The photoenhanced toxicity of weathered Alaska North Slope crude oil (ANS) was investigated in the eggs and larvae of Pacific herring (Clupea pallasi) with and without the chemical dispersant Corexit® 9527. Oil alone was acutely toxic to larvae at aqueous concentrations below 50 μg/L total polycyclic aromatic hydrocarbons (tPAH), and median lethal (LC50s) and effective concentrations (EC50s) decreased with time after initial oil exposure. Brief exposure to sunlight (∼2.5 h/d for 2 d) significantly increased toxicity 1.5- to 48-fold over control lighting. Photoenhanced toxicity only occurred when oil was present in larval tissue and increased with increasing tPAH concentration in tissue. Ultraviolet radiation A (UVA) treatments were less potent than natural sunlight, and UVA + sunlight caused greater toxicity than sunlight alone. The toxicity of chemically dispersed oil was similar to oil alone in control and UVA treatments, but oil + dispersant was significantly more toxic in the sunlight treatments. The chemical dispersant appeared to accelerate PAH dissolution into the aqueous phase, resulting in more rapid toxicity. In oil + dispersant exposures, the 96-h no-observed-effect concentrations in the UVA + sunlight treatment were 0.2 μg/L tPAH and 0.01 μg/g tPAH. Exposure of herring eggs to oil caused yolk sac edema, but eggs were not exposed to sun and UVA treatment did not cause phototoxicity. These results are consistent with the hypothesis that weathered ANS is phototoxic and that UV can be a significant and causative factor in the mortality of early life stages of herring exposed to oil and chemically dispersed oil.
Traditionally, toxicological studies used to define the hazards of polycyclic aromatic compounds and oil have been conducted in the absence of ultraviolet radiation (UV) . Laboratory studies have demonstrated that the toxicity of specific polycyclic aromatic compounds, oil products, and weathered oil increases 2 to greater than 1,000 times in the presence of UV compared with standard laboratory lighting conditions with fluorescent lights and minimal UV [2,3]. Photoenhanced toxicity occurs at the UV wavelengths and intensities that occur in the water column of aquatic environments, i.e., ultraviolet radiation B (UVB; 280–320 nm) and UVA (320–400 nm) . The photoenhanced toxicity of oil in fish and aquatic invertebrates appears to occur through activation of chemical residues that have bioaccumulated in aquatic organisms (termed photosensitization) rather than through photomodification of the chemical in water . The specific region of the UV spectrum that activates polycyclic aromatic hydrocarbon (PAH) compounds is considered to be the wavelengths of active UV absorption by the chemical. Current research indicates this includes both UVA and UVB , with UVA possibly the more important because of less rapid attenuation and greater PAH activation .
In a hazard assessment of oil and UV in Alaska waters, Barron and Ka'aihue  concluded that photoenhanced toxicity of spilled oil may occur in Prince William Sound and Gulf of Alaska waters because of the potential for significant levels of oil and UV in the water column. This finding was important because the laboratory studies used to assess the impacts of the Exxon Valdez oil spill in Prince William Sound did not account for photoenhanced toxicity nor do available laboratory databases on the toxicity of oil to Alaskan marine species [9–12]. Pelletier et al.  showed that unweathered Prudhoe Bay crude oil was greater than 100 times more toxic to shrimp and bivalve embryos when tested under UV. Several factors that may modulate photoenhanced toxicity of the aqueous phase of crude and refined oils require elucidation. These include identification of phototoxic components of petroleum, the variability of organism sensitivity with species and life stage, oil exposure methods, and the intensity and quality of UV. We conducted the study reported here to evaluate some of these factors, including the photoenhanced toxicity of chemicals dispersants, life stage, and artificial UVA compared with natural sunlight as UV sources.
The interaction of chemical dispersants used in spill response on the photoenhanced toxicity of oil has not been previously investigated. Chemical dispersants break up free product oil into small droplets (e.g., 0.01–50 μm), which disperse in the water column. Dispersants generally increase the total concentrations of petroleum compounds (dissolved + particulate oil), but the relative environmental hazards of chemically dispersed and nonchemically dispersed oil are uncertain and are likely spill specific . To date, laboratory studies determining the toxicity of chemically dispersed oil have only been conducted under standard light conditions (minimal UV). These studies may substantially underestimate the risks of chemically dispersed oil in the environment because of the potential for photoenhanced toxicity.
