Ria Formosa is a shallow, meso-tidal coastal lagoon located on the south coast of Portugal (Fig. 1). This lagoon is a paradigmatic example of an ecosystem with ecological and economic importance although it is under stress from human activities and displays alterations of its morphology reflecting its fragility to extreme natural conditions, such as sea storms (Cunha et al. 2005, Guimarães et al. 2012, Vila-Concejo et al. 2002). The lagoon has several channels and a large inter-tidal area covered by sand, muddy sand-flats, and salt marshes (Falcão and Vale 1990). It extends for 55 km along the coastline and has a maximum width of 6 km (Newton and Mudge 2003). The hydrodynamics is forced mainly by the semidiurnal tidal regime exchanging water to the sea through several inlets (Nobre et al. 2005). Except in periods of heavy rains, the freshwater input to the lagoon is negligible, salinity remaining approximately 36 all year long (Águas 1985, Falcão and Vale 1990). Spatial distribution of nutrients and chlorophyll a is highly influenced by internal processes, such as sediment-water interactions, and exchanges with the sea (Cravo et al., 2014, Falcão and Vale 1990, Newton et al. 2003). In addition to diffuse sources in the watershed, discharges of domestic effluents and boat traffic increase drastically in summer with tourism (Cravo et al. 2012, Mudge and Bebianno 1997). Several studies have documented the trace element contamination in sediments (Bebianno 1995, Caetano et al. 2002, Cortesão et al. 1986), in water (Caetano et al. 2007, Falcão and Vale 1990) and in bivalves, namely clams (Bebianno and Serafim 2003, Cravo et al. 2012, Serafim and Bebianno 2001). Ria Formosa is traditionally used for the production of the clam R. decussatus, which represents 90% of the national production of bivalves. The production of R. decussatus is 47% of total aquaculture production in Portugal (DGRM 2013). The decline of clam production in the last two decades has been attributed to a combination of factors, such as clam mortalities probably related to pathologies and less favourable environmental conditions (Matias et al. 2011).

A field experiment was performed with the clam R. decussatus in inter-tidal ground plots. Clams were collected from a natural recruitment area (N) and transplanted to sand squares of 25 m² with a clam density of 1 kg m^–2 at the three sites (Fig. 1). Site F is located in a clam growth ground near Faro city and is consequently influenced by urban activities. Site O is in a clam ground that receives the wastewater of the Aquaculture Research Station (7 ha) through a small tributary that increases its flow in winter in response to precipitation. In contrast to the locations of these inner sites, site L is positioned in an inter-tidal clam ground of the inlet channel Lavajo that connects the lagoon to the adjacent coastal waters. Transplantation of clams started on 23 November 2010 and sampling ended on 25 October 2011.

The upper layer of the clam ground substrate (0-2 cm) was collected from sites F, O and L (Fig. 1) in November 2010 and March, August and October 2011. Samples were homogenized and separated into three sub-samples kept at 4°C for the determination of loss in ignition, grain size, and the following elements: aluminium (Al), chromium (Cr), zinc (Zn), copper (Cu), nickel (Ni), lead (Pb), arsenic (As), cobalt (Co) and cadmium (Cd). Elements were quantified in sediment samples that had been previously freeze-dried. Transplanted clams with a mean length of 2.8±0.4 cm were collected in February, May and August 2011 from the sites F, O and L and carefully dissected to remove digestive glands, gills, mantle plus siphons, and remaining tissues. Three pools of composite samples (n=10 specimens) were prepared, weighed and frozen at –20°C. Samples were freeze-dried, ground and homogenized for determination of Al and trace elements. Forty clams were collected monthly from each transplanted site and placed in filtered seawater (0.45 μm) at 20°C in the laboratory for 24 h to purge their stomachs. Twenty clams were stored at –20°C to determine condition index and gross biochemical composition (proteins, glycogen and total lipids). In another set of 20 individuals, the visceral mass was separated from siphons and gills and fixed in the Davidson solution for 48 h, and then transferred to 70% ethyl alcohol for storage until histological analysis to evaluate the gametogenic stages. Osmolarity and pH were determined in the hemolymph collected immediately after sampling from the foot and adductor muscle of ten clams with a 21G needle and 1-mL syringes. Then, it was centrifuged for 10 min (4500 g, 4°C), snap-frozen in liquid nitrogen and stored at –80°C until analysis (Gustafson et al. 2005).

