQ) are reported in Table 1. To evaluate the analytical accuracy of the acid digestion procedure, experimental values and the certified values of the reference material (BCS No. 368 - dolomite) were compared, and the percentage recovery (Table S1) of each metal was calculated as:

$$ \%\mathrm{Recovery}=\frac{\mathrm{Mean}\ \mathrm{value}\ \mathrm{of}\ \mathrm{the}\ \mathrm{measured}\ \mathrm{concentration}\ \left(\frac{\mu g}{l}\right)}{\mathrm{Certified}\ \mathrm{concentration}\ \left(\frac{\mu g}{l}\right)}\times 100 $$
Table 1 Sulfur and trace element concentration in bivalve shells. Precision, expressed as standard deviation/mean concentration of reference solution, < 5%. Chamelea and Donax from San Nicolò were digested in duplicate, the mean concentrations of each element are reported (precision 0.1–10%)
Full size table

To estimate potential contaminations, the accuracy and precision of trace element analysis, procedural blanks, and standard and reference solutions (SRM 1643e and EnviroMAT Drinking Water, High EP-H-3 and Low EP-L-3) were analyzed after every five samples.

X-ray microscopy techniques

Soft x-ray microscopy, low-energy XRF mapping, and TEM microscopy techniques were employed to investigate specific metal distributions from the micro- to the nanoscale. Iron and Zn distribution was investigated by soft x-ray microscopy combined with low-energy XRF mapping analysis. Sections of shell samples were prepared by dehydration in a graded series of acetone solutions (25, 50, 75, and 100%) followed by 100% toluol. All steps were carried out at room temperature for 15 min each. The samples were then left overnight in a mixture of 50:50 toluol/epon resin (Epon 812) followed by immersion in 100% resin for 30 min at room temperature, and then polymerization at 60 °C for 48 h. Semi-thin sections (1 μm) were cut with a diamond knife on ultramicrotome and collected on ultralen films. The analysis were performed at the TwinMic beamline (Gianoncelli et al. 2016a, b; Kaulich et al. 2006) at Elettra-Sincrotrone Trieste (Trieste, Italy).

The TwinMic microscope was operated in the scanning transmission x-ray microscope (STXM) mode, where the sample is scanned across a microprobe delivered by a suitable zone plate diffractive optics. Absorption and phase contrast images are generated from the transmitted x-rays (Gianoncelli et al. 2006; Morrison et al. 2006) collected through a fast readout charge-coupled device camera (Andor Technology), while the simultaneous XRF signals emitted by the specimen is acquired by eight silicon drift detectors (Gianoncelli et al. 2009, 2013). This setup allows the simultaneous collection of morphological and chemical information.

STXM data were mainly addressed to individuate the Fe and Zn distribution in the shell with micrometer resolution, the x-ray beam energy was 1.5 keV, to ensure the best excitation and detection of Zn and Fe, with a spatial resolution (x-ray spot size) of 1.2 × 1.2 μm2 as a compromise between a good XRF signal and dimension of the features of interest. The XRF elemental maps were deconvoluted and analyzed with PyMCA software (Solé et al. 2007).

TEM analysis provided elemental distribution with subnanometric resolution. The shell samples for TEM analysis were gently crushed in an agate mortar and dispersed in ultrapure water in an ultrasonic bath for 3 min. A droplet of this suspension was then deposited onto a TEM Cu grid. For analytical scanning TEM (STEM) investigations, we used a JEOL 2100 microscope (NIMP, Laboratory of Atomic Structures and Defects in Advanced Materials, Magurele, Romania) equipped with a LaB6 gun and JEOL JED-2300T energy-dispersive x-ray spectrometer. The acceleration voltage was 200 kV. Energy-dispersive x-ray spectroscopy (EDS) mapping in STEM mode was performed to map the distribution of the main component elements in the samples (Zn, Fe, Ca, and S). This technique provides simultaneously both spatial and spectral information in each acquired pixel. A selected area was scanned using a convergent beam with a diameter of approximately 0.3 nm, and the x-ray signal from each point of the scan was collected by the detector.

