Bioconcentration of Organotin Cations during Molting Inhibits Heterocypris incongruens Growth.

The densely populated North Sea region encompasses catchments of rivers such as Scheldt and Meuse. Herein, agricultural, industrial, and household chemicals are emitted, transported by water, and deposited in sediments, posing ecological risks. Though sediment monitoring is often costly and time-intensive, modeling its toxicity to biota has received little attention. Due to high complexity of interacting variables that induce overall toxicity, monitoring data only sporadically validates current models. Via a range of concepts, we related bio-physicochemical constituents of sediment in Flanders to results from toxicity bioassays performed on the ostracod Heterocypris incongruens. Depending on the water body, we explain up to 90% of the variance in H. incongruens growth. Though variable across Flanders' main water bodies, organotin cations and ammonia dominate the observed toxicity according to toxic unit (TU) assessments. Approximately 10% relates to testing conditions/setups, species variabilities, incoherently documented pollutant concentrations, and/or bio-physicochemical sediment properties. We elucidated the influence of organotin cations and ammonia relative to other metal(oxides) and biocides. Surprisingly, the tributylin cation appeared ∼1000 times more toxic to H. incongruens as compared to "single-substance" bioassays for similar species. We inferred indirect mixture effects between organotin, ammonia, and phosphate. Via chemical speciation calculations, we observed strong physicochemical and biological interactions between phosphate and organotin cations. These interactions enhance bioconcentration and explain the elevated toxicity of organotin cations. Our study aids water managers and policy makers to interpret monitoring data on a mechanistic basis. As sampled sediments differ, future modeling requires more emphasis on characterizing and parametrizing the interactions between bioassay constituents. We envision that this will aid in bridging the gap between testing in the laboratory and field observations.

Introduction

Despite decreasing emissions,[1,2] persistent chemicn>an class="Chemical">als can remain in the aquatic ecosystem for decades (legacy pollutants)[3] because they bind physicochemically to sediment.[4] Changing bio-physicochemical conditions can render pollutants bioavailable again to let them re-initiate adverse effects on aquatic organisms.[5,6] Thus, sediment “quality” is crucial to the health of an aquatic ecosystem.[7,8]

As a combination of species from different trophic levels and effect endpoints characterizes an ecpan class="Chemical">osystem’s resilience,[9−11] bioassays/test batteries for n>an class="Species">daphnia, algae, midge larvae, or bacteria[12,13] are common to establish a weight of evidence approaches for sediment quality. Toxic effects vary not only between organisms and life stages[14,15] but also due to differences in exposure regimes and sediment properties: one can use the solution as extracted from sediment[16] to evaluate “dissolved” contaminants or exposure to whole sediment to elucidate more integrated effects.

A variety of chemometric methods attempan class="Chemical">pt to expn>lain the observed effects[17−19] and have shown that n>an class="Disease">toxicity often does not relate to dry weight pollutant concentration.[20] Establishing dose–effect relationships with data from field studies is more difficult than that from spiking with single chemicals. Testing may not agree with field observations at all.[21,22] Many interactions between sediment constituents can attenuate or amplify toxic effects, more so than for surface water, which hampers selection of a proper dose metric.[23] Relations between bioassays and physicochemical constituents of sediment in the field are scarce, and quantitative predictions based thereon are elusive.

Among recent progress,[24−27] insight into the extent of additionpan class="Chemical">al available binding capn>acity via sediment propn>erties allowed correct identification of toxicity.[28] Both the apparent bioavailability and toxicity relate to chemical speciation. Speciation is the quantitative distribution of all chemical forms of a compound over all possible phases, as a function of bio-physicochemical properties of the endemic solution: pH, hardness, redox conditions, and natural catalysts (clays, microorganisms). These factors vary geographically.[29,30]

The North Sea region has a dopan class="Chemical">cumented history in pollution linked to industrin>an class="Chemical">al (heavy metals, acidification[31]) and agricultural (e.g., eutrophication) activities.[32−34] An extensive measurement campaign by the Flemish Environmental Agency (VMM)[32] has existed in Flanders for approximately 20 years, gathering information on pollutants in 1000+ sampling locations (e.g., metals, polycyclic hydrocarbons), in conjunction with characteristics known to affect speciation/bioavailability. Built hereon, models linked >60% of the observed variance in Hyalella azteca mortality bioassays to heavy metals and NH3/NH4+.[35]

Given the amount of pollutants and sediments to be evpan class="Chemical">aluated and monitored, bioassay(s) should entail limited redundancy so as to can>an class="Chemical">pture the bulk of the potential effects with minimum effort and cost.[36] Given their sensitivities, Chlorella vulgaris and Ceriodaphnia dubia may be suitable to evaluate metals,[37] whereas Daphnia magna and Raphidocelis subcapitata may be more applicable to metallic/metal oxide nanomaterials.[38] For Flanders, the question would be: which causal relationships between constituents of polluted sediment and ecotoxicological effects allow us to prioritize or complement bioassays?

pan class="Chemical">Ostracod microcrustaceans are keystone meiofauna species in production and community metabolism of micro- or mesocosm freshwater beds.[39] Present in diverse freshwater benthic habitats (ponds, lakes) in all continents, ostracods prove to be bioindicators when other species fall short.[40] Many campaigns entail tests with Heterocypris incongruens (seed shrimp).[41−44]H. incongruens is substantially more sensitive than midge larvae[45] and comparably sensitive (to heavy metals, NH4+) with H. azteca,[41] but H. incongruens growth is particularly sensitive.[39] As Flanders is an urbanized coastal (maritime) and agricultural region, concentrations of organotins (antifouling endocrine disruptors)[46−48] and phosphate[49,50] could be elevated and particularly relevant for growth.

