Introduction

The Gulf of Mexico, including the Western coast of Florida (FL), is frequently affected by red tide events caused by the proliferation of the athecate dinoflagellate Karenia brevis. Occasional outbreaks have also been reported on the US Atlantic coast (Tester and Steidinger 1997) and in New Zealand (MacKenzie et al. 1995). Karenia brevis produces lipid-soluble polyether neurotoxins called brevetoxins (Ptychodiscus brevis toxin, i.e. PbTx) and its proliferation is associated with significant environmental, human health and economic impacts. Brevetoxins cause massive fish kills at concentrations as low as 250 cells ml−1, can affect human health through ingestion of contaminated shellfish (Kirkpatrick et al. 2004), and induce adverse respiratory symptoms due to toxin released in marine aerosols (Pierce 1986; Pierce et al. 2003). Blooms of K. brevis are also associated with mass mortalities of marine mammals such as the endangered Florida manatees and bottlenose dolphins (O’shea et al. 1991; Flewelling et al. 2005).

The threat to marine ecosystems due to harmful algal blooms (HABs) in general has prompted research on control and mitigation methods (Anderson 1997). One of these methods, involving the use of ecologically inert clays to induce sedimentation of harmful algae, has the potential to be environmentally benign and cost-effective. It has been successfully used at field aquaculture sites in South Korea to control fish-killing blooms of Cochlodinium polykrikoides (Na et al. 1996). Sengco et al. (2001) investigated the effectiveness of 25 clays against the dinoflagellate Karenia brevis (Florida red tides) and the picoplanktonic pelagophyte Aureococcus anophagefferens (brown tides), two major HAB species occurring in Atlantic waters. Phosphatic clays were found to be particularly efficient in removing K. brevis from the water column at a concentration as low as 0.04 g dry weight l−1 (80% removal efficiency) and 0.25 g dry weight l−1 (90% removal efficiency), but clay did not significantly remove A. anophagefferens (<60% removal efficiency with clay at 0.04 to 2 g dry weight l−1). Further experimental studies in the laboratory confirmed the effectiveness of phosphatic clay for removing K. brevis cells as well as its extracellular toxins from the water column (Pierce et al. 2004). Clay dispersal might provide a potential method to control recurrent blooms of the dinoflagellate K. brevis in nearshore Florida coastal waters. However, the biological impacts of this strategy need to be assessed before it can be used at any significant scale on natural bloom populations. Previous studies investigating the effect of clay on shell and tissue growth rates of the hard clam Mercenaria mercenaria (Archambault et al. 2004) suggested that a high-energy environment leading to prolonged in situ resuspension of clay might affect suspension-feeding communities. Lewis et al. (2003) investigated the detrimental effect of the K. brevis/clay flocculates compared to the toxicity of K. brevis alone on several benthic species and showed that clay flocculation would neither increase, nor decrease toxicity to benthic organisms relative to that attributable to an untreated bloom.

The mechanism underlying the clay mitigation strategy is the entrainment, flocculation and sedimentation of clay-cell aggregates following clay dispersal at the surface (Sengco et al. 2001). This method transfers the algae from the water column to the bottom sediment and could thus alter the fate of PbTxs in the food web (Fig. 1). During a K. brevis bloom, toxins are primarily transferred to suspension-feeding bivalves that filter out large volumes of water and accumulate PbTxs in their tissues (Poli et al. 2000; Fletcher et al. 1998; Plakas et al. 2002). Suspension-feeding bivalves include many commercially important species, and ingestion of contaminated shellfish is responsible for neurotoxic symptoms (i.e. NSP, Neurotoxic Shellfish Poisoning) and acute gastroenteritis in humans (Kirkpatrick et al. 2004). Direct toxin transfer from shellfish to humans is therefore a major public health concern. Another important concern is the possibility of food web toxin transfer via benthic invertebrates such as blue crabs and whelks (Poli et al. 2000), or omnivorous and carnivorous fish (striped burrfish, longnose killfish, pinfish, spot, puffer fish) that feed on bivalves. Flewelling et al. (2005) indeed demonstrated that massive dolphin mortalities in Florida were due to consumption of brevetoxin contaminated fish that can accumulate toxins from their food source. This routing of toxins through secondary consumers might potentially lead to human contamination but no evidence of such poisoning has been reported. Clay treatment of K. brevis blooms could reduce direct public health impacts by changing the fate of the PbTxs from suspension- to deposit-feeders as those are generally not commercially exploited for human consumption. However, the ecological impact due to food web transfer from deposit-feeders to higher trophic levels needs to be assessed. This risk assessment will primarily depend on the ability of deposit-feeders to accumulate and retain significant concentrations of PbTxs in tissues, which has not been previously demonstrated.