Calcium and S distribution was further investigated by scanning transmission x-ray microscopy (STXM) combined with XRF elemental mapping analysis at the I08-SXM beamline in Diamond Light Source, Didcot, UK. Sections of shell samples were selected and prepared using the same procedure for the TwinMic beamline and fixed on silicon nitride thin membranes. The x-ray beam was focused down to about 800 nm with a tungsten Fresnel zone plate (40-nm outermost zone width and 333-um diameter) optimized to ensure a good XRF sensitivity with a compromised spatial resolution for this work. X-ray fluorescence mapping was conducted using a 4.1-keV photon energy to collect the Ca K-edge fluorescence, and simultaneously, transmission images were recorded with a photo diode. Preliminary S K-edge x-ray absorption near edge structure (XANES) spectromicroscopy data were collected to speciate the S chemical states of the Chamelea sample. Quantitative STXM data analysis and element mapping were performed using MANTiS (Lerotic et al. 2014) and PyMca (Solé et al. 2007), respectively.

Correlation analysis

In order to evaluate the relative distribution of selected elements in the shells, the linear correlation ρA-B between the fluorescence intensities of A and B elements in the maps was calculated (De Giudici et al. 2018b). Higher correlation values point out major co-localization of the A and B elements in the maps, lower correlation suggests the A and B elements are in different regions of the shells, thus in different phases. The correlation analysis was carried out using a python script based on the Numpy (2014) and Scipy (2014) libraries. The script may use a cutoff intensity value to automatically individuate voids in the images and select only the image regions corresponding to shell regions. Noticeably, the correlation estimate is affected by statistical noise on pixel intensity (Poisson counts); in EDS TEM images, the counts per pixel in the energy region of interest (ROI) may be quite low, even close to the unit, giving high Poisson noise that may reach the 50% of the signal, worsening the correlation estimate. In order to reduce the statistic noise, the script allows integrating ROI counts over N × N pixel squares. We selected N = 4 in the EDS TEM correlation analysis, while N = 1 is used for analysis of STXM images.

X-ray absorption spectroscopy at the Zn K-edge

X-ray absorption spectroscopy (XAS) measurements were carried out at the Zn K-edge (9.659 keV) at the x-ray absorption fine structure (XAFS) beamline of Elettra-Sincrotrone Trieste (Trieste, Italy; Di Cicco et al. 2009) with the aim to individuate the chemical nature of Zn species in the shells. The cleaned shell samples were ground and mixed with polyvinyl pyrrolidone (PVP) matrix (2:1 weight ratio), then pressed in thin solid pellets suitable for XAS measurements. The Zn K-edge absorption spectra were measured in fluorescence geometry at room temperature. Reference materials, including inorganic and organic materials (Table 2 and Fig. S1), were measured in transmission geometry and used for XANES interpretation by linear combination analysis (LCA; Benfatto and Meneghini 2014). Zn K-edge raw XAS data were treated following the standard methods (Meneghini et al. 2005, 2012) for background subtraction and edge-jump normalization. The XAS data have been analyzed in the near-edge region (XANES), using a set of reference compounds (Table 2) for the LCA (Torchio et al. 2010). Because the relatively low Zn concentration in the shells and the complex average Zn coordination (see LCA), the quantitative analysis of the extended EXAFS signal cannot provide further details.

Table 2 Reference compounds analyzed by x-ray absorption spectroscopy
Full size table

Results

Mineralogical composition and trace metals concentration

Figure 3 shows the XRD pattern of the bivalve shells. All the investigated genera produce mainly aragonite biomineralization. Trace element concentrations in the bivalve shell are reported in Table 2. The most abundant metals were Zn (2.0–80 mg/kg), Fe (5.4–60 mg/kg), Pb (80–350 mg/kg), and Mn (0.5–4.5 mg/kg), having an ample concentration variability in all the analyzed samples. Nickel and Co showed less variable contents, ranging from 0.57–0.65 mg/kg and 0.30–0.33 mg/kg, respectively. Noticeably, S concentration is the highest, ranging from 1950 to 2250 mg/kg, similar among the investigated samples. Looking at the metal concentrations (Table 2), we found that the variability of Zn content as a function of genus is strongly correlated (r = 0.98, p = 0.02), suggesting the Zn concentration in bivalve shells could be a genus-dependent parameter.