Given the potentipan class="Chemical">ally higher sensitivity of n>an class="Species">H. incongruens growth to sediments in Flanders and its dissimilarity with other bioassays, we set out to build a quantitative model for predicting ostracod growth inhibition. Using extensive monitoring data in Flanders, we accounted for bioavailability and chemical speciation to develop a predictive model based on toxic unit (TU)[51] calculations linking concentration to effect. We predict ostracod growth inhibition with appreciable certainty, ∼90%. Our method underpins that ostracods complement other bioindicators and supplements the existing suite of models.

Materials and Methods

In the current sn>an class="Chemical">tudy, we quantitatively predict the growth inhibition of H. incongruens as would be observed when exposed to sediment. As an empirical exercise, we utilized the results from bioassays as obtained from our Flemish VMM monitoring campaign. Bioassays are detailed in Section and elsewhere.[45]

Input variables for our predictions were dry weight concentrations of various chemicpan class="Chemical">als. Following a recent study,[35] we considered heavy metals, nitrogen, and phosphorus. Given exposure history, we also hypothesized local toxic effects by organotins. As variables, we also used common sediment characteristics: pH, organic carbon, clay, and the nitrogen status. Physicochemical sampling of sediments and analytical methods are detailed in the Supporting Information (Section S0) and elsewhere.[13,32] We describe the parametrization below (Sections –2.5).

Estimation of Ammonia and Phosphate in Sediment

Data on aqueous concentrations of pan class="Chemical">NH4+ and are either scattered, unrepn>resentative, or their uncertainties are fairly high.[13,32,35] This is partially due to their high mobilities. Therefore, we estimate and from the total phosphorus, PTwith f being a ratio applied to total P to distinguish for available fractions: a typical natural water body (PT ∼23 or ∼30 mg/m3) constitutes 3 mg/m3 soluble + and usual ( + )/PT ratios of 1:10. Hence, fP=0.1.[52]

In pan class="Chemical">turn, we estimate n>an class="Chemical">NH4+ and NH3 concentrations from the Kjeldahl nitrogen (KjN) and eq wherein fN is taken as the fraction of the KjN that is [NH4+] + [NH3], taken to be 0.5(±0.1).[53,54] Exclusive for the data in West Flanders (s = 1, else s = 0), we applied a sinoid term with maximal amplitude (peak) in mid-march (d = 65)[55,56] to account for local seasonal (agricultural) fluctuations[57,58] (S1). If the data adhering to negative growth inhibition (Section ) did not entail KjN determination, we applied the log-average of samples analyzed throughout Flanders. To prevent bias, we excluded those sediments that we already selected from curation (Section ).

Bioavailability

Toxicity is mn>an class="Chemical">ost often associated with the fraction of the chemical that is freely bioavailable. Though test vessels contained sediment,[13,59] we had no information on burrowing behavior. We assumed that all ostracods are exposed to chemicals in the water phase. From our campaign, we obtained only the total sediment concentrations. Since those sediments are applied to the (aqueous) bioassay, we estimate in situ water concentrations from the total in sediment.

We converted sediment (sed) concentrations to water (aq) concentrations via phase partitioning[60,61]wherein 0.2 repn>resents the sediment to test volume ratio (1:5[45]). Considering that the test medium is presn>an class="Chemical">cribed as moderately hard,[13] we took the value for the solids-to-water partitioning coefficient KSW, local for NH4+/NH3 and H2PO4–/HPO42– to be 6.0(±3.0)[62−65] and 100 L/kg,[66,67] respectively, for field concentrations[64−66] representable for Flanders[13,32,35] and pH = 7.4.

The binding of pan class="Chemical">metals is stronger in sediments with inn>an class="Chemical">creasing clay and organic carbon (OC),[68,69] creating a natural “bias”. Thus, we calculated KSW,local based on %clay and %OC content, analogous to Bruijn and Denneman[70]wherein we obtained the solid-water partition coefficient (KSW) from the literature,[60,71,72] representing a mix of speciation states. KSW values were assumed to represent 11% clay and 5% OS.[13,68]a and b are the concentrations (mg/kg dissolved substances) of metals per %clay and %OS, respectively, based on (log–log) multiple linear regression (MLR) for the entire VMM (Flanders) data set (N = 1762):Considering their hydrophobicity, we estimated KSW,local for organotins from their organic carbon partition coefficients (KOC) and the %OC content (varying 10 ≳ %OC ≳ 0.1[13,32]):Sediment and biota affect the in situ ionic composition of the medium,[73,74] though the influence of salinity[75,76] on organotin sorption is conflicting (salting out vs. cation exchange). We also lacked information on the relevant OC type.[75] Hence, we did not consider influences of salinity and pH on KOC. Instead, we applied log-normal averages for KOC values reported for fresh and marine sediments (Table ).