Fig. 1
figure 1

Schematic flow diagram illustrating potential food web pathways of brevetoxin-producing Karenia brevis cells: sedimentation of toxic cells following inert clay application could reduce toxin incorporation by commercially important suspension-feeding bivalves (indicated by the striped X) and thus prevent direct public health impacts, but could potentially lead to brevetoxin (PbTx) accumulation by benthic deposit-feeders. Striped arrows indicate unknown pathways prior to this study. Drawing shows a tellinid bivalve; is and es = inhalent and exhalent siphons, respectively

Full size image

The main objective of this study was therefore to determine whether surface deposit-feeding could provide a pathway for toxins from deposited clay-K. brevis aggregates. Toxin uptake via deposit-feeding was investigated and compared to that via suspension-feeding using the same model organism, the tellinid bivalve Macoma balthica. This species is mainly a deposit-feeder but can facultatively adopt suspension-feeding in response to flow and food availability in the water column (Hummel 1985; Levinton 1991). We determined the toxic effects of K. brevis on both feeding modes and compared the effectiveness of brevetoxin-transfer. For both feeding modes, Karenia brevis was offered alone or mixed with a non-toxic diatom (Fragilaria familica) to test whether the presence of non-toxic algae stimulates feeding activity [as previously shown by Bricelj et al. (1991) and Bardouil et al. (1996) for the toxic dinoflagellate Alexandrium tamarense] and to investigate conditions of maximal toxin incorporation.

Materials and methods

Test organism and algal culture

Macoma balthica was chosen as a model species for tellinids, the main family of deposit-feeding bivalves (Ward and Shumway 2004), represented by species such as Tellina versicolor, Tellina lineata or Macoma constricta in Florida (Camp et al. 1998). Clams were collected from an intertidal mudflat of the Minas Basin, Bay of Fundy, along the northwestern shore of Nova Scotia, Canada. They were acclimated to experimental conditions in a flow-through raceway provided with ambient seawater at 20°C and were maintained in sieved (500-μm mesh) natural sediment from the sampling site. Natural benthic microflora present in the sediment was supplemented once a week by sedimentation of the cultured diatom Fragilaria familica. This local species isolated from Mahone Bay, Nova Scotia, was chosen because diatoms constitute the main component of M. balthica’s diet (Rossi et al. 2004). We also observed that once added in the water this diatom rapidly settled at low flow and formed a surface biofilm that was readily ingested by M. balthica via deposit-feeding.

The toxic dinoflagellate Karenia brevis (Wilson strain, CCMP 718) was grown in 2.8 l Fernbach flasks containing autoclaved filtered seawater (0.22 μm millipore membrane filters) enriched with L1 nutrient medium (Guillard and Hargraves 1993). Algae were grown on a 14:10 h light–dark cycle, at 22°C and 30‰ salinity. The non-toxic dinoflagellate Heterocapsa triquetra (CCMP 449) was batch cultured in aerated 20 l carboys in L1 medium at 16°C and a 14:10 h light–dark cycle. Fragilaria familica was batch cultured in aerated 20 l carboys in F/2 + Si medium, at 16°C, under continuous illumination. All cultures were harvested in late-exponential phase.

Effect of Karenia brevis on suspension- and deposit-feeding and comparison of brevetoxin transfer effectiveness

The objective of this experiment was to (1) determine whether clay treatment may provide a trophic pathway for toxins by deposit-feeding of settled clay-cell aggregates and (2) to compare brevetoxin trophic transfer from deposited aggregates (deposit-feeding following clay treatment) to that from suspended cells (suspension-feeding during an untreated bloom). For this, clams (∼13 mm shell length, SL) were exposed to suspended (∼300–400 cells ml−1) or settled K. brevis cells (loading of ∼3 × 105 cells cm−2, equivalent to removal of 300 cells ml−1 from a 1 m-deep water column) and sampled for toxin content after 24 h. This interval was chosen to determine whether deposit-feeders were able to accumulate a significant level of PbTxs over a short period of time as we assumed that in the field, winds and currents would rapidly spread the clay layer following deposition.