Fig. 3
figure 3

XRD patterns of bivalves Donax, Glycymeris, Lentidium, and Chamelea collected along Buggerru (1, 2, 3, and 4) and San Nicolò beaches (5, 6, 7, and 8). A = aragonite

Full size image

Elemental distribution

Selected representative STXM-XRF images (TwinMic and I08-SXM instruments) of thin cross sections from the central part of the shells are shown in Figs. 4 and 5, these provide selected element distribution with micrometer resolution (Table 3). Zinc and Fe were found weakly correlated to Ca (ρZn-Ca = 0.23, ρFe-Ca = 0.08), pointing out that Zn and Fe were not homogeneously dispersed in the aragonite matrix but may likely form Fe/Zn rich specific phases (De Giudici et al. 2018b). Moreover, Zn and Fe were also weakly correlated (ρZn-Fe = 0.09–0.15) likely meaning they are located in different phases. The relatively high Ca–S correlation (ρCa-S = 0.47) suggested that S is more homogeneously dispersed into the aragonite structure and preliminary μ-XANES investigation at the S K-edge showed that S was mainly present as sulfate (unpublished data) according to Yoshimura et al. 2013.

Fig. 4
figure 4

Selected samples of Glycymeris collected at San Nicolò (a, b) and Donax collected at Buggerru (c, d). Ordinary light stereo-microscope image with the location of the acquired maps (a, c); bright field (absorption) images and LEXRF maps of Zn and Fe (b, d). Maps 1 and 2 of Glycymeris, size 80 × 80 μm2, scan 64 × 64 pixels; map 3 of Glycymeris, size 80 × 40 μm2, scan 64 × 32 pixels. Maps 1, 2, and 3 of Donax, size 80 × 40 μm2, scan 64 × 32 pixels; map 3 of Donax, size 80 × 80 μm2, scan 64 × 64 pixels

Full size image
Fig. 5
figure 5

Transmission image (a) and XRF maps of Chamelea (b, c) collected at I08 during SP16496 experiment with pixel size of ca. 800 nm. Ca K and S K-edges fluorescence signals are co-located in the section cut perpendicular to the shell

Full size image
Table 3 Elemental intensity correlation (Pearson) from STXM-XRF and TEM-EDS images analysis (see text). Values are averaged over several images, uncertainties are estimated around 5–10%
Full size table

TEM image (Fig. 6a) shows a fragment with irregular shape and diameter of ~ 600 nm. The corresponding selected area-electron diffraction (SAED) pattern (Fig. 6b) demonstrates that the fragment had the aragonite crystalline structure (ICSD card no. 034308). EDS mapping (Fig. 6c, d) of aragonitic grain fragments provides elemental distribution details with nanometric spatial resolution. EDS images show that S and Ca are collocated in the shell (ρCa-S = 0.52) accordingly to STXM-XRF analysis. Zinc and Fe are widely distributed in the shell, and Fe also can occur as concentrated spots (see arrows, Fig. 6c, d). Zinc and Fe clusters are evident in different regions (Fig. 6), explaining the low correlation coefficient (ρZn-Fe = 0.12). The Fe (ρFe-Ca = 0.29) and Zn (ρZn-Ca = 0.28) correlation with Ca is weak, confirming at the nanometric scale the presence of Fe/Zn rich specific phases.