Statistical analyses were performed using SigmaPlot13.0 (Systat Software Inc.®) for Windows with a p-value of 0.05. Whenever the pre-requisites of normality and homogeneity of variances were fulfilled, an ANOVA test was performed to determine the significance of the differences found between sites or sampling dates for the sediment characteristics, biological parameters and trace elements in clam tissues. Principal component analyses (PCAs) were performed using Statistica 8.0 (StatSoft®). Conventional descriptive multivariate analyses were used to summarize high-dimensionality data in a simpler two-dimensional space (Abdi and Williams 2010). The aim was to interpret major trends of similarities between sites, clam tissues and sampling dates, using correlation-based element variables. Sample matrix of the whole dataset was composed of 45 tissue samples and 9 descriptors (As, Cd, Cr, Cu, Mn, Ni, Pb, V and Zn). PCA was also applied to digestive samples and remaining tissue samples.

The major characteristics of the sediment samples collected in the inter-tidal ground plots F, O and L are presented in Table 2. The sediments consisted of sand (>63 µm) with a small fraction of fine particles. The proportion of coarser particles (mean±SD) ranged from 87%±3.8% (site F) to 100%±0.07% (site L). Accordingly, Al content was low in sediments from the three sites, although it was significantly (p<0.05) higher at sites F and O than at site L. Loss on ignition varied from 0.50% to 2.0% and was significantly (p<0.05) higher at site O than at site L.

Due to the narrow variation of Al content and its low values, trace element concentrations in sediments from the three sampling sites were compared without normalization with Al commonly used recurrently to minimize differences related to grain size (Windom et al. 1989). Significant differences of trace element concentrations among sites are presented in Table 2. Copper concentration (µg g^–1) was significantly (p<0.05) higher at site O (6.6±0.78) than at sites F (4.2±0.61) and L (1.1±0.24). A different distribution was encountered for Cd and was significantly (p<0.05) higher at site F (0.035±0.0045) than at sites O (0.015±0.0042) and L (0.0091±0.0017). Chromium in sands from site L (30±1.5) displayed significantly (p<0.05) higher values than in sands from sites F (19±1.8) and O (13±1.6).

Maturation stages over an annual cycle of R. decussatus transplanted to sites F, O and L are presented in Figure 2. The histological phase II, indicating the beginning of the gametogenesis, was first observed in January at the three sites. The histological phase V, indicating partially spawned development, started in May (sites F and O) and June (site L), and was highly marked in July and August at the three sites. During the spawning season (May to August), microscopic observation of gonads revealed simultaneous spawning (phase IV) and partial spawning (phase V). This observation was clearer in clams from the inner sites F and O. Most of the clams had already spawned in October and remained inactive thereafter.

Table 3 gives the monthly mean values and SD of the condition index, proteins, glycogen and total lipids of clams from each transplanted site over the annual survey cycle. From January to July (site F) or to August (sites O and L), which corresponds to the reproductive period, CI varied between 6 and 11, and then remained within the interval 4-7 until October (sites F and L). Considering all sampling dates, no significant (p<0.05) differences were found among the three sites. Clams transplanted to sites F, O and L in November showed lower protein content (102±26 μg mg⁻¹) than those transplanted in subsequent months, reaching a maximum ranging from 378±50 μg mg⁻¹ (site F) to 436±73 μg mg⁻¹ (site L). Clams from site O showed significantly (p<0.05) higher protein contents than those from sites F and L. Mean glycogen varied between 9.0±5.0 and 19±7.1 μg mg⁻¹ (site F), 5.9±1.1 and 16±7.6 μg mg⁻¹ (site O), and 6.5±1.4 and 25±11 μg mg⁻¹ (site L), differing (p<0.05) significantly among the sites. Mean total lipids were statistically (p<0.05) higher atsite L (27±9.7 to 104±47 μg mg⁻¹) than at sites F (17±7.1 to 84±20 μg mg⁻¹) and O (31±10 and 72±30 μg mg⁻¹).