The primary objective of the current study was to investigate the photoenhanced toxicity of a weathered Alaska North Slope crude oil (ANS) to larvae of the Pacific herring (Clupea pallasi) exposed to UVA and to determine the relative toxicity of chemically dispersed and aqueous-phase oil under these exposure conditions. Additional experiments were performed to determine the relative sensitivity of eggs and larvae of herring exposed to aqueous-phase oil followed by artificial UVA treatment and the effects of artificial UVA compared with natural sunlight on herring larvae after exposure to aqueous-phase oil. Also, five combinations of UV and oil exposure were evaluated to determine if weathered ANS had a photo-sensitization mode of action (UV activation of bioaccumulated residues) or a photomodification mode of action (UV photo-oxidation to form more toxic molecules in the water column). Based on prior studies with oil  and components of oil , we hypothesized that weathered ANS would have a photo-sensitization mode of action. Herring were studied because they are ecologically and economically important in Alaska. Both life stages of herring were evaluated separately to allow a comparison of life-stage sensitivity. Herring eggs are spawned in the photic zone of near-surface tidal areas potentially impacted by oil spills and other sources of PAHs, and larvae develop and grow in these areas. Bioaccumulation of total PAHs was quantified in both eggs and larvae to allow a comparison of exposure and sensitivity based on tissue residues in addition to aqueous concentrations.
MATERIALS AND METHODS
Pacific herring were exposed to a series of aqueous-phase doses prepared with high-energy mixing of ANS with chemical dispersant either present or absent. Oil exposures occurred at only one life stage, i.e., either in embryos or larvae. Following oil exposure, larvae and eggs were exposed to control lighting or artificial UVA. Additional UV treatments were performed with larvae in natural sunlight (∼2.5 h for 2 d) or UVA + sunlight. Toxic concentrations of oil were reported as total PAH (tPAH) concentrations in exposure water and as wet weight concentrations in tissue. Mean dry/wet weight ratios were 0.211 (0.118–0.388) for eggs and 0.087 (0.068–0.108) for larvae. Tests were performed under environmentally realistic conditions of temperature, salinity, oil exposure concentrations, and UV doses.
Reproductively ripe Pacific herring were collected by dip net from a purse seine net in Sitka Sound, Alaska, USA, on March 24, 2001. Herring were placed over ice in coolers, transported to the laboratory, and artificially spawned within 7 h. Gametes from 4 of the 10 females and several males included in the experiment were dissected from fish before transportation and were transported chilled, similar to the methods of Yanagimachi et al. . Eggs were spawned as described by Carls et al. . In brief, eggs from 17 females were spawned onto glass microscope slides (2.5 × 7.6 cm). Fertility was qualitatively assessed the following day, and slides from 10 females with good fertility and egg abundance were selected for study. Slides from each female were randomly distributed among 18 slide racks, and excess eggs along slide margins and in areas more than one layer deep were removed. Additional eggs were spawned onto 202-μm plankton net to provide eggs for uptake measurements and larval experiments. Laboratory seawater was supplied by a single-pass flow-through system with a – 24-m intake in Auke Bay, Alaska. All water was passed through 1-μm polyester filters before use. Hatching larvae were collected every 1 to 2 d (1–3 d for uptake studies) to provide fish of known age for separate larvae experiments. Eggs and larvae were reared at 4.9 to 6.6°C and 30.1 to 31.2 ppt salinity. Dissolved oxygen ranged from 89 to 98% of saturation and was not affected by the presence of eggs or larvae.
