Comparative toxicity of five dispersants to coral larvae

Oil spill responders require information on the absolute and relative toxicities of chemical dispersants to relevant receptor species to assess their use in spill response. However, little toxicity data are available for tropical marine species including reef-building corals. In this study, we experimentally assessed the sub-lethal toxicity of five dispersants to larvae of the coral Acropora millepora over three short exposure periods (2, 6 and 24 h) reflecting real-world spill response scenario durations. Inhibition of larval settlement increased rapidly between 2 and 6 h, and was highest at 24 h: EC50 Corexit EC9500A = 4.0 mg l−1; Ardrox 6120 = 4.0 mg l−1; Slickgone LTSW = 2.6 mg L−1; Slickgone NS = 11.1 mg L−1 and Finasol OSR52 = 3.4 mg L−1. Coral larvae were more sensitive to dispersants than most other coral life stages and marine taxa, but the toxic thresholds (EC10s) exceeded most realistic environmental dispersant concentrations. Estimating toxic threshold values for effects of dispersants on coral should benefit the decision-making of oil spill responders by contributing to the development of species sensitivity distributions (SSDs) for dispersant toxicity, and by informing net environmental benefit assessment (NEBA) for dispersant use.


Results
In uncontaminated water (control treatments) and in the presence of the lowest concentrations of dispersant, ~1 mm long larvae were typically cigar-shaped and highly motile (Fig. 1a). As the concentration and exposure time increased (for all dispersants) larvae became less motile, forming a squat, bullet shape with ill-defined and transparent outer membranes (Fig. 1b). Similar effects were observed for the reference toxicant SDS. After exposure to uncontaminated seawater (2, 6 and 24 h) and following the addition of chemical inducer, 85-95% of control coral larvae underwent successful attachment and metamorphosis into single polyp juveniles (Fig. 1c, Table 2).
The effects of exposure to all dispersants and SDS followed typical concentration-response curves with increasing concentrations (Fig. 2). Fitting the % inhibition data to 4-parameter sigmoidal equations (R 2 values 0.80-0.98, Table 3) enabled the interpolation of concentrations that inhibited metamorphosis in A. millepora larvae by 10% (EC 10 ) and 50% (EC 50 ) ( Table 3). Of the five dispersants, Slickgone LTSW was the most toxic (24 h EC 50 = 2.6 mg l −1 ), while Slickgone NS was the least toxic (24 h EC 50 = 11.1 mg l −1 ). Exposure to the lowest concentration of SDS (1.25 mg l −1 ) at 24 h caused 79% inhibition so EC XX values for SDS were not reliable at this time point.
The exposure duration had a strong effect on the toxicity of all five dispersants and SDS to coral larvae, with EC 50 values decreasing by an average 2.9-fold as the exposure duration increased from 2 to 24 h ( Table 3). The EC 50 for SDS also decreased dramatically as the duration of exposure increased from 2 h to 6 h ( Table 3). F-test comparisons from the fitted curves demonstrated a significant effect of exposure duration on EC 50 values for each dispersant and SDS ( Table 3). The time-dependent relationship between the EC x values and concentration was captured using a three parameter exponential decay function (Fig. 3). The EC x values can be predicted for other exposure times using Equation 1 with the estimated parameters in Table 4.