Hemolymph osmolarity of clams from the three sites varied within a narrow interval (mean values of between 1.0 and 1.2 mOs m kg^–1) over the annual surveyed period. Mean pH values of clam hemolymph ranged from 7.6 to 8.2, and no significant (p>0.05) differences were observed among the three sites. Clams from sites F and L exhibited significantly (p<0.05) lower pH from January to September and from May to July, respectively. A less clear variation was found at site O, and pH was only significantly higher (p<0.05) in November.

Table 4 gives the mean concentrations (±SD) of Zn, As, Cu, Mn, Ni, V, Cr, Pb and Cd in digestive gland, gills, mantle plus siphons, and remaining tissues of the clams from the recruitment area (N) collected in November and of the transplanted clams collected in February, May and August 2011 from sites F, O and L. Significant differences among sites in trace elements of each clam tissue are marked in Table 4. Remaining tissues displayed differences in more elements (Zn, Cu, Cr, Pb and Cd) than digestive glands (Cu, Ni, Cr). Gills and mantle plus siphons showed differences only for Zn and Cu, respectively. In those ten statistical differences, eight enhanced concentrations were observed at site O, while only five were observed at sites F and L.

In order to have a better integration of trace element partitioning among the four clam parts/tissues, a PCA was applied to the whole dataset (Fig. 3, PC1=43.7% and PC2=23.1%). Samples of the four analysed tissues were generally projected in different quadrants of the plan PC1-PC2; the digestive gland was projected closer to the trace elements and the mantle plus siphons far from these descriptors. Arsenic, Cu, Mn, V, Cr and Pb were significantly (p<0.05) higher in digestive gland than in gills. Cadmium was the exception, being more abundant in gills. Zinc, As, Cu, Mn, Ni and V were significantly (p<0.05) higher in the digestive gland than in the remaining tissues. Another two PCAs were applied separately to digestive gland (Fig. 4A, PC1=34.5% and PC2=20.0%) and to remaining tissues (Fig. 4B, PC1=48.0% and PC2=26.5%). The analysis of gills and mantle plus siphon data are not presented due to the lower variance obtained (PC1+PC2 <50%). The digestive gland plot shows a clear difference between February and May/August, because all February samples are closer to the trace element projections, while May and August samples are represented in quadrants II and III. This dichotomy between winter and late spring/summer was not observed at site O, because May samples are represented near the winter samples. The remaining tissues plot also shows a similar separation between winter and late spring/summer. In addition to this temporal variation, samples collected in May and August at site O are projected separated from samples of the other sites and associated with Ni, Cu, As and Zn.