Table 1

Sorption Equilibrium Partitioning Coefficients

chemical	log KOC[75]	log KSW	ref
tetrabutyltin	5.5(±0.1)	4.2(±0.1)a	(75)
tributyltin	5.3(±0.1)	4.0(±0.1)a	(75)
dibutyltin	5.2(±0.2)	3.9(±0.2)a	(75)
monobutyltin	5.0(±0.2)	3.7(±0.2)a	(75)
triphenyltin	4.9(±0.1)	3.6(±0.2)a	(75)
diphenyltin	4.7(±0.2)	3.4(±0.2)a	(75)
monophenyltin	4.4(±0.3)	3.1(±0.2)a	(75)
NH4+/NH3	n/a	0.8b	(62, 65)
H2PO4–/HPO42–	n/a	2.0(±0.5)b	(66)
Cd	n/a	2.1a	(60)
Cr	n/a	2.5a	(60)
Cu	n/a	1.7a	(60)
Hg	n/a	2.2a	(60)
Ni	n/a	0.9a	(60)
Pb	n/a	2.8a	(60)
Zn	n/a	2.0a	(60)

Taken to represent sediments in Flanders with 5% OC and 11% and clay.

Taken for all sediments in Flanders.

Taken for pan class="Chemical">all sediments in Flanders.

Speciation

Toxicity is often the result of a spn>ecific spn>eciation state of a chemicn>an class="Chemical">al; heavy metals, organotins, phosphate, and ammonia are subject to adopt different speciation. Thus, we describe explicitly the influence of the (bio)chemical properties of the endemic solution.

Phosphate may act both as a toxicant[49] and a nutrient,[50] whereas n>an class="Chemical">organotins and ammonium are toxicants. Prominently, the fraction of nonionized ammonium often relates to observed toxicity.[77] The species (de)protonate according toWe thus determine their speciation via Henderson–Hasselbalch[78]wherein pKa(XH = NH4+) = 9.25 and pKa(XH = ) = 7.2. We took the pKa values for organotins from the literature (Table ).

Table 2

Proton Dissociation Constant (pKa) Valuesa

name	formula	pKa	ref
tetrabutyltinhydroxide	Sn(By)4	n/a	n/a
tributyltinhydroxide cation	(Sn(OH2))+(By)3	6.25	(79−81)
dibutyltinhydroxide cation	(Sn(OH2)1)+(OH)1(By)2	5.1(±0.2)	(82)
monobutyltinhydroxide cation	(Sn(OH2)1)+(OH)2(By)1	5.9(±0.1)	(82)
triphenyltinhydroxide cation	Sn(OH2)+(Ph)3	5.2	(79, 81, 83)
diphenyltinhydroxide cation	(Sn(OH2)1)+(OH)1(Ph)2	4.0	(84)b
monophenyltinhydroxide cation	(Sn(OH2)1)+(OH)2(Ph)1	4.8	**c
dihydrogenphosphate	H2PO4–	7.2	n/a
ammonia	NH4+	9.25	n/a

See Table S2 for full details. By = butyl; Ph = phenyl.

In 75% dioxane–water solution. pKa values of ligands in 75% dioxane–water solutions are higher than those reported in water.[84]

Double asterisks mean an estimation assuming the substitution of butyl by phenyl has a constant effect on pKa: pKa MPT = pKa MBT – ((pKa TBT – pKa TPT) + ( pKa DBT – pKa DPT)/2).

See Table S2 for full details. By = butyl; pan class="Chemical">Ph = phenyl.

In 75% dioxane–n>an class="Chemical">water solution. pKa values of ligands in 75% dioxane–water solutions are higher than those reported in water.[84]

Double asterisks mean an estimation assuming the substitution of butyl by phenyl has a constant effect on pKa: pKa Mn>an class="Chemical">PT = pKa MBT – ((pKa TBT – pKa TPT) + ( pKa DBT – pKa DPT)/2).

Cr n>an class="Disease">toxicity relates to the presence of CrO42–[85,86] via microbial (nitrification,[87−89] eutrophication[33]) and redox processes:[90] [CrO42–] correlates with [NO3–] and pH[91] due to (bio)chemical reactionsBecause KSW(Cr3+) ≫ KSW(), the total sediment Cr concentration is dominated by [Cr3+], i.e., Thus, based on eqs and 13, we characterized the ratio between and Cr3+ empirically via[35]wherein we take on average to be 4.13% of total Cr[91] and c, d, e, and f are regression coefficients[35] (Figure S4). Since KSW of CrVI (<1–50 L/kg) ≫ Crlll (850–5600 L/kg),[92] ∼90% of aqueous Cr is .

Median Effect Concentrations (EC50)

With the exception of n>an class="Chemical">Cr and Sn, we assumed that the total concentration of metals is equal to a single free valent (uncomplexed) M2+ oxidation state, as the sole contributor to the observed toxicity. We assume that Cr expresses toxicity only via the CrO42– form (Table ).