Preparation of dispersed and aqueous-phase oil solution
Water-accommodated fractions of oil were prepared in 50-L, cone-bottomed fiberglass tanks fitted with 2.5-cm polyvinyl chloride ball valves at the bottom. Tanks were partially filled with 32 L of seawater filtered through 1-μm polyester filters, which were presoaked for 24 h in flowing seawater, rinsed with freshwater, and dried before use. A motor-driven 5.7-cm paddle shaft extended from above the water surface to within approximately 2.5 cm above the tank bottom. Shaft speed was approximately 600 rpm and resulted in vigorous mixing. The vortex extended from the water surface down to the paddle, and a mist formed above the surface. Mixing began a few minutes before addition of oil and dispersant. Separate water-accommodated fractions (WAFs) were prepared with either oil (300 μl) or oil + dispersant (300 μl oil + 12 μl Corexit 9527; Nalco/Exxon Energy Chemicals, Sugar Land, TX, USA). Our 1:25 ratio of dispersant:oil compares with the generally recommended ratios of 1:10 to 1:50. The oil:water ratio (0.3 ml oil to 32 L water) was selected based on preliminary analyses of part per billion concentrations of tPAH, which was known to be toxic to herring embryos .
Oil was added near the margin of the vortex with a positive-displacement pipette, followed immediately by dispersant using a gas-tight syringe at the vortex margin. Mixing time was 12 h in all tests, followed by a 1-h separation period. Mixtures were drained from the bottom of the tanks through a clean, presoaked 1-μm polyester filter, and a sample of filtered WAF was collected in 4-L hydrocarbon-free glass jugs, spiked with a surrogate standard, and extracted immediately. Stock WAF was chilled to testing temperature, then diluted with filtered seawater to achieve each desired oil concentration. For the larval experiments, an additional high WAF concentration was prepared as described above but using 3,000 μl oil and 120 μl dispersant.
Exposures. Herring eggs were exposed to WAF of weathered ANS, chemically dispersed WAF, and UVA in a three-factor experiment. There were six oil levels (including controls), two dispersant levels (absent, present), and three UV treatments (controls, 2.5 h UVA, 15 h UVA). Each treatment was conducted in triplicate with approximately 100 eggs per slide in 1-L wide-mouth jars. Eggs from each female were present in each WAF × dispersant combination. Static exposure of eggs to WAF began approximately 44 h after fertilization and continued for 4 d, followed by UV treatment. Racks containing slides with eggs from each female were randomly transferred into 3-L jars containing chilled WAF dilutions and suspended from mobile overhead slide racks . Nylon plankton netting that contained additional eggs for determining bioaccumulation was also suspended in each jar. Maximum biomass was estimated as 1.6 to 3.2 g tissue/L during the first 2 d and 0.6 to 0.8 g/L thereafter. Eggs were exposed to hydrocarbons for 4 d with daily WAF renewal, then transferred to clean seawater and randomly allocated to one of three UV treatments (each replicated three times). Each replicate consisted of a glass slide with approximately 100 eggs placed inside a 1-L wide-mouth glass jar. Nets with fertilized eggs were rinsed in seawater and frozen in hydrocarbon-free jars for analysis. Qualitative fertilization success was determined, without knowledge of treatment, at 9 to 11 d after spawning.
Endpoints. Hatch timing and success, larval viability, and larval abnormalities were observed daily during peak hatch (which peaked sharply on April 24, 2001) without knowledge of treatment. Observation frequency was reduced to 2- or 3-d intervals during periods of low hatch, and larvae were collected only if five or more were present per jar. Living larvae were assessed for swimming ability and gross morphological deformities, anesthetized with tricaine methanesulfonate, and preserved in 5% phosphate-buffered formalin. Dead larvae were enumerated and discarded. At each observation, the slide with eggs was transferred within a few seconds to a jar containing freshly filtered (1 μm) seawter. The swimming ability of live larvae was categorized as effective, ineffective, or incapable. Effective swimmers were active, frequented the water column, and avoided capture. Ineffective swimmers were more lethargic and were more likely to be found on jar bottoms. Incapable larvae were unable to swim in a straight line and were often only capable of spasmodic twitching. Preserved stage 1a and 1b larvae (yolk sac ≥ 2.5 diameter of muscle mass) were indiscriminately subsampled blind from each female in each treatment group to score yolk sac edema, indicated if the anterior margin of the yolk membrane was bounded by an area of clear fluid or irregularities in yolk shape (n = 40 per replicate except where fewer larvae were available). After hatch was complete (as indicated by no viable eggs remaining), any dead embryos were inspected and enumerated.