If the incorporation of Zn or Fe is mediated by a given element, then their abundance should be higher around the sites of this element. In this case, a high spatial correlation is expected between the intensity of this element and that of Zn or Fe. The inhomogeneous distribution of Zn across each map results in generally low Zn–Ca correlations. In the b1 region (Fig. 4), the highest Zn concentration is observed in small clusters located at the very bottom of the map. Away from this area, the Zn appears rarely co-located with Fe in some regions (left-hand side in the map), whereas it is almost absent where most of the Fe is observed (top right area in the map). This inhomogeneity results in the above-mentioned weak Zn–Fe correlation. In b2 and d4 (Fig. 4), the Zn–Fe correlation is slightly higher (up to ρZn-Fe = 0.25) and can be visually predicted from the maps, this seems related to higher Fe concentration in these regions. In the sample of Fig. 4d, the Zn–Fe ratio is rather higher compared to the one observed in Fig. 4b. Furthermore, while Fe is more evenly spread on the surface, Zn appears to be mostly localized in small areas.

Fig. 6
figure 6

TEM image (a) and the corresponding SAED pattern (b) of a fragment from Chamelea from Buggerru; c, d STEM images with the EDS maps (in false colors) of Zn, Fe, S, and Ca. White arrows indicate Fe-concentrated spots

Full size image

Zinc chemical environment in the aragonitic shells

XANES analysis provides details on Zn average valence state and coordination chemistry. Here, Zn speciation in bivalve shells was investigated by XANES analysis in samples having the highest Zn concentration (Lentidium, Donax, and Chamelea from Buggerru). We selected an ample set of reference compounds, both inorganic and organic (Table 2, Fig. S1). Figure 7 presents Zn XANES spectra of bivalve shells and a selected subset of relevant reference compounds. The Zn K-edge XANES spectral features of bivalve samples show that Zn is incorporated as Zn2+ ions in the shells. The XANES spectral features are smoother and broader respect to the reference compounds (Fig. 7), likely due to several Zn phases and a more structurally disordered Zn environments in the shell samples. Comparing the Zn XANES spectral features of bivalve samples with those of the reference compounds, the similarity with the Zn phosphate spectrum is clear (Fig. 7, labels A–D). Specifically, the XANES features, labeled as A at the Zn phosphate edge white line, and the minimum labeled as D, were present in the bivalve spectra. In addition, the spectral features labeled as B and C in the Zn phosphate spectrum (Fig. 7) agreed with the spectral features observed in all the bivalve spectra.

Fig. 7
figure 7

XANES spectra at the Zn K-edge of Lentidium, Donax, and Chamelea collected at Buggerru and selected reference compounds: Zn phosphate (Zn3(PO4)2), hydrozincite (hdz), and Zn cysteine (Zn cyst). Spectra are vertically shifted for clarity

Full size image

Linear combination analysis (LCA) of XANES spectra (Benfatto and Meneghini 2014) was performed to determine the main Zn phases present in these bivalve samples. We selected the reference spectra for the LCA from the compounds reported in Table 2. We followed a trial and error procedure based on a statistical analysis of the best fit residue (χ2 test and F test) to determine the components giving the best agreement. Results are shown in Fig. 8. We found that in Lentidium and Donax the main contribution came from the Zn phosphate, followed by hydrozincite and Zn cysteine in the same amount. On the other hand, in Chamelea, Zn cysteine was the most abundant Zn specie, followed by Zn phosphate and hydrozincite (Fig. 9).

Fig. 8
figure 8

Linear-combination analysis of the XANES spectra. The dashed lines are the experimental data, the solid lines are the fit, and the gray lines are the residue

Full size image
Fig. 9
figure 9

Results of linear combination analysis of Zn XANES spectra. The sum of contribution fractions is fixed to 100%; the incertitude on the fraction values is around 5–8%

Full size image

Discussion

Bivalve shells are built by a biologically controlled process occurring at the interface between the cells and the environment, resulting in an organo-mineral biocomposite, of which the Ca-carbonate mineral makes up 95–99% (Weiner et al. 1976; Lowenstam and Weiner 1989; Arivalagan et al. 2017). Aragonite and calcite dominate the mineral composition of the bivalve shells, which may wholly be aragonitic or contain both aragonite and calcite in separate monomineralic layers (Taylor et al. 1969). Calcite was found in the outer layer of superfamilies belonging to the subclass Pteriomorphia, excluding two species from the Heterodont superfamily Chamacea. (Kennedy et al. 2008). The remaining two superfamilies, the Arcacea and Limopsacea, have wholly aragonitic shells (Taylor et al. 1969). The organic and inorganic constituents are secreted by the cells into the extrapallial fluid (EPF) that lies in the extrapallial cavity between mantle and shell (Kennedy et al. 2008). The calcium used to build the shell is taken from the diet or seawater. Carbonate is derived from the CO2/bicarbonate pool in the tissues of the animal (Lutz 2004).