The effect of Karenia brevis on Macoma balthica’s suspension-feeding behavior was assessed by measuring the clams’ individual clearance rate (volume swept clear of particles per unit time). Three experiments were performed in feeding chambers (450 ml capacity, without sediment) placed in a recirculating system (Fig. 2). The algal suspension was delivered from an aerated 60 l header tank. Cell concentration was monitored using a Beckman-Coulter Multisizer (100 μm aperture), and algae were added to maintain the cell concentration within the target range in the system (300–400 K. brevis cells ml−1). Cell concentration was monitored every 30 min during the 2 h preceding each CR measurements and every hour the rest of the time. During nighttimes, the system was supplemented with algal culture with a peristaltic pump, and cell concentration was adjusted to the target range in the morning when necessary. The suspension was gravity-fed from the header tank (∼120 ml min−1) and outflow from the chambers was collected in a trough and returned to the header tank with a Masterflex peristaltic pump. The flow generated by the pump was gentle enough to maintain the integrity of dinoflagellate cells. Two chambers without animals were used as controls for cell settlement or lysis and five additional chambers contained five juvenile clams each (13.58 mm SL ± 0.85 standard deviation, SD). Algae were held in suspension with motor-driven magnetic stirrer mounted at the top of the chamber, allowing mixing without resuspension of biodeposits. Before the experiments, clams were allowed to clear their gut of sediment in filtered seawater for 24 h. Experiments were conducted in a temperature-controlled walk-in environmental chamber held at 20°C.

Fig. 2
figure 2

Experimental design of the suspension-feeding experiments. These were performed in seven feeding chambers placed in a recirculating system. Clams (n = 5) were placed in five feeding chambers and two chambers were used as controls for cell settlement

Full size image

Clearance rate (CR) was determined from the depletion of cells over a 60 min incubation (period corresponding to ∼30% depletion). CR was estimated from the equation: CR = (LnC0-LnCf) × V/t where C0 and Cf = initial and final cell concentration (cells ml−1), respectively, V = volume of suspension (ml) and t = time interval (h). CR was standardized to the mean dry body mass of experimental animals, equal to 22.91 mg, using the following equation (Bayne et al. 1993): CRs = CR × (22.91/W)b , where CR = measured clearance rate and W = dry body mass, and b = slope of the allometric equation relating to CR and body mass. The value of 0.6 was used for b, corresponding to that determined for the tellinid, facultative suspension- deposit-feeder, Scrobicularia plana (Hughes 1969).

For each experiment, the clams were exposed successively to a volume equivalent concentration of the conditioning suspended diet of Fragilaria familica (18 × 103–21 × 103 cells ml−1, equivalent spherical diameter, ESD = 7.52 ± 2.5 μm SD) and an experimental diet: (1) the non-toxic dinoflagellate Heterocapsa triquetra (800–900 cells ml−1; ESD = 17.19 ± 1.55 μm SD) (2) the toxic dinoflagellate Karenia brevis (300–400 cells ml−1; ESD = 21.96 ± 2.78 μm SD) or (3) a 50:50 mixture of K. brevis and F. familica. Heterocapsa triquetra was used as a control to test whether the clams were affected by a change in diet from diatoms to non-toxic dinoflagellates. Mixed versus unialgal suspensions were used to determine conditions of maximal toxin incorporation. After 2 h of acclimation, two consecutive CR on the conditioning diet were determined; the animals were then switched to the experimental diet for 24 h with measurement of CR at t2h, t4h and t24h. The K. brevis stock was replaced 4 h before the t24h measurement. Finally clams were exposed again to the conditioning diet to determine whether they could recover their initial CR. For diets (2) and (3), five replicates (1 × 106 cells each) of the initial K. brevis cell suspension were sampled for toxin analysis and two chambers were sampled after toxic exposure to determine toxin concentration in clam tissues. Soft tissues were dissected, rinsed with filtered seawater and weighed (wet weight). Tissues of the five individuals contained in each chamber were pooled and stored at −80°C until toxin analysis.