Table 3

H. incongruens Effect Level Concentrations of Chemicals from the Open Literaturea

chemical	effect (EC50) concentration (ng/L)	effect (EC50) concentration (ng/L cation)b	ref
tetrabutyltin	50	n/a	Table S3
tributyltinhydroxide	50	3.5	Table S3
dibutyltinhydroxide	50	0.3	Table S3
monobutyltinhydroxide	50	1.6	Table S3
triphenyltinhydroxide	50	0.3	Table S3
diphenyltinhydroxide	50	0.02	Table S3
monophenyltinhydroxide	50	0.1	Table S3
Cu2+	950,000	n/a	(95)
Ni2+	2500,000	n/a	(95)
Hg2+	400,000	n/a	(95)
Zn2+	14,775,880	n/a	(95)
Cd2+	70,000	n/a	(95)
CrO42–c	4,310,000	n/a	(95)
Pb2+	39,200,000	n/a	(95)
NH3	2000	140,000d	(96, 97)
HPO42–/H2PO4–	3,905,460,000	2,000,000,000e	(49)

The values for organotins are the means from hatching, developmental, growth, and mortality experiments with varying durations (4 to 28 days); details are shown in Table S3.

Assuming pH = 7.4.

Since dichromate hydrolyzes in water (Cr2O72– + H2O ⇌ 2 CrO42– + 2H+), we multiplied EC50 by 2.

Represents NH3.

Represents HPO42–.

The values for n>an class="Chemical">organotins are the means from hatching, developmental, growth, and mortality experiments with varying durations (4 to 28 days); details are shown in Table S3.

Since dichromate hydrolyzes in n>an class="Chemical">water (Cr2O72– + H2O ⇌ 2 CrO42– + 2H+), we multiplied EC50 by 2.

Represents pan class="Chemical">NH3.

Represents pan class="Chemical">HPO42–.

In contrast to uncomplexed heavy pan class="Chemical">metals (Table ), to our knowledge, there are no 6 day EC50 vn>an class="Chemical">alues available for H. incongruens growth when exposed to organotins, NH3, or H2PO4–/HPO42–. Since we assumed all H. incongruens to reside in the water layer, we screened the literature for aqueous phase EC50 values for related species/studies. Though tributylated organotin is generally more toxic than other organotins,[93,94] to our knowledge, there is no information on the (relative) H. incongruens toxicity due to substituents.

Based on Table S3, we initially estimated (based on log-averages) the EC50 of (neutrn>an class="Chemical">al) organotins, NH3, and HPO42– for H. incongruens growth to be 50, 2000, and 2,000,000,000 ng/L (Table ), respectively.

Toxic Unit (TU)

Initial Approximation

The TU estimates mixn>an class="Chemical">ture-toxicity risks to both terrestrial and aquatic communities.[8] We assessed the total exposure from the sediment by summing the exposures for all of the contaminants (eq )[98]wherein n reflects the number of pollutants i, taken as heavy metals, organotins, NH3, and H2PO4–/HPO42–.

For the current purpn>n>an class="Chemical">ose, it was not feasible to parametrize a priori the toxic interaction between chemicals based on their EC50 values only (i.e., without SSD shape parameters).[51,99] Taking γ = 1 reflects all chemicals to have a similar mode of action (MoA) and negligible interaction between chemicals. As the simple response addition (RA) method does not necessarily produce better results than concentration addition (CA),[35]eq produces an “educated guess” for the TU of H. incongruens.

Calibration

Assuming a logistic distribution of sensitivities toward the polluted sediment, we despan class="Chemical">cribed the percentage of growth inhibited as a function of exposure via a cumulative logistic distribution function[98]wherein β is a steepness function and TU is initially the ratio of the measured (bioavailable) chemical concentration [C]aq to its effect concentration [EC50]aq.[100] Fitting mortalities with these TU values (Section ) produced suboptimal results (Figure A; Figures S5, S7).

Figure 3

H. incongruens growth inhibition (%) versus the toxic unit. GI = growth inhibition. (A) EC50 for organotins (cations) represented 3 ng/L; pH assumed to be 7.5. R2 = 0.26. (B) EC50 for organotins (cations) represented 0.003 ng/L (eq ); pH assumed to be 7.5. R2 = 0.82. (C) EC50 for organotins (cations) represented 0.003 ng/L(eq ); pH taken for the porewater. R2 = 0.90.

To acquire statistically better predictions based on the explained variance, we n>an class="Chemical">allowed variation in EC50 values to fit the growth inhibition data (Section ) via eq . This was done via the factor γ in eq . The preselected range for γ was 1000. The fitting implies the average TU fixed at the exposure for which 50% of the growth is inhibited. We envisioned potential divergences between values for γEC50 and EC50 to be the result of unknown testing procedures and conditions altering speciation, bioavailability (e.g., metals as oxides or organometals), or MoAs. Final γEC50 values are given in Figure .

Figure 2

H. incongruens growth inhibition (%) versus the toxic unit. EC50 for organotin cations was taken to be 0.003 ng/L, representing By3SnOH2+. Simulation was performed using β = 4.5 (unitless). The far-left triangle denotes the weighted average of values with log TU < −1. Squares denote data wherefore KjN was unknown and taken as a regional average (Section ). The table provides corresponding uncertainties/ranges. For EC50 values, the errors are those carried over from errors in bioavailabilities.