Discussion
Coral larval settlement was sensitive to acute exposures to all five individual oil spill dispersants formulations, with the threshold concentrations for toxicity (EC 10 s) all relatively low in comparison with other coral life stages and marine taxa. The toxicity of each dispersant increased with exposure duration until 6 h, indicating rapid uptake of the toxic components of the formulations and this could be related to the small size and high surface area to volume of the larvae. Although the toxic thresholds were low, they were higher than dispersant concentrations that would be expected in the real-world following applications in tropical waters. Differences in the toxicity observed between dispersants was likely due to the combined toxic effects of different formulation components and their relative concentrations. The toxic threshold values determined here for each dispersant should contribute to the development of species sensitivity distributions (SSDs) for dispersant toxicity and support net environmental benefit assessments (NEBAs) for dispersant applications, to improve decision making by oil spill responders.
Comparative toxicity between dispersants. All of the dispersant formulations were observed to affect larvae in a similar way. Settlement and metamorphosis was reduced at low concentrations, while impaired SCIeNTIfIC RePORTS | (2018) 8:3043 | DOI:10.1038/s41598-018-20709-2 mobility, changes to larval shape and increased membrane transparency were typically observed at higher concentrations. Although each dispersant formulation is different, similar toxic modes of action between formulations might be expected as all contain the anionic surfactant DOSS, which is likely to be a major contributor to toxicity 9,15,30 . The reference toxicant SDS caused a similar response pattern and (like DOSS) was likely to have damaged membrane integrity, leading to a failure of cellular homeostasis 30 . Additional impacts on membrane proteins, lipopolysaccharides and other cellular functions cannot be discounted 31 . Surfactants have a wide range of toxicities, with no effect concentrations of environmentally relevant endpoints reported between 0.01 and 100 mg l −1 , and sub-lethal responses often occurring at concentrations an order of magnitude lower 32 . The 24 h thresholds for toxicity (EC 10 ) of between 1.6-5.6 mg l −1 for the five dispersants were similar to the toxicities of other dispersants to coral larvae summarised in Table 1, but direct comparisons with these other studies are difficult as EC 10 s have not previously been calculated for corals from concentration response curves. The only other study to publish EC 10 or EC 50 values on the effects of dispersants to coral, indicated that Corexit EC9500A was more than 10-fold less toxic to adult fragments of a temperate octocoral 33 than to A. millepora larvae. The current study allowed direct comparisons of toxicity between the five dispersants to the same species under identical conditions. This matched dataset indicated that although there were differences in toxicity between the dispersants to coral larvae, the range of EC 50 s was relatively narrow (2.6 and 11 mg l −1 ) after 24 h. The acute toxicity of each formulation should be linked to the mixture toxicity (combined potency and concentration of surfactants and other components) of the dissolved fraction 30 , but without detailed chemical analysis (not available here), assigning specific reasons for these differences would be speculative. Time-dependent toxicity. To examine the time-dependency of toxicity, coral larvae were exposed to dispersants over three short periods representing short peaks in concentration expected during an oil spill response 19,28 . There was a rapid exponential change in toxicity between 2 and 24 h with the majority of the ~3-fold increase occurring between 2 and 6 hours (Fig. 3). The minor changes in toxicity between 6 and 24 h exposures indicate that maximum acute toxicity to the small coral larvae occurred rapidly for all of the dispersants tested.
In adult soft corals, Corexit EC9500A exposure caused bleaching (the loss of symbionts) but no time-dependent effects were observed between 24 and 72 h 34 . The rapid onset of toxicity for larvae could be related to their small size and high surface area to mass ratio, leading to fast uptake kinetics 35 . The acute toxicity results reported here are relevant to short-term field exposures expected in the majority of response scenarios where applications are short or tides and currents reduce exposure from longer applications 23,28 . The consistent exposures applied over three short periods were followed by settlement in uncontaminated water. Additional short and long term exposures tests should be conducted to assess a variety of other exposure scenarios. Acute exposures of larvae to spikes of dispersant followed by gradual dilution during settlement is a more likely natural scenario and should be tested, but deriving effect thresholds for comparison with other studies is more difficult. Extended low concentration exposure tests should also be conducted to assess the potential chronic effects posed by longer term dispersant responses 4 .
Toxicity relative to regulatory values. Ideally, responders to an oil spill would know the relative toxicity of each available dispersant in combination with the specific oil type for species relevant to the spill site 19 . Since these data are rarely available, the comparative toxicity of dispersant formulations to the relevant taxa (as measured here for corals) may be valuable for combination with modelled concentrations to assess risk and to help choose the least harmful dispersant for the response. Each national or state jurisdiction has its own regulations or guidelines on the appropriate use of dispersants. For example, some European nations require toxicity tests to demonstrate that dispersant formulations are less toxic than a reference toxicant or do not enhance the toxicity of oil in comparison to mechanically dispersed oil 17 . Guidelines set by the Australian Maritime Safety Authority (AMSA) under the National Plan limit the acceptability for use of dispersant products that show a toxic effect (EC 50 , LC 50 or similar) to a diversity of marine species, including microalgae, copepods, amphipods, sea urchins, scallops and fish, at concentrations of 10 mg l −1 or less 27 . Inhibition of larval settlement and metamorphosis is considered an ecologically relevant endpoint for water quality derivation 29 . If the 24 h coral larval exposure results (Table 3) were to be included within this AMSA multi-taxa approach, then only Slickgone NS would have met this requirement. Another approach that should be considered to compare and regulate dispersant safety involves the application of species sensitivity distributions (SSD); probability models of the sensitivity of multiple taxa to dispersant exposure. Adams 36 recently developed preliminary SSDs for a range of dispersants registered for use in Australia based on the very limited set of toxicity data available for marine species. The SSDs were used to determine the predicted no-effect concentration which was assigned as 95% species protection (PC95, this is equivalent to the HC5 where 5% of species are affected). The PC95s (Table 3), were generally derived from chronic EC 10 values and were of low to moderate quality due to the lack of toxicity data (only 3 to 7 species available per dispersant). More   toxicity data is available for Corexit EC9500A and a PC95 of 6.6 mg l −1 was derived by Barron et al. 37 . A comparison of the acute EC 10 s for coral larval settlement with the sensitivity data reported by Adams 36 and reviewed by Hook and Lee 15 , indicates that coral larvae are relatively sensitive to Ardrox, Slickgone LTSW, Slickgone NS and Finasol compared with many other species tested, especially if the acute toxicity thresholds derived here were converted to chronic values (by dividing by a factor of 2-5) 29 . However, the PC95 values derived from limited toxicity data would be broadly protective of coral larvae to these dispersants. Coral larvae were also relatively sensitive to Corexit EC9500A in comparison to many other fish, mollusc and crustacean species 37,38 but would not be protected by the PC95 derived by Barron et al. 37 . When a NEBA is undertaken to assess whether to apply dispersants to help degrade oil slicks and protect intertidal and shoreline habitats, it requires input of a very broad range of information that includes the type and size of a spill, weather conditions, hydrographic and oceanographic conditions, the relative presence and distribution of vulnerable receptors to both dispersed and undispersed oil, accessibility and availability of application, and monitoring resources and predictive modelling 9,15 . Dispersant application may increase the exposure of sub-surface species to greater doses of petroleum hydrocarbons, and this additional exposure and potential harm from dispersants themselves is also a critical consideration for the NEBA process 19 . While specific studies should continue to be conducted to compare the toxicity of physically versus chemically dispersed oil, the current study demonstrated that these five dispersants were approximately an order of magnitude less toxic to coral larvae than dissolved aromatic hydrocarbons from a water accommodated fraction of light crude oil in very similar tests 25 .  Table 3. This is consistent with the general notion that dispersants alone are less toxic than oil, but chemically dispersed oil is more toxic due to greater concentrations of oil in water 5,12,39 . Comparing toxic thresholds for dispersants derived from laboratory studies with concentrations measured in the field are particularly difficult as dispersant formulations are comprised of multiple potentially toxic components. These components are likely to retain their original proportions in short acute experiments, but these proportions will change rapidly when formulations are applied in the field. Field and laboratory comparisons are further complicated as concentrations can be expressed as dilutions of the formulation (which can only be modelled in the field) or as concentrations of marker/proxy components, such as DOSS (which doesn't account for different behaviour and toxicity of other components). The highest concentration of Corexit EC9500A at the water surface soon after application by air (dispersant:oil ratio 1:20) has been estimated as ~5 mg l −140 and up to 13 mg l −1 of dispersant measured in sea trials 41 . The EC 10 thresholds for 24 h exposure of coral larvae to dispersants were 1.6-5.6 mg l −1 , indicating that coral larvae in the immediate vicinity of dispersant application could be at risk. However, rapid dilution at sea is expected to reduce exposure concentrations 1,28,41 . For example, measured concentrations of DOSS as a presumed marker for Corexit EC9500A and EC9527A (since other sources of DOSS were identified in the Gulf 42,43 ) reported in weeks following the Deepwater Horizon spill were far less than the EC 10 values for dispersant formulations reported here for coral larval settlement 44,45 . Nevertheless, more tropical species toxicity data, especially for sensitive reproductive stages of reef-building corals and for relevant exposure durations, are crucial to improve the SSD approach to assess environmental safety and to contribute to a more robust information base for the NEBA process and decision support for oil responders 15,16,36,37 . Other operational considerations reinforce this conclusion. For example, Australian dispersant operations protocols recommend that dispersants not be applied without sufficient depth of water, or water exchange, or near coral reefs 1,9,46 . The most sensitive reproductive processes for many reef-building corals occur during and soon after discrete mass spawning events that usually occur annually at a given reef, and where spawn and larvae are present in the water column between reefs for days to weeks 47 . This limits the likelihood of exposing mobile coral larvae to oil spills for most of the year, but dispersant applications during the annual coral spawning seasons (even long distances from coral reefs) should be carefully considered in the NEBA process.