Herring larvae were exposed to WAF of weathered ANS or WAF + dispersant and UV in a three-factor experiment. There were four oil levels (including controls), two dispersant levels (absent, present), and four UV treatments (control, 15-h UVA, sunlight, UVA + sunlight). Each treatment was conducted in triplicate with approximately 40 larvae in 1-L wide-mouth glass jars; biomass loading was <0.1 g/L. Larvae (1-2 d old) were exposed to WAF for 24 h, followed by UV treatment. Numbers of live, moribund, dead, and swimming larvae were enumerated daily for 8 d following UV treatment; exposure water was not replaced during this period. The total number present was determined at the end of testing to calculate percent responses. Bioaccumulation of petroleum hydrocarbons was determined in separate groups of 1- to 3-d-old larvae (∼2,100 larvae/jar) at three dose levels (control, two highest WAF treatments) in 3-L jars for 24 h.
Laboratory UV exposures
Exposure system. The laboratory UV exposure system consisted of paired tanks (225 cm × 55 cm) with black bottoms and reflective sides (aluminum foil, dull side out). Fluorescent lamps were positioned approximately 60 cm above tank bottoms. Exact positioning of the light fixtures above the tank was determined in preliminary measurements to minimize variation of UV radiation within the exposure system. Each tank was covered with a cracked-crystal-style light panel (0.4 cm, translucent white) as light diffusers and subdivided into a UVA chamber (130 × 55 cm) and a control light chamber (65 × 55 cm) by an approximately 30-cm black plastic partition consisting of two layers of black plastic and aluminum foil lining. Each UVA chamber was illuminated with two UVA lamps (GE 123-cm F40BLB; General Electric, Louisville, KY, USA) and two Chroma 50 lamps (GE 123-cm F40C50) oriented longitudinally. Spectral output of the UVA lamps was approximately 320 to 420 nm, with maximum intensity at 340 to 380 nm (Fig. 1). Control light chambers were illuminated with two Chroma 50 lamps, which provided visible light plus minimal UV. Light diffusers were positioned immediately below the lamps (∼60 cm above tank bottom). Flowing seawater provided temperature control in each chamber.
Light measurements. A broad wavelength radiometer (Macam UV 203; Macam Photometrics, Livingston, Scotland) was used to measure visible light (400-700 nm, square-wave response) and ultraviolet radiation (UVA, 320-400 nm, max response at ∼365 nm; UVB, 280-320 nm, maximum response at ∼315 nm). Light intensity was measured with the detector positioned at the bottom of a test jar filled with filtered seawater. Light intensity was measured in multiple positions in each light chamber and at multiple times to capture the variability associated with tank position and measurement time. Laboratory light exposures were designed to determine if environmentally relevant doses of UVA would cause phototoxicity in the absence of significant UVB. Control and UVA treatments provided nearly identical UVB and visible light but substantially greater UVA in the UVA treatment. For both egg and larval experiments, light intensities for the control light exposures ranged from 10 to 16 μW/cm2 UVA, 0.2 to 0.6 μW/cm2 UVB, and 380 to 640 μW/cm2 visible light. Light intensities in the UVA exposures ranged from 470 to 650 μW/cm2 UVA, 0.2 to 0.5 μW/cm2 UVB, and 370 to 560 μW/cm2 visible light. Ambient UV in the laboratory was very low, with light intensities ranging from 0.2 to 1.0 μW/cm2 UVA, 0.001 to 0.02 μW/cm2 UVB, and 29 to 130 μW/cm2 visible light.
UV treatments. Egg UV treatments were initiated in clean seawater within 1 h of the termination of WAF exposure; larval UV treatments began after approximately 24 h of WAF exposure without transferring the fish to clean seawater. Larvae were not transferred to clean water to avoid potential injury. Eggs and larvae were exposed to control light or UVA for 15 h, and a separate group of eggs were exposed to UVA for 2.5 h. UV doses (μW·h/cm2) were computed from the specific light intensities measured during each exposure multiplied by the duration of light exposure. All treatments occurred in temperature-controlled tanks. Test jars were placed under ambient lighting in the laboratory when not being exposed to UV treatments.