Bioassay Description and Data Curation

We cpan class="Chemical">alibrate and test our methods (eqs –16) using data on whole-sediment n>an class="Disease">toxicity assessed by VMM via laboratory bioassays performed on the ostracod H. incongruens, a small benthic freshwater crustacean. The data represents a 6 day growth inhibition test (ISO 14371[101]) with freshly (≤52 h old) hatched cysts (i.e., neonates, with lengths of 150–250 μm) incubated at 25(±1) °C, in darkness. Details are shown elsewhere.[45,102] The percentage % of growth inhibition μ because of exposure to sediment is defined[45] aswhere the lengths are the averages of a total of 60 organisms distributed across six replicates. The reference refers to a standard sand. For the reference, lengthday6, reference ≥ 1.5·lengthday0 and mortality is ≤20%. To enlarge our modeling domain, we specifically included also bioassays reporting negative growth inhibition (i.e., growth stimulation).

Since the bioassay is aerobic, it should represent aerobic sediments in the field. This implies minimizing speciation-related changes due to in sipan class="Chemical">tu anaerobic–aerobic transformations. n>an class="Chemical">Also, previous O2 deprivation in the field may create bias for germination/growth (chemical cues) at the start of the bioassay: crustaceans are sensitive to O2 deprivation;[103,104] low O2 induces dormancy (HOEC = 5.5 mg O2/L).[104] Taking 44 ± 4 mg/L and 0.21 as solubility (T ≈ 20 °C) and partial pressure, respectively, this amounts to 60 ± 5% O2. H. incongruens does not tolerate acidic conditions.[39]

Therefore, we exclude all assays for which the field poren>an class="Chemical">water O2 < 65%, O2 ≥ 100%, and pH ≤ 6.5, which we consider nonrepresentable sediments.

Results

We performed anpan class="Chemical">alyses to elucidate the n>an class="Disease">toxicity of sediments in Flanders to H. incongruens. According to initial TU results (Figures S5, S7), NH3 (TU = 0.1) contributed most to H. incongruens growth inhibition. Other contributors were Ni and Cu (TUs ≤ 0.1), whereas the individual TUs of Zn, Cd, CrO4, Hg, and Pb were ≤0.01. Predicting the % growth inhibition via initial TU using initial (standard) EC50 values (Table , Section ) yielded maximal explained variance lower than 30%. Initially, speciation calculations (eqs , 14) did not appear to improve these results. In contrast to Figure S7 (wherein there is a low contribution), organotins are among the species most highly correlating with H. incongruens % growth inhibition (Figure S9).

We then performed “semiempiricpan class="Chemical">al” ann>an class="Chemical">alyses to further elucidate the toxicity of sediments by letting γEC50 values fit the growth inhibition data (Section ). This “calibrating” procedure allowed a robust TU analysis for H. incongruens. According to the semiempirical analysis, too, NH3 dominates the TU (TU = 0.3). Additional toxic pressure arises from organotins (as represented by tributyltin, TU = 0.3). Then, the contributions from Ni and Cu appear minor (≤0.1) (Figure ). Inclusion of chemicals other than NH3, HPO42–, and By3SnOH2 only marginally improved the fit. Instead, we found a positive influence by HPO42–, i.e., growth-promoting (TU = −0.3).

Figure 1

Toxic units for H. incongruens ([C]aq/EC50aq) for individual pollutants in sediments from water bodies in Flanders. Table denotes EC50aq values for free metals; Figure denotes EC50 values for NH3, By3SnOH2+, and H2PO4/HPO4, contrasting ranges in Table . Hence, the TUs may be both underestimations/overestimations. Error bars denote variability throughout Flanders. For MFT (PhSn(OH)2OH2+), assuming porewater (pH = 7.5), 99% of samples were below the detection limit (DL); negative errors are 10-fold the SD including the values set at the DL.

Toxic units for pan class="Species">H. incongruens ([C]aq/EC50aq) for individun>an class="Chemical">al pollutants in sediments from water bodies in Flanders. Table denotes EC50aq values for free metals; Figure denotes EC50 values for NH3, By3SnOH2+, and H2PO4/HPO4, contrasting ranges in Table . Hence, the TUs may be both underestimations/overestimations. Error bars denote variability throughout Flanders. For MFT (PhSn(OH)2OH2+), assuming porewater (pH = 7.5), 99% of samples were below the detection limit (DL); negative errors are 10-fold the SD including the values set at the DL.

After calibration, we found that spn>eciation cn>an class="Chemical">alculations (eq ) significantly improved the results (Figure B,C). Then, NH3, organotins, and HPO42– explain up to 90% of the observed variance in the % growth inhibition for water bodies throughout Flanders. We established γEC50 values for NH3, HPO42–, and TBT cation (By3SnOH2+) at 150,000–39,000 and 0.003 ng/L. The relative importance of the toxicants differs between water bodies. In general, there is a higher toxic pressure in West Flanders than in inland regions (Table S4). In general, predictions for H. incongruens growth inhibition are more precise for “uniform” (e.g., large rivers and canals) water bodies than for small “nonuniform” water bodies.

Discussion

For sediments in Flanders, we report a maximum growth of pan class="Species">H. incongruens of 56% (Supn>porn>an class="Chemical">ting Information). The curated data depict a leveling (for low toxicant concentrations) between 0 and −40% (Figure ). In comparison, H. incongruens grows in unpolluted soft and hard water (pH = 7) with 0.063(±0.003) and 0.072(±0.003) mm/day, respectively.[105] Assuming linear growth, this amounts to 84(±42)% in 6 days, similar to sediments in Flanders with “low” toxicant concentrations. This indicates toxic pressure to H. incongruens in the majority of bioassays on sediments from Flanders.