Materials and Methods
Coral larvae were exposed to eight concentrations of five different dispersants for durations of 2, 6 or 24 h. After these exposure periods the larvae were rinsed and the larvae allowed to settle and undergo metamorphosis over an additional 18 h in uncontaminated seawater. The concentrations of dispersant that inhibited metamorphosis by 10% and 50% (EC 10 and EC 50 ) were calculated from concentration-response curves for each exposure duration.

Coral collection and larval culture. Colonies of the common Indo-Pacific broadcast spawning coral
Acropora millepora (Ehrenberg, 1834) >20 cm were collected from ~3 m depth in October 2015 from Esk Island, on the central Great Barrier Reef (GBR, 18°46.420′S 146°31.372′E). This species has been used in similar larval assays for over a decade and has predictable spawning and settlement behaviour [48][49][50][51][52] . Gravid colonies were transported to the National Sea Simulator (SeaSim) at the Australian Institute of Marine Science (Townsville, Australia) and placed in flow-through tanks at ~27 °C until spawning. Gametes were collected from seven parental colonies on a single night, fertilized and the symbiont-free larvae were cultured at less than 500 larvae l −1 in flow through tanks as previously described 49 . A. millepora larvae reach maximum competency for settlement after six days (reviewed in Jones et al. 47 ) and seven day old larvae were used in these exposure experiments.  Dispersant preparation. Dispersants (Table 5)    are also assisted by expressing the treatments in mg (formulation) l −1 . Sodium dodecyl sulphate (SDS), an anionic surfactant, was prepared in the same way and used as a reference toxicant.