The bivalve shell consists of three layers (Lutz 2004; Marin et al. 2012): a thin outer periostracum of horny conchiolin, a middle prismatic layer, and an inner nacreous layer. The periostracum is a quinone-tanned unmineralized protein layer that covers the external surface of the shell, enhancing abrasion resistance, and acting as barriers against predation (Yao et al. 2014), and it provides a nucleation site for calcium carbonate (Checa 2000). The prismatic layer grows below the periostracum (Yao et al. 2014) and it is secreted by the outer mantle fold. The calcite or aragonite prismatic layer contains organic matrices consisting of chitin (Suzuki et al. 2007) and proteins (Suzuki et al. 2004; Tsukamoto et al. 2004; Kong et al. 2009). Finally, the nacreous layer is deposited on the outer prismatic layer (Yao et al. 2014) by the general mantle surface (Lutz 2004). It is a micro-laminate composite material with highly oriented aragonite crystals. The organic matrix consists of several macromolecules, polysaccharides, proteins, and glycoproteins, located in both intercrystalline and intracrystalline positions within the nacre and constitutes 1–5% of the nacre shell weight structure (Yao et al. 2014).

Trace elements are actively incorporated in bivalve shells (Klein et al. 1996; Gillikin et al. 2005; Yan et al. 2014); these elements enter the EPF through the epithelial mantle cells via intracellular or intercellular transport ways (Klein et al. 1996; Gillikin et al. 2005) and affect the bivalve shell trace elements content (Yan et al. 2014). Despite all this extensive available information, the speciation of Zn in bivalve shells has not been investigated in the literature. In this study, we analyzed bulk shells and identified three main different Zn phases (Fig. 9): (i) Zn phosphate, (ii) hydrozincite, and (iii) Zn cysteine, motivating the low Zn–Ca correlation. Complementary XRF images of thin sections from the central part of the shells and TEM images of the aragonite (nano)crystals allowed us to determine that Zn was mainly located in the bulk of the aragonite crystal matrix (Figs. 5 and 6).

We recognized a difference in Zn chemical environment as a function of Zn concentration: Zn phosphate was found to be the most abundant species in the shells with the lowest Zn content (Lentidium, with Zn concentration of 29 mg/kg and Znphoph of 51%; Donax, with Zn of 27 mg/kg and Znphoph of 55%). George et al. (1978) found that the detoxication of Zn by the oyster Ostrea edulis (L) takes place by Zn immobilization in membrane-limited vesicles associated with P. In our study, P also seems to be an excellent ligand in the aragonitic shells. This is confirmed by our correlation calculations, which shows a high Pearson’s coefficient (ρCa-P = 0.54). Furthermore, Zn cysteine became more relevant in Chamelea, characterized by higher Zn concentrations (80 mg/kg). It has been shown that cysteine-rich proteins have an important role in metal detoxification, as they act as metal-chelating agents for the excess metals in the soft parts of mollusks (Géret et al. 2002). Thus, we formulated a hypothesis that Chamelea genus has a higher fraction (56%) of Zn cysteine than Lentidium (24%) and Donax (23%) genera, indicating that cysteine synthesis in bivalve cellules increases when the Zn uptake from the organism is higher and the physiological response is the incorporation of excess of non-toxic Zn in the shell. Previous researches (Adediran et al. 2016) demonstrated that the coordination of metals with cysteine-rich peptides can occur in different organisms such as fungi (Ahmad et al. 2002), sulfate-reducing bacteria (Cunningham and Lundie 1993; Holmes et al. 1995). and plants growing in Zn-extreme environments (Medas et al. 2017b, 2015; Terzano et al. 2008; De Giudici et al. 2015; Adediran et al. 2016). Biomineral processes occurring in polluted environments were referred to as active detoxification mechanisms (Carney et al. 2007; Caldelas and Weiss 2017).