We explain up to 90% of the variance in pan class="Species">H. incongruens growth. The standard error of model prediction (Figure ) is lower than that of a similar model for n>an class="Species">H. azteca.[35] This is likely related to a better defined testing system and lower physical or metabolic complexity of the organism.[13,45] Predictions appear more precise for neonate growth than for survival (Figures S9, S13), similar to earlier observations.[96] A coefficient of variation of 6–20% was reported using parallel mortality assays.[106] Thus, the contribution of other toxicants, erroneous speciation/bioavailability, or mixture effects to the residual variance (∼10%, Figure ) is limited. 10-fold intraspecies uncertainty factors are widely used;[107] our prediction errors are well within this margin.

H. incongruens growth inhibition (%) versus the toxic unit. EC50 for n>an class="Chemical">organotin cations was taken to be 0.003 ng/L, representing By3SnOH2+. Simulation was performed using β = 4.5 (unitless). The far-left triangle denotes the weighted average of values with log TU < −1. Squares denote data wherefore KjN was unknown and taken as a regional average (Section ). The table provides corresponding uncertainties/ranges. For EC50 values, the errors are those carried over from errors in bioavailabilities.

The average toxic pressure for pan class="Species">H. incongruens is compn>arable to that for n>an class="Species">H. azteca.[35] Both initial and calibrated TU analyses support our hypotheses that NH3, along with heavy metals, dominates the effects on H. incongruens. Meanwhile, in contrast to H. azteca,[35] CrO42– did not appear to contribute to the apparent toxicity. The analysis also shows that the mixture of organotins, NH3, and HPO42– acts distinctly toward H. incongruens. We interpreted this partially as a bioconcentration effect, which we will explain stepwise below. From its distinct sensitivity and the 90% explained variance, we deem the H. incongruens sediment contact test to be a complementary bioassay to evaluate sediment and water quality.

Organotins

Ni and Cu seemed to only marginn>an class="Chemical">ally inhibit growth (TUs ≤ 0.1), though heavy metal loading as low as 100 μg/kg affecting aquatic species is generally accepted.[108] Organotins, as characterized by TBT, appear significantly more toxic to H. incongruens than to similar aquatic species in single-substance tests (Figure , Table S3). This is on average a factor of 1000, irrespective of speciation. In other words, the term γTBT in eq is ∼1 × 10–3. Metals react with receptors, enzymes, DNA, protein and lipids or produce structural and functional changes resulting in toxic damage. The toxicant can be a derivative of the parent metal or reactive species (ROS or RNS) from (bio)transformations.

The correlation between TBT cation concentrations and growth inhibition is stronger than for the neutrn>an class="Chemical">al TBT (Figure B,C; Figure S10). Since H. incongruens is small, the internal (e.g., cytoplasmic) pH may be correlated to medium/porewater pH.[109,110] This indicates that growth inhibition is associated with the cation. In contrast, organotin toxicity toward yeast increases with cations < hydroxides < chlorides.[111] From pH 5.8 to 8.0, the KOW of TBT and TPT increased.[112,113] As it partitions into the lipid fraction of the biomembrane, the neutral (hydrophobic) form is more bioavailable:[114] uptake and bioconcentration of TBT are higher at pH 8.0 (where neutral TBTOH predominates) than at pH 5.0–6.0.[115]

H. incongruens growth inhibition (%) versus the toxic unit. GI = growth inhibition. (A) EC50 for n>an class="Chemical">organotins (cations) represented 3 ng/L; pH assumed to be 7.5. R2 = 0.26. (B) EC50 for organotins (cations) represented 0.003 ng/L (eq ); pH assumed to be 7.5. R2 = 0.82. (C) EC50 for organotins (cations) represented 0.003 ng/L(eq ); pH taken for the porewater. R2 = 0.90.

However, biomembrane–water distribution of the n>an class="Chemical">TBT cation exceeds liposome–water distribution, and distribution of the TBT cation exceeds that of the neutral hydroxo-complex. This relates to the cation complexing with ligands and proteins.[114] This indicates that the high sensitivity of H. incongruens growth on organotin as compared to other aquatic species (Figure ; Table S2) does not arise from nonpolar (baseline) narcosis.[116] In turn, this implies interactions between the TBT cation and proteins with growth regulation activity.[117]

Endopan class="Chemical">crine activity of n>an class="Chemical">TBT in crustaceans refers to the interactions with the ecdysteroid receptor–retinoid-X receptor dimer (CrcEcR-CrcRXR) complex:[48,117,118] TBT binds in the electrostatic and neutral areas of CrcRXR.[117] The ecdysone (nuclear) receptor is responsible for transcriptional regulation of molting,[119] which is critical for growth of crustaceans accumulating (i.e., “feeding on”) phosphate and carbonate. Interaction with vitellogenin and cuticular proteins may cause a second endocrine disruption on the Y-organ–ecdysteroidal endocrine axis and impair growth.[117]