Table Table 1.. Range of UV and visible light doses (μW·h/cm2) in each larval light treatment
Visible light dose
aUVA = ultraviolet radiation A (320-400 nm); UVB = ultraviolet radiation B (280-320 nm).
UVA + sunlight
Sunlight exposures were only performed in the larval experiments. The sunlight exposure system consisted of a temperature-controlled 1.0-m diameter × 19-cm-high aluminum-lined basin filled with filtered laboratory seawater. Ambient sunlight exposures were conducted on two sequential days in Juneau, Alaska, and ranged from 1.6 to 2.8 h/d, with the duration of exposure adjusted to provide similar daily doses of ambient UVB. Light intensities were measured every 15 to 30 min depending on the variation in ambient conditions, with the radiometer detector positioned at the bottom of a test jar filled with filtered seawater. The environmental conditions ranged from complete to partial clouds, solar disk visible to not visible, haze, and occasional rain. Average light intensities during ambient sunlight exposures ranged from 570 to 1,250 μW/cm2 UVA, 7.0 to 14 μW/cm2 UVB, and 4,850 to 11,600 μW/cm2 visible. Ultraviolet doses (μW·h/cm2) were computed from the average light intensity measured during each exposure multiplied by the duration of light exposure. Table 1 summarizes the light treatments used in the larval experiments.
A separate larval experiment was conducted to discriminate between two possible modes of action of the photoenhanced toxicity of oil, photosensitization (activation of hydrocarbons in tissue), or photomodification (photooxidation of hydrocarbons in water). These experiments were performed similarly to the main larval experiment to facilitate interpretation of the toxicity in the presence of both UV and oil. Five treatment combinations of WAF (larval exposure to WAF either before or after UV exposure) and UV (sunlight present or absent) were evaluated for each dispersant treatment group (dispersant present or absent): (1) no WAF in water or larvae, no sunlight, (2) no WAF in water or larvae, 4 h sunlight, (3) WAF in water and larvae (24-h preexposure to WAF), no sunlight, (4) WAF irradiated 4 h with sunlight, then larvae with no prior WAF exposure were added, and (5) no WAF in water, WAF in larvae, 4 h sunlight. Aqueous tPAH concentrations in WAF exposures were 8.0 μg/L for oil-only treatments and 9.0 μg/L for oil + dispersant treatments. The UV exposures occurred in sunlight (4 h, 1,350 μW/cm2 UVA, 14 μW/cm2 UVB, and 14,900 μW/cm2 visible) with mostly sun, some haze, and solar disk always visible. Exposures with no sunlight were conducted under ambient laboratory lighting of 0.59 μW/cm2 UVA, 0.009 μW/cm2 UVB, and 79 μW/cm2 visible.
Water and tissue samples were extracted with dichloro-methane after addition of six internal standards according to the methods of Short et al. . Isolation and purification of calibrated and uncalibrated compounds were completed by silica gel/alumina column chromatography followed by size-exclusion high-pressure liquid chromatography and fractionation; seawater samples were not fractionated by high-pressure liquid chromatography. Extracts of PAHs were separated and analyzed by gas chromatography equipped with a mass selective detector. Calibrated PAHs were identified by retention time and two mass fragment ions characteristic of each PAH and were quantified using a five-point calibration curve. Uncalibrated PAH homologs (which included alkyl-substituted isomers of naphthalene, fluorene, dibenzothiophene, phenanthrene, and chrysene) were identified by retention time and the presence of a single-characteristic mass fragment ion. Uncalibrated PAHs were quantified by using calibration curves of their respective parent homologs. Experimentally determined method detection limits depended on sample weights and generally were 1 ng/g in tissue and 1 to 8 ng/L in water. Concentrations below the method detection limits were treated as zero. Tissue concentrations are reported on a wet-weight basis, but wet- to dry-weight ratios were measured by dehydrating 1 g wet samples for 24 h at 60°C and weighing the remaining mass. The accuracy of the hydrocarbon analyses was about ± 15% based on comparison with National Institute of Standards and Technology (Washington, DC) values, and precision expressed as coefficient of variation was usually less than about 20%, depending on the PAH. Total PAH concentrations were calculated by summing concentrations of individual PAHs. Relative PAH concentrations were calculated as the ratio of PAH concentration to the tPAH concentration.