Hydrozincite can be recognized in marine shells only via chemical selective/local structure-sensitive techniques such as XAS (De Giudici et al. 2018b) because of the low concentration of Zn, owing to the characteristic features of the Zn K-edge spectra of hydrozincite. Hydrozincite extracellular biomineralization has been found to have an important role in cyanobacteria preservation (De Giudici et al. 2014b; Medas et al. 2013, 2014a; Podda et al. 2014), and recent studies from our group (De Giudici et al. 2018b) indicated that Zn can occur as a separate hydrozincite nanostructured phase in foraminifera shells grown in heavily Zn-polluted environment (Zn concentration in sediments exceeding 1% w/w), while distribution and crystallinity depend on specific cellular processes. To our knowledge, revealing a hydrozincite biomineralization in mollusk shells is completely novel. Noticeably, we found such hydrozincite phase in shells (foraminifera in a previous study and bivalves in the present study) grown in highly Zn-polluted sediments; this suggests that the hydrozincite formation must have a role in Zn incorporation/detoxification mechanisms. Current data do not allow us to assess if hydrozincite in mollusk shells forms because of specific cellular machinery devoted to detoxification in heavily polluted environment or as a by-product of Ca-carbonate biological synthesis.

Theoretically, trace elements such as Mg, Sr, Zn, Mn, and Pb can substitute for the Ca2+ ion and thus become incorporated in the Ca-carbonate structure (Tynan et al. 2005). Previous researches (Soldati et al. 2016) found that, in different shells of freshwater bivalves, Mn was incorporated in the inorganic carbonate in the shell nacre of the internal region, and the structure of the aragonitic host was locally altered such that Mn attained an octahedral, calcitic coordination. The XANES analysis performed on the bivalve Arctica islandica indicated that Mg was not substituted into aragonite but was hosted by a disordered phase, e.g., organic components or nanoparticles of an inorganic phase (Foster et al. 2008), whereas Sr randomly substituted for Ca within the shell aragonite (Foster et al. 2009). In this study, Zn mainly occurred in independent Zn phases (ZnPO4, Zn cysteine, and hydrozincite) being related to the synthesis of the shells. Thus, we argue that the Zn in our samples was not incorporated into the aragonite lattice or adsorbed onto the surfaces of Ca-carbonate (nano)crystals.

Conclusions

This study showed that bivalves can synthesize phases other than Ca-carbonate when exposed to high Zn concentrations. The elemental distribution analysis (STXM and TEM) suggested that Zn was incorporated as independent phases at the micrometric as well nanometric scale. Zn K-edge XANES chemical speciation demonstrated that these phases involved cysteine molecules, phosphate ions, and carbonate ions to regulate trace element concentration. The relative amount of Zn-rich phases are likely related to specific biogeochemical reactions responding to Zn incorporation. These findings support the relevant role of biominerals in acting as critical detoxification sinks within certain organisms to remove potentially toxic elements from their immediate environments (Carney et al. 2007).

We were not able to assess if the Zn-independent phases were produced by the investigated mollusks, exposed to high availability of Zn, as a by-product of the molecular machinery during the Ca-carbonate synthesis or due to specific tolerance mechanisms. We cannot conclude that Zn-independent phases can also be formed in bivalve shells having a Zn content lower than 30 ppm due to the weakness of the XANES signal, which would require an experimental setup specifically addressed to ultradiluted compounds.

Future studies should comprise (i) combined analysis of the soft tissues and shells to observe if potential correlations between the metal content and Zn coordination environment occur, and (ii) XAS microbeam analyses at a micro-nanoscale on shell cross-sections to clarify the chemical environment of Zn within the shells of these marine bivalves and to understand the mechanisms involved in Zn incorporation.