For the deposit-feeding experiment, juvenile clams (average size 13 mm SL ± 0.7 SD) were exposed to the following three treatments: (1) 3 × 105 K. brevis cells cm−2 of substrate (Wilson strain, CCMP 718), equivalent to removal of 300 cells ml−1 from a 1-m-deep water column, (2) a mixed diet of ∼3 × 105 K. brevis cells cm−2 and a 20% volume equivalent concentration of F. familica, or (3) a control diet of F. familica (at a concentration equivalent in cell volume to treatment 1). For each treatment, 3 microcosms (made out of Plexiglas, W6 × L12 × H8 cm3) were placed in a recirculating system. Microcosms were filled with a 2 cm bottom layer of muddy substrate obtained from the clam sampling site and pre-sieved through a 500 μm sieve. The application of toxic cells and clay was repeated twice to attain the desired bottom loading. Each time, 400 ml of suspension at the appropriate concentration was added to the microcosms and treated with phosphatic clay (IMC-P2, International Mining Corporation, Bartow, FL, USA) at a concentration of 0.25 g dry weight l−1. Five replicates (1 × 106 cells each) of K. brevis suspension were previously sampled for toxin analysis. Once K. brevis cells had completely settled, the water column in each microcosm was sampled to estimate the removal efficiency. After renewal of water by flushing, five clams were placed in each microcosm and allowed to burrow; t0 was taken once all the animals had burrowed (∼1 h). After 24 h exposure, animals were collected, soft tissues dissected, rinsed with filtered seawater and weighed (wet weight). The five individuals contained in each chamber were pooled and stored at −80°C until toxin analysis.

Toxin extraction and analysis

Toxins from K. brevis cells were extracted by elution of water samples through C-18 solid phase extraction tubes (Supelco) and eluted from the tubes with 15 ml methanol under gentle vacuum. The methanol was dried in a rotary-evaporation unit, and samples were then brought to a final volume of 3 ml methanol.

Clam tissue samples were extracted in 3–5 ml methanol using a Brinkmann Polytron homogenizer. Slurries were centrifuged (10 min at 3,000 rpm), and the supernatant was collected and dried in a rotary-evaporation unit. The dried residues were re-solubilized in acetone at a concentration of 0.2 g to 1 g wet tissue ml−1 by bath sonication and vortex mixing.

Pre-weighed sediment samples were first centrifuged at ca. 2,500 rpm for 2 min to discard any remaining water. Toxins were extracted in ca. 10 ml acetone using an ultrasonic probe. Samples were centrifuged and the supernatant was decanted into a flask. The acetone sonic extraction and centrifugation was repeated until the solvent layer was colorless and successive extracts were pooled in the flask. The extracts were then dried in a rotary-evaporation unit and re-solubilized in 3 ml acetone.

The toxin content of K. brevis cells, sediment and clam tissues was analyzed by competitive Enzyme-Linked Immunosorbent Assay (ELISA) as described in Naar et al. (2002). This method allows detecting brevetoxins accurately in a variety of sample types without matrix effects at a level as low as 4 ng PbTx g−1. The anti-brevetoxin goat polyclonal antibodies used are specific for the last H–K rings of the PbTx type-2 brevetoxins (PbTx-2, PbTx-3, PbTx-9) and recognize all of these compound derivatives (Naar et al. 2004; Dickey et al. 2004; Plakas et al. 2004). As ELISA analysis cannot differentiate between toxins present in the sample tested, results are expressed as PbTx-3 equivalent and reflect the overall concentration of brevetoxins and brevetoxin-like compounds present in the sample.

Statistical analysis

Measurement of clearance rates (CR) during suspension-feeding experiments consisted of repeated measurements over time from the same sampling units (i.e. 5 chambers containing 5 individuals each). Such repeated sampling of the same population might introduce risks of temporal non-independence of data and statistical analysis can proceed only if it can be assumed that there is no interaction among times and measurements (Underwood 1997). This interaction might be expected to occur during the toxic treatments (suspension-feeding with Karenia brevis and the mixed diet) as CR may be more affected with increasing exposure time. Interactions among time and CR measurements during K. brevis exposure were tested by linear regressions. For both treatments, the slope of the regression line was not significantly different from zero (P > 0.05) and the coefficient of determination R2 was low (<0.2) indicating that CR and time were not interdependent. For further statistical analysis we assumed that each measurement was independent and analyses of variance (ANOVA), followed by multiple comparison tests (Tukey), were performed. Normality of data and homogeneity of variances were also tested.