The elemental composition of estuarine and coastal sediments has long been used to identify possible sources of contamination because of the high affinity of trace elements to particles, namely those with a large specific area and covered by organic matter films (Libes 1992). Despite the sandy nature of the substrates (87-100% sand) the almost negligible proportion of fine particles at site L appears to explain the lower concentrations of Zn, Cu, Ni, Pb and Cd in comparison with sites F and O. Conversely, the enhancement of Cr in sands of site L was most likely due to natural variation related to the lithology, as observed in sandy beach nourishment in other areas of Portugal (Vale, unpublished data). Other factors might account for the slight differences observed at sites O and F (Table 2). The significantly (p<0.05) higher Cu concentrations in the substrate of site O in comparison with sites F and L may be interpreted as the influence of localized sources of contamination. Most likely, wastes of food artificially enriched with Cu that is supplied daily to fish produced in the Aquaculture Station located near site O is partially retained in the lagoon sediments localized nearby. In addition to the regular supply of food, copper sulphate is added occasionally to the aquaculture tanks to eliminate fish parasites that develop mainly in spring, which involves a pulse release of Cu to the ecosystem. The possible influence of a small tributary draining agriculture fields that ends in the area should also be considered. Due to the torrential precipitation regime, discharges may occur during short periods in winter (Falcão and Vale 1990). However, freshwater discharge into the lagoon should not have been dramatic in the year studied because osmolarity in clams, which is a parameter sensitive to salinity, did not differ among individuals from the three transplanted sites. The enhanced Cu values found in the present study is in agreement with previous studies in Ria Formosa showing higher Cu/Al ratios in sediments near the site O (Caetano et al. 2002, Cortesão et al. 1986). Enhanced concentrations of Cd and Ni at site F in comparison with sites O and L probably result from human activities related to Faro city, whose population increases pronouncedly in summer due to tourism (Cravo et al. 2012).

Copper, Zn and Cd concentrations measured in gills and digestive gland of R. decussates in the present work were lower than the values reported by Serafim and Bebianno (2001) and Bebianno and Serafim (2003) Cu, and Zn was followed in different tissues (gills, digestive gland, and remaining tissues for the same species from Ria Formosa. Cadmium concentrations in this current work (0.10-0.56 µg g^–1) were one order of magnitude lower than the values reported by Serafim and Bebianno (2001) (1.4±0.19 µg g^–1) and far below the limit recommended for human consumption, 5 µg g^–1 dry weight (obtained by the conversion of 1 µg g^–1 wet weight, assuming 80% as the mean percentage of water in clam flesh) (EC 2001). This decrease suggests a lower availability of this element in food or water over a period that exceeds one decade. Most likely, the reduction of anthropogenic inputs to Ria Formosa contributed to that difference (Cravo et al. 2012). Copper concentrations showed a less marked decrease in comparison with previous studies, and large variations were encountered among the sampling periods and sites (Table 4). Values of Zn observed in the present work (81-99 µg g^–1) are comparable to those reported by Serafim and Bebianno (2001) (98±8.6 µg g^–1), possibly reflecting the metabolic regulation of Zn concentrations in bivalves (e.g. Bryan et al. 1979).

The similar size and condition index of clams at the three transplanted sites are in favour of comparable filtration rates and hence trace element concentrations (Reinfelder et al. 1998). Nevertheless, trace elements in clam tissues, particularly Cu and Cr, displayed statistical differences (p<0.05) among the three sites (Table 4). Spatial differences were observed more recurrently in remaining tissues and digestive gland. Spatial availability of trace elements in water and food is a plausible cause for the observed variation in digestive gland concentrations. For example, enhanced values of Cu in site O may result from aquaculture practices, which is in line with the sediment composition. Concordance between substrate and clam was less clear for other elements, such as Cr and Cd.