Phosphate

In contrast to pan class="Disease">toxicity (γn>an class="Chemical">PO > 0), phosphate appears to promote H. incongruens growth (γPO < 0) (eq ). The apparent effect level is ∼0.04 mg/L, far lower than levels causing toxicity (∼2000 mg/L[49]). Given the aqueous concentration of phosphate in the bioassays, 0.02–2 mg/L, and an LC50 value of approximately 3900 mg/L,[49] toxicity of HPO42– to H. incongruens is negligible. H. incongruens is one of the only two ostracods wherefore Ca2+ positively correlates with its natural abundance; Ca2+ associates with phosphate.[120] P content of the carapace differs geographically.[39]

pan class="Chemical">Prior to moln>an class="Chemical">ting, ostracods store CaHPO4 (apatite) granules and chitin precursors in epidermal cells. The granules expulse their content, which is transformed into platelets of CaCO3, disintegrated to amorphous calcite and, ultimately, crystals to build the carapace.[121] Note that HCO3– is not a limiting factor (relatively constant) since it is added via the synthetic medium.[45] The carapace may need ∼7 days to complete its calcification, via up to nine growth stages (instars).[50]

We find a higher correlation for HPO42– than for n>an class="Chemical">H2PO4– (Figure B,C; Figure S11). Indeed, an alkaline environment (high pH) accelerates tissue calcification.[122,123] Moreover, calcium phosphate crystallinity decreases with an increase in reaction pH,[123,124] and solubilities decrease with increasing pH.[125] The combined results indicate that phosphate in bioassays performed with sediments from Flanders, in the HPO42– form, promotes H. incongruens growth.

Ammonia

Whereas we initipan class="Chemical">ally assigned an EC50 vn>an class="Chemical">alue for NH3 of 0.002 mg/L, we found an empirical value for γNHEC50 of 0.15 mg/L (implying γNH > 1) (eq ). In analogy, the LC50 value for H. azteca exposed to NH3 is 2–5 mg/L,[78,126] whereas in bioassays, it appears empirically to be 5000 mg/L.[35] The average TU of NH4+/NH3 is 0.3 across Flanders, and NH3 explains 30% of the variance in growth inhibition. We noted a relatively high toxicity of sediments from West Flanders (e.g., Yser river) collected during (fertilization[56] in) spring; levels of NH4+/KjN in the Yser are structurally high (Table S4); anaerobic conditions are more likely for those sediments.

We found a slightly higher correlation for NH3 than for n>an class="Chemical">NH4+ (Figure B,C; Figure S12). Aquatic ecotoxicity studies in the 1990s identified NH4+/NH3 (esp. NH3) as a contributing cause.[127] More so than NH4+, NH3 can pass through biological surfaces into tissue fluid. When pH of the tissue fluid is lower than that of the surrounding water, NH3 can disrupt ionoregulatory functions.[49,128] In addition, NH3 may activate N-methyl-d-aspartate receptors in the brain,[129] leading to intoxication.

Mixture Effects

The γEC50 values for n>an class="Chemical">HPO42–, NH3, and organotin used for the analysis differed from the initial estimation (Table ; Figure ). With respect to literature EC50 values, much variation exists (Table S3); the values obtained represent extreme values. Thus, the extreme γEC50 values may not only represent a direct mode of toxic action from HPO42–, NH3, and organotin individually but also suggest toxicity as the result of interaction(s) between them, i.e., γ ≠ 1 (see eq ).

There are many documented interactions between n>an class="Chemical">P, N, and metals.[130] However, models (e.g., the TU, eq ) often describe toxicity additively or independently,[98] i.e., not including synergistic/antagonistic mixture effects.[131] Mixture toxicity entails detailed description of the interactions between sediment fractions and chemicals: depending on numerous conditions, metals form organic complexes or oxides, each with its own characteristic influence on the intoxication pathway.

Phosphorus–Tin Interaction

Though phosphorus can be toxic,[132] it can diminish n>an class="Chemical">metal toxicity via precipitation.[133] Phosphate is a polydentate ligand forming negatively charged complexes with triorganotins mono- and dimers;[73,134] organotin is 102–105 times more selective toward phosphate than toward other oxoanions[135] and aids in its extraction.[136] In fact, extraction of organotins from sediment co-extracts phosphorus compounds.[137] Plants absorb phosphate; ostracods feed on dead/living plants. Thus, the influence of sediment’s organic phosphate on the distribution of TBT in ostracods may relate to dietary assimilation via periphyton.[120,130] Acute toxicity of TPT on ostracods may be supplemented by elimination of food.[138]

pan class="Chemical">Prominently, n>an class="Species">H. incongruens accumulates HPO42–, hence TBT, via molting (Section ). Additionally, the TPT+ cation sorbs onto liposomes via complex formation with phospholipids.[114] Guanosine-5-monophosphate (G-5-MP) most efficiently forms a competitive organotin complex;[139] as an essential enzyme for cell growth, G-5-MP kinase converts GMP to GDP. Combined, the interactions with phosphate compounds indicate an effective bioconcentration larger than based on the thermodynamic (hydrophobic) maximum ((1) in Figure ). Reported BCFs for species (e.g., bivalves) capable of accumulating phosphate are ∼1000 times larger than those without the capability.[81,140] This value matches with 1/γTBT, insinuating a similar TMoA:

Figure 4

Simplified schematic of interactions between sediment constituents affecting bioassay toxicity. Green and red arrows denote positive and negative effects on H. incongruens growth, respectively. According to de Deckere et al.,[13] O2 ought to remain ≥60% during bioassays.

Simplified schematic of interactions between sediment constituents affecn>an class="Chemical">ting bioassay toxicity. Green and red arrows denote positive and negative effects on H. incongruens growth, respectively. According to de Deckere et al.,[13] O2 ought to remain ≥60% during bioassays.