Relationships among larval and embryo responses (e.g., mortality and edema) to experimental treatments were examined with analysis of variance to identify significant treatment effects based on mean responses. Where overall factors (oil, dispersant, or light) were significant, treatment groups were compared with controls with pairwise comparisons and the acceptance criterion for statistical significance (α) was 0.05 divided by the number of comparisons. The lowest-observed-adverse effect concentration (LOEC) was defined as the lowest treatment (e.g., oil) where response was significantly different than in controls and the no-observed-effect concentration (NOEC) was the next lower treatment. Percentage data were arc-sine transformed before analysis of variance and corrected for small n, but the same general conclusions were reached with untransformed data. Calculation of both LC50 and EC50 concentrations required two models (logit and Spearman-Karber). Not all data could be analyzed with the logit method because iterations did not always converge to a solution, and not all data could be analyzed with the Spearman-Karber method because responses did not always range from 0 to 100%. Treatment responses were corrected by corresponding control response  before LC50 and EC50 calculation. Regressions were applied to relate tPAH concentrations in tissue to initial aqueous tPAH concentrations and alkane concentrations in tissue to initial aqueous alkane concentrations (log-log transforms). Alkane concentrations were restricted to calibrated compounds (C10- to C34-alkane, pristane, and phytane) to avoid confusion with nonpetroleum alkanes in tissue.
Percent response was based on the total number of larvae present in a given treatment at the beginning of the test in larval experiments. In egg experiments, embryo responses included percent hatch, percent moribund plus dead, percent with spinal abnormalities, percent effective swimmers, and yolk sac edema. Percentages of eggs that hatched were based on the initial number of eggs determined before hatch. Hatched larvae were categorized as live, moribund (not moving), or dead (tissue necrosis evident), and percentages were based on the total number of larvae enumerated during hatch including dead embryos remaining on slides. Percentages of effective swimmers were based on the number of live larvae, i.e., percentages of larvae with spinal aberrations were based on the total of live and moribund larvae. Larvae were not inspected for morphological damage.
The extent of PAH dissolution in the exposure WAFs was evaluated by calculating the ratio of phytane and total PAH (phyt:PAH). Phytane is a branched aliphatic hydrocarbon of very low solubility (<1 μg/L) and is used here as a proxy for the amount of whole oil in oil-phase droplets; PAHs are more soluble. The phyt:tPAH is 0.12 for unweathered ANS, and values of the ratio decrease as water extracts of oil become enriched with dissolved PAH in addition to any PAH associated with dispersed, whole-oil droplets.
Effects on larvae
Oil-only effects. Water-accommodated fractions of weathered ANS were directly toxic to herring larvae and exhibited a dose-response relationship of increasing mortality and morbidity with both increasing tPAH in water and tPAH residues in tissue (Fig. 2). Under control lighting, the 96-h NOECs for larvae were 86.5 μg/L tPAH and 34.8 μg/g tPAH, but there was substantially greater mortality than in the no-oil controls (bottom panel; Fig. 2). The control results are the lowest measured tPAH concentrations in Figure 2, i.e., 0.013 μg/g tissue and 0.03 μg/L (oil-only treatment) and 0.18 μg/L (oil + dispersant treatment).
The LC50s decreased over time (p <0.001; Fig. 3). For example, LC50s in the oil-only control-light treatment decreased approximately 50% at 4 d to 26 μg/L at 8 d. The EC50s, based on impaired swimming in addition to mortality and morbidity, were similar to LC50 values (Fig. 3).
UV + oil effects. The toxicity of ANS was enhanced by exposure to sunlight, and the degree of photoenhanced toxicity increased with increasing UV treatment (UVA + sunlight > sunlight > UVA) and with increasing tPAH concentration in tissue. Sunlight, which had approximately 50% of the UVA and six times more UVB than the UVA treatment, caused significantly greater photoenhanced toxicity than UVA alone (Fig. 2). For example, in the sunlight treatment, there was >90% mortality and morbidity in larvae exposed for 4 d to 87 μg/L, 63% affected in the UVA treatment, and 37% affected under control light (Fig. 2). The sunlight treatment significantly lowered 8-d LC50s and EC50s to approximately 50% of the control light treatment (Fig. 3). The combination of UVA and sunlight caused the greatest enhancement of ANS toxicity. The LC50s, EC50s, and LOECs were 18- to 450-fold lower in the UVA + sun treatment compared with control lighting with minimal UV. The 96-h NOECs in the UVA + sun treatment were 0.7 μg/L and 0.1 μg/g tPAH in oil-only exposures.