Settlement assays.
Coral larvae were statically exposed in acetone-rinsed 20 ml glass vials containing 10-12 coral larvae and 15 ml dispersant solutions with six replicate vials used for each dispersant concentration.
Vials were sealed with aluminium-lined caps and a ~7 ml headspace allowed oxygen exchange. Vials were then transferred to an orbital incubator/shaker (50 rpm) to maintain gentle water movement and to discourage settlement 25 . Temperature was set at 27 °C and light at 40 µmol quanta m −2 s −1 . Under these conditions dissolved oxygen was maintained at >7 mg l −1 , and pH (8.0-8.2) and salinity 33-35 psu (Table S-1, Supplementary information). Three exposure durations were tested with vials removed from the incubator after 2, 6 or 24 h exposures.
To enable the testing of multiple treatments, the start times for each group of exposure durations (24 h, 6 h and 2 h in sequence) were offset by 2 h. Control (uncontaminated seawater) exposures were conducted (n = 6) for each dispersant type at each time point. Following the exposures, larvae were transferred into small (15 mm wide) 100 µm nylon mesh filters which were partially submerged and the larvae gently washed with an excess of uncontaminated FSW. The dispersant-free larvae were then transferred into individual 6-well cell culture plates containing 10 ml FSW (12 ml, Nunc, NY, USA). Settlement and metamorphosis of coral larvae in the plastic well plates was initiated by the addition of a slightly sub-optimal (to maximise the sensitivity of the assay) concentration (5 µl) of crustose coralline algae (CCA) extract 51 prepared using 4 g of the crustose coralline algae Porolithon onkodes 48 . Following exposure the settlement inducer, larvae cease swimming within minutes and elongate before undergoing early metamorphosis within 12 h 48,53 . Metamorphosis was assessed after 18 h and larvae scored as normal and functional if they had changed from either the free swimming or casually attached pear-shaped forms to squat, firmly attached, disc-shaped structures with pronounced flattening of the oral-aboral axis and with septal mesenteries radiating from the central mouth region 48 . This assay therefore assessed whether larvae were functional following the dispersant exposure. Larval settlement above 70% in the controls was considered acceptable as an endpoint based upon several previous studies using CCA or extracts of CCA to initiate settlement of Acropora spp. 25,54-56 . Data analysis. Inhibition of metamorphosis (% inhibition relative to 0% WAF control) was calculated from treatment data as Inhibition (%) = 100 × [(% metamorphosis control − % metamorphosis treatment )/% metamorphosis control ]. The concentration of dispersant that inhibited 10% and 50% of metamorphosis (EC 10 and EC 50 ) was calculated from concentration-response curves (four-parameter sigmoidal models) fitted to the % inhibition and total aromatics data of each treatment using the program GraphPad Prism (v7, San Diego, USA). The probability that EC 50 values generated by the logistic curves were statistically different was tested by applying the F test in Graph Pad Prism v6. EC 50 values were considered different when p < 0.05.
A Monte Carlo approach was carried out using the R software 57 to propagate uncertainty in the EC 10 and EC 50 thresholds (calculated independently for each experimental exposure time) and to provide parametric model fits (and associated uncertainty) for estimating thresholds as a continuous function of exposure time that might be more readily incorporated into oil spill modelling. This was achieved in two steps: (i) a gamma distribution was fitted to the mean, lower and upper confidence bands for the EC 10 and EC 50 estimated threshold concentrations for each experimental time; (ii) secondly 1000 random new concentration values were generated using these fitted distributions for each of the three exposure times, and fitted these using a three parameter exponential decay function:  Table 4. Parameter estimates for the time-dependence of EC 10 and EC 50 concentrations for each dispersant for the three parameter exponential decay function (y = m + a * exp(−b * log(x)) where y is the EC 10 or EC 50 threshold concentration, x is exposure time in hours, exp(−b * log(x)) is the exponential function, m is the concentration at which an infinite exposure time would theoretically yield no effect, a is the theoretical threshold concentration at zero exposure (y-intercept) and b is the exponential rate of decay. 95% confidence intervals in parentheses.

SCIeNTIfIC
where y is the EC 10 or EC 50 threshold concentration, x is exposure time in hours, a is the initial value, exp(−b * log(x)) is the exponential function and m is the concentration at which an infinite exposure time would theoretically yield no effect. Both the gamma and non-linear relationships were fitted using simple least squares procedures via the function optim in R.