Filtration rate may also have changed in clams of the three sites due to its sensitivity to contaminant exposure. Bioturbation in clam production grounds contributes to the release of trace elements and nutrients from sediments (Falcão et al. 2006). However, bioturbation is not expected to differ considerably among the transplanted sites because clam density and mortality along the experiments were similar. Despite the renewal of large water volume during spring tides, which promotes a vigorous mixing of water and particles inside the lagoon (Falcão et al. 2006), the PCA of the digestive gland data suggests higher concentrations of most elements in February, particularly at site F. Enhanced values were also observed for remaining tissues in February, which emphasizes the importance of diffusive sources in winter for the accumulation of trace elements by clams. Indeed, the precipitation regime in the region was marked by episodic periods of heavy precipitation (Lima et al. 2013, Soares et al. 2015) and consequent runoff from the city and agriculture fields. The increase inhuman population and leisure activities in summer may have led to higher availability of trace elements in Ria Formosa (Cravo et al. 2012). It may also be hypothesized that internal fluxes of trace elements contribute to the increase in their availability in warmer periods. However, the results obtained suggest the predominance of the rain effect in winter. Previous works have shown that nutrient fluxes from sediments increase with temperature, as well as phytoplankton biomass (Falcão and Vale 1998). In confined areas of the lagoon, dissolved oxygen in the water column tends to be consumed during the night, decreasing to under saturated values. The seasonal decrease of pH in clam hemolymph, which may be attributed to less oxygenation of the water (McFarland et al. 2011), is in line with the reduction of dissolved oxygen during the night (Falcão and Vale 1998) and consequently the internal input of trace metals to the water column.

Despite the presence of trace elements in the ingested food, the affinity of most determined elements with the digestive gland may be followed by mechanisms that facilitate their retention in this tissue. The sequestration of the elements may be enhanced by the induction of metallothioneins if they are present above a threshold value (Bebianno and Serafim 2003). Cadmium appears to be linked with gills and at a lower level with digestive gland (Fig. 3), which may be influenced by the higher stability of Cd chloro-complexes in seawater (Simões et al. 1981). In addition to the passage of dissolved Cd through gills during filtration, the presence of metallothioneins in that organ (Bebianno et al. 1994) may also contribute to the retention of Cd.

The reproductive cycle of R. decussatus observed in the present work is in agreement with previous studies on this species at Ria Formosa (Matias et al. 2013, Pacheco et al. 1989). The cycle was characterized by the onset of gametogenesis in January followed by the rapid development of gametes (Fig. 2). Clams reached the highest sexual maturity in May-June (sites F and O) and in June-July (site L). In late spring/summer, condition index and total lipid content remained high, decreasing sharply as rest stage was observed. This change was better observed in clams transplanted to sites F and O. The spawning was expanded throughout summer until early autumn, clams ensuring a consistent supply of gametes. Indeed, histological analyses showed simultaneously gonias, maturing gametocytes and variable proportions of fully matured gametes in some individuals. The co-occurrence of spawning and gonad recovery (phase V) during a long period of the year, concomitantly with nutrient storage and consumption, pointed to high reproductive effort of clams with a high capacity for gonadal regeneration. The continuous supply of gametes mirrors an advantageous strategy of this bivalve species (Matias et al. 2013).

Variations in reproductive output of bivalve molluscs are highly sensitive to environmental perturbations (Weinberg et al. 1997). In particular, accumulated contaminants in bivalves could deplete energy reserves that were initially intended for growth and reproduction (Capuzzo et al. 1988). Under stressful circumstances, energy reserves are allocated to defence mechanisms, so animals living in contaminated areas often show poor tissue condition and retarded growth (Nicholson and Lam 2005). The period of spawning and gonad recovery found in clams transplanted to the inner sites F and O (May-October) was longer than in clams of site L (July-October) located in the inlet channel. Probably, the accumulated trace elements have negligible effects on the gametogenesis of R. decussatus, even in inner areas of the lagoon that may be affected by localized or diffuse sources of contamination. The successive production of gametes and spawning seems to obscure a trace element pattern during the sexual maturity period. In fact, the PCA indicate dissimilarities between samples of May and August and trace element projections (Figs 3 and 4). However, condition index and total lipids of clams after spawning were better at site L than at sites F and O.

These results indicate that trace element partitioning in R. decussatus produced in extended inter-tidal flats of Ria Formosa are influenced by environmental changes in winter, most likely associated with rain and diffuse sources. Despite the massive tourism in spring and summer, and the consequent increase of domestic effluent discharges, trace element concentrations in clam tissues are low, presumably influenced by the long period of production of gametes and spawning.