Results

For all treatments and both feeding modes, no clam mortality occurred during the course of the experiments. The first suspension-feeding experiment (Fig. 3a) showed that Macoma balthica maintained a constant weight-standardized clearance rate (ANOVA, P = 0.07) of ∼41 ml h−1 when successively exposed to the diatom F. familica and the non-toxic dinoflagellate H. triquetra. When exposed to the toxic dinoflagellate K. brevis (300–400 cells ml−1; 46.8 pg PbTx-3 eq cell−1 ± 1 SD), M. balthica showed a significant decrease in its mean clearance rate (ANOVA, P < 0.05) from ∼38 to 12 ± 4 ml h−1 (68% reduction) over 24 h (Fig. 3b). It was also observed that the inhalant siphon was less extended in the water column when the clams were exposed to the toxic dinoflagellate. The animals rapidly recovered their initial clearance rate after toxic cells were replaced with the F. familica conditioning diet. When exposed to a mixed diet of F. familica and K. brevis (30.5 ± 3.2 pg PbTx-3 eq cell−1), clearance rates were less affected (37% reduction) attaining a mean value of 24 ± 4 ml h−1 over 24 h (Fig. 3c), which was twice as high as that obtained with K. brevis alone (12 ± 4 ml h−1). The clams cleared K. brevis cells and F. familica cells from the mixed suspension at similar rates (24 ± 4 ml h−1 for F. familica and 33 ± 14 ml h−1 for K. brevis, t test, P > 0.05) indicating that the animals cleared both toxic and non-toxic algae without showing selective feeding behavior.

Fig. 3
figure 3

Mean wet weight-standardized clearance rate (ml h−1; mean ± SD) of Macoma balthica successively exposed to a control non-toxic diet of Fragilaria familica and an experimental diet (cell volume equivalent concentrations). Experimental diets were a the non-toxic dinoflagellate Heterocapsa triquetra, and b the toxic dinoflagellate Karenia brevis and c a 50:50 mixture of K. brevis and F. familica. Toxic diets are indicated by black bars. Letters (a, b, c and d) correspond to the results of Tukey multiple comparison test and indicate statistically significant differences at P > 0.05

Full size image

The K. brevis cell removal efficiency by clay in the deposit-feeding experiment was 98% ± 2 SD for both applications. During the 24 h exposure, clams showed reduced feeding activity when exposed to K. brevis (21.2 ± 6.1 pg PbTx-3 eq cell−1) compared to that of the non-toxic control (F. familica). In the latter, the inhalant siphon exhibited wide movements and explored the sediment over a large surface area, and the algal biofilm deposited on the sediment surface was completely removed after 24 h exposure. In contrast, in the presence of K. brevis the inhalant siphon did not emerge much and only a small area of the biofilm was removed in the vicinity of the orifice of the siphonal channel (Fig. 4). No obvious behavioral differences were observed between the treatment with K. brevis alone and the mixture of K. brevis and F. familica.

Fig. 4
figure 4

Partial removal of the clay layer by deposit-feeding after 24 h-exposure (see text)

Full size image

The toxin level in M. balthica tissues after 24 h-exposure reached 1.2 ± 0.12 μg PbTx-3 eq (g wet weight tissues)−1 by suspension-feeding in the presence of K. brevis alone (Fig. 5) and was not significantly different when K. brevis was present in a mixture with non-toxic cells [0.9 ± 0.19 μg PbTx-3 eq (g wet weight tissues)−1; t test, P > 0.05].

Fig. 5
figure 5

Brevetoxin concentration in tissues of the clam Macoma balthica [μg PbTx-3 eq (g wet tissues)−1] after 24-h exposure: a to suspended Karenia brevis (at 300–400 cells ml−1) and a 50:50 mixed diet including the non-toxic diatom Fragilaria familica; b to deposited Karenia brevis (loading of ∼3 × 105 K. brevis cells cm−2, equivalent to removal of 300 cells ml−1 from a 1-m-deep water column) and to a mixture of K. brevis (∼3 × 105 K. brevis cells cm−2) and F. familica (20% volume equivalent concentration)

Full size image

For the deposit-feeding experiment, toxin accumulation in tissues was not influenced by the addition of the non-toxic diatom F. familica (t test, P > 0.05) and toxin levels reached ∼1.6 μg PbTx-3 eq (g wet weight tissues)−1 in both treatments. Since the initial toxin loading was the same in both treatments, this suggests that no preferential ingestion of non-toxic cells occurred when clams were exposed to the mixture.