Nitrogen–Tin Interaction

Since the bioassays are aerobic,[45,102] NH3/n>an class="Chemical">NH4 may be converted to NO2–/NO3– (Figure ), which would be enhanced by bioturbation.[74] Conversion to NO3–[35] could contribute to the high apparent γEC50 value of NH3 (Figure ). Though affecting ostracods,[141−143] NO3– is likely not directly toxic.[35,144] Instead, metal-nitrates are well-soluble, hence bioavailable.[145] Heavy metals may influence nitrification.[88,146] In turn, nitrification and volatilization (NH4+ → NH3(↑) + H+) affect pH. Changes in pH in sediment/ecotoxicity tests are reported;[35,147] we cannot assume that buffering was consistent.

Surplus H+ increases the cation concentration and interactions of the n>an class="Chemical">TBT metal-type behavior by complex formation with ligands in phospholipids and proteins,[48,148] illustrated by (2) in Figure . NO3– and SO42– (also in the EPA medium) may solubilize organotins. In addition, the relative concentrations of NH4 and metals may influence uncoupling of mitochondrial oxidative phosphorylation.[149]

Limitations, Recommendations, and Outlook

Sorpan class="Chemical">ption to varying organic n>an class="Chemical">carbon did not appear to strengthen correlations (eq ). Apart from those considered here, other sediment constituents (e.g., Al, Mg, Ca, S) affect speciation and bioavailability.[30,150] Prominently, acid-volatile sulfide (AVS)[20,69] relates to invertebrate toxicity.[85,86,151] We can determine bioavailability more comprehensively by distinguishing sediment fractions (e.g., humic acid[75]) and binding mechanisms,[152] pending available data.[75,153]Section S3 contains a more complete discussion of limitations. Despite increased monitoring,[13] it is laborious to measure all such relevant parameters, which depend on the sediment/organism. Even if we have such direct detailed information, it is difficult to describe all interactions explicitly.[154−156] We therefore ought to consider also (semi)empirical- and evidence-based approaches.

Based on the current sn>an class="Chemical">tudy, we argue for an approach based on (bio)chemical interactions between NH4+, organotins, and HPO42–. Incomplete monitoring data for N or P might obscure or underestimate ostracod toxicities.[35] Since phosphate was not present in reference sediments, this may obscure the toxic effect on the ostracods. Thus, we propose to standardize total phosphorus (i.e., phosphate) in the sediment bioassays. Prediction of ostracod growth is probably more straightforward than that of mortality, which relates to, e.g., complexity of the toxic mode of action. Due to the higher sensitivity of juveniles, we recommend using the neonate test or making age-based extrapolations. BCF tests should elucidate the validity of eq .

pan class="Chemical">Pollutants are not n>an class="Chemical">always (directly) measured, which would be important for remote (small) water bodies where an “exotic” chemical may skew local toxicity. More direct data for NH4/PO4 may help to solidify the relationships.[35,157,158] Large, uniform water bodies are less susceptible to pressures from point releases of chemicals (dilution). Higher temperature and drought (concentrating pollutants) may influence (bio)transformation, O2 uptake, and elevate toxicity.[159] This determines parametrization, representability, and implementability of models. We may optimize via identifying “hotspot” locations, e.g., bridges/harbors.[160] Concentrations of organotins appear to decrease over time in sediment in Flanders. Removal mechanisms include leaching into sea or biodegradation (Figure S3; Figure ), which need to be elucidated.

Figure 5

First-order disappearance rate constant for TBT in Flemish sediment (x, in year–1) versus salinity and concentration (y1 and y2). Symbols denote distinct sampling locations, having a uniform temporal trend in concentration (continuously increasing or continuously decreasing).

First-order disappearance rate constant for TBT in Flemish sediment (x, in year–1) versus sn>an class="Chemical">alinity and concentration (y1 and y2). Symbols denote distinct sampling locations, having a uniform temporal trend in concentration (continuously increasing or continuously decreasing).

In this span class="Chemical">tudy, we have apn>plied explicit desn>an class="Chemical">criptions to characterize the speciation and bioavailability of pollutants in ostracod bioassays performed using sediments from Flanders. Semiempirical toxic unit analyses elucidated the effects of heavy metals (esp. organotin) and ammonia inhibiting H. incongruens growth. Based on standard approaches, 26% of the variance in the observed effect could be explained. Yet, by acknowledging speciation and interactions between bioassay constituents, we can explain up to 90%. Organotin and ammonia appear more and less toxic, respectively, than what is reported in the existing literature for controlled (single-substance) laboratory tests. We can attribute this to a cascade of bio-physicochemical interactions affecting speciation and transformations of the chemicals in situ. Prominently, organotins appear to bioaccumulate during the molting process of H. incongruens. Future modeling ought to focus on the factors controlling the interactions, via more explicit measurement and monitoring of speciation and bioavailability. This is likely to help in bridging the gap between laboratory and field tests so as to enable a more reliable risk assessment for sediments. Despite the need for further study, we deem our model a worthwhile complementary method to evaluate sediments in Flanders. Applying the approach to more spatiotemporal and geochemical settings[11,161] will evaluate its broader applicability. Thus, we deem it a starting point for more integrated ecotoxicity modeling in general.