Discussion

Karenia brevis bloom concentrations are highly variable but typically in the range 100–1,000 cell ml−1, although densities up to 10,000 cells ml−1 have been reported (Tester and Steidinger 1997; Magnana et al. 2003). The experimental concentration of ∼300–400 cells ml−1 used in the suspension-feeding experiment was selected to minimize pseudofeces production and maximize ingestion of toxic cells. For the deposit-feeding experiment, the toxic cell loading of bottom sediment aimed to simulate removal of a bloom of a concentration similar to that of the suspension-feeding experiment from a 1 m water column. The brevetoxin concentration obtained in our K. brevis cultures using ELISA varied in the range 21–47 pg PbTx-3 eq cell−1 for cultures harvested in late log-phase. Roszell et al. (1990) reported brevetoxin concentration of ∼11 pg PbTx cell−1 for cultures in log-phase and in Florida waters cell toxin content usually varies in the range 10–17 pg PbTx cells−1 (Hua et al. 1996; Pierce et al. 2003). Exceptionally high toxin levels have been recorded during a Karenia bloom in FL with values in the range 47–126 pg PbTx cells−1 (Greene et al. 2000). Our values are consistent with all these data but it is worth noting that toxin content of algae and clam tissues was measured in the present study with the ELISA method, which reacts with PbTx type-2 backbone structure (Naar et al. 2002), including PbTx-2, 3, 9 and derivatives. Values in the study cited above were obtained with high pressure liquid chromatography (HPLC) or liquid chromatography/mass spectrometry (LC/MS) methods, which quantify only the predominant toxins PbTx 1, 2 and 3. Among the nine brevetoxins fully elucidated in terms of structure (PbTx 1 to 9), PbTx-2 is the dominant toxin in K. brevis cells (Baden et al. 1988) and represents ∼80% of the total brevetoxin content (Roszell et al. 1990; Hua et al. 1996). We can then assume that suspension- and deposit-feeding experiments simulated two comparable scenarios, representative of field conditions.

Karenia brevis reduced both suspension- and deposit-feeding activity. The clams’ clearance rate decreased when exposed to K. brevis while it remained constant when exposed to Heterocapsa triquetra, a non-toxic dinoflagellate of similar size. Clearance rate was also inhibited with the 50:50 mixture of K. brevis and F. familica and no selective feeding behavior was observed. Similarly, although based only on qualitative observations, deposit-feeding was adversely affected with both diets (i.e. K. brevis alone and mixture including F. familica) and no preferential ingestion of non-toxic cells appeared to occur based on comparable toxin levels accumulated in tissues. It thus appears that, with both feeding modes, clams respond to the presence of toxic K. brevis cells by reducing their feeding rate but have no ability to preferentially sort between toxic and non-toxic cells. According to Colin and Dam (2003), such a feeding response is attributed to a toxic effect, as opposed to a deterrent effect resulting from the grazers’ ability to detect a compound in the algae and to select against this specific algal species. Accordingly, Bricelj et al. (1991) showed that the hard clam Mercenaria mercenaria did not selectively ingest the non-toxic diatom Thalassiosira weissflogii over the PSP-producing dinoflagellate Alexandrium fundyense when offered in a mixed suspension. Li et al. (2001) obtained similar results with the clam Ruditapes philipinarum and the scallop Chlamys nobilis exposed to a mixture of the toxic dinoflagellate A. tamarense and the non-toxic diatom T. pseudonana. However, feeding responses to toxic dinoflagellates may be highly variable among bivalve species, and clearance rate is either inhibited (Bricelj et al. 1996; Matsuyama et al. 1999; Lassus et al. 1999), enhanced (Lesser and Shumway 1993) or not affected (Shumway et al. 1985; Bricelj et al. 1996). Supplementation with non-toxic cells appears to stimulate suspension-feeding activity compared to that on toxic cells alone in relatively toxin-sensitive species such as hard clam, oysters and soft shell clams (Bricelj et al. 1991; Bardouil et al. 1996). This can lead to accumulation of higher toxin levels in mixed suspensions (Bricelj et al. 1996). Indeed, the clearance rate of M. balthica fed with the 50:50 mixture was twice as high as that with K. brevis alone and consequently, similar brevetoxin concentrations in tissues were obtained in both treatments (i.e. K. brevis alone and mixture including F. familica). These results imply that even at relatively low K. brevis concentrations in the water column, toxin body burden may reach a significant level in tissues when K. brevis is mixed with non-toxic microalgae.

Reduction of feeding activity in the presence of K. brevis may ultimately affect bivalve growth and survival. This would particularly affect suspension-feeders as we assume that under most natural conditions the clay layer would be rapidly dispersed by tidal currents or waves after clay application and that consequently deposit-feeders would be only exposed to sedimented brevetoxins for a short period of time. Leverone et al. (2007) showed that clearance rates of juvenile bay scallops, Argopecten irradians, were highly affected at concentrations as low as 100 cells ml−1. They suggest that recurring blooms of K. brevis could hamper restoration projects for recovery of bay scallop populations in Florida estuaries. By reducing the exposure time to K. brevis, clay mitigation could thus prevent the impact on shellfish resources.

Toxin incorporation by Macoma balthica occurred by both suspension and deposit-feeding and toxin body burden after 24-hr exposure attained comparable values in both feeding modes [1 to 1.6 μg PbTx (g wet weight)−1], exceeding the guidance level of 0.8 μg PbTx-3 eq g−1 for edible tissues (based on the established mouse toxicity of PbTx-3). A similar brevetoxin body burden of ∼1.8 μg PbTx-3 eq (g wet weight)−1 was measured by ELISA in tissues of suspension-feeding bivalves collected in Florida waters (Naar et al., 2003). High levels of contamination achieved by suspension-feeding bivalves result from their ability to concentrate and consume high numbers of cells by filtering large volumes of water (Ward and Shumway 2004). Surface deposit-feeding bivalves use their inhalant siphon to vacuum surface sediment and ingest a mixture of sediment and organic matter including bacteria, detritus and microalgae (Hughes 1969; Hummel 1985). They are able to process large quantities of material to extract the diluted organic food source from the inorganic matrix (Taghon 1981). Clay mitigation of K. brevis concentrates PbTxs into a thin layer of clay-cell aggregates at the sediment surface thus making toxins available for deposit-feeders. Toxin levels in shellfish tissues obtained in this study should however be considered with caution as M. balthica is a species from North Atlantic and experimental conditions, representatives of Florida water temperatures, are not necessarily optimal for this species. M. balthica was chosen for its unique property of allowing test of both suspension- and deposit-feeding with a single species. We then demonstrate that clay mitigation of K. brevis blooms may provide a new trophic pathway for toxins by deposit-feeding of settled clay-cell aggregates and that toxin transfer via deposit-feeding could be of comparable magnitude to that attained via suspension-feeding during an untreated bloom (see Fig. 1). Deposit-feeders do not pose a direct threat to humans but may then provide a pathway for brevetoxin transfer to secondary consumers. The present study provides a first demonstration of brevetoxin accumulation by this trophic group and our findings suggest that toxin accumulation by deposit-feeders might also result from natural deposition of particulate matter during a bloom in the absence of clay treatment. This may occur due to settlement of K. brevis cells, deposition of copepod fecal pellets (Tester et al. 2000) or accumulation of biodeposits from benthic suspension-feeders. Toxin transfer to deposit-feeders during untreated blooms would then need to be investigated in order to fully assess the ecological impact of K. brevis blooms.

Conclusions

This study shows that significant concentrations of brevetoxins may be accumulated in shellfish tissues from deposited clay-cell aggregates after only 24-h exposure. However, the residence time and composition of brevetoxins in the deposit-feeding compartment remain to be determined, and are important in evaluating the potential for toxin transfer to secondary consumers. Furthermore, several studies have shown that shellfish are able to rapidly metabolize brevetoxins (Morohashi et al. 1999; Poli et al. 2000; Plakas et al. 2002; Ishida et al. 2004) and Poli et al. (2000) suggested that PbTx metabolites contribute to shellfish toxicity and may be the true cause of NSP in humans. Finally, this study provides the first evidence of brevetoxin transfer to deposit-feeders. Beyond the context of clay mitigation of HABs, this result highlights that benthic deposit-feeders should be considered as an important compartment in future studies on the fate of algal toxins in the food web.