Oil + dispersant effects. As in oil-only exposures, oil + dispersant treatment caused a dose-dependent increase in mortality and morbidity (Fig. 2). Under control lighting, the toxicity of chemically dispersed oil was similar to oil-only toxicity at equivalent tPAH concentrations in tissues and in water (Figs. 2 and 3). The 96-h NOECs in the oil + dispersant treatment were 9.2 μg/L and 2.5 μg/g, and 96-h LOECs were 440 μg/L and 70.7 μg/g tPAH (Fig. 2). Mortality in dispersant controls (0.4 and 4 μl/L Corexit 9527) was not significantly greater than mortality in water-only controls in any light treatment (0.658 ≤ p ≤ 0.935) and averaged 10 ± 4% across all light treatments at 4 μl/L after 8 d of exposure. Statistical comparison of EC50 and LC50 values (Fig. 3) indicated there was no significant difference between oil-only and oil + dispersant treatments under visible light.
UV + oil + dispersant effects. Chemically dispersed oil was significantly more toxic in the sunlight treatment compared with oil-only exposures with sunlight (Fig. 3). As in oil-only exposures, toxicity increased with increasing UV treatment (sunlight + UVA > sunlight > UVA). Total PAH concentrations in larvae in the oil + dispersant treatment were elevated approximately two times above oil-only exposure, resulting in increased mortality and morbidity (Fig. 2) and more rapid mortality in sunlight and UVA + sunlight treatments (data not shown). In the highest UV treatment (UVA + sunlight), the 96-h NOECs were 0.2 μg/L and 0.01 μg/g in oil + dispersant exposures (Fig. 2).
The methods used to prepare the exposure WAFs resulted in different PAH compositions depending on the amount of oil added initially and whether chemical dispersant was present. The WAF prepared from 3,000 μl oil without dispersant resulted in a PAH composition only slightly different than whole ANS (Fig. 4a and d), with a phyt:tPAH of 0.041 and a tPAH concentration of 87 μg/L. This WAF was dominated by the naphthalenes, as was whole ANS. Addition of dispersant to the 3,000 μl oil preparation dramatically altered the PAH composition and increased the tPAH concentration to 440 μg/L. The phyt:tPAH was 0.021 and contributions from the naphthalene and phenanthrene homologs were nearly equal, indicating much greater dissolution of three-ring PAH than when dispersant was absent (Fig. 4d and e). Some losses to the atmosphere of the lowest molecular weight PAH were also evident with addition of dispersant to these high-oil treatments.
In contrast with the 3,000-μl oil treatments, the presence of chemical dispersant during WAF preparation had less effect on WAF chemistry prepared from 300 μl oil. Despite an overall similarity among PAH homologs, differences within homologous groups between oil-only and chemically dispersed oil were apparent (Fig. 4b and c) and the phyt:tPAH was significantly higher without dispersant (0.070 ± 0.006) than with dispersant (0.051 ± 0.006, p < 0.02), which indicates the dispersant promoted PAH dissolution. Both 300-μl preparations had more losses of naphthalene homologs, with the phenanthrene homologs dominant or equal, when compared with the 3,000 μl oil preparations. Dispersant promoted dissolution of the less substituted, more soluble alkyl homologs (including the naphthalenes, which subsequently volatilized).
Bioaccumulation of petroleum compounds
Egg and larval tissue bioaccumulated PAHs from oil-only and oil + dispersant exposures, but relatively few alkanes were detected in tissues. The composition of PAHs bioaccumulated in larvae and eggs was generally similar to the composition of the exposure water (Fig. 5), and concentrations of tPAH in tissue were highly correlated with aqueous tPAH concentrations (0.92 ≤ r2 ≤ 0.98, log-log transform). Eggs accumulated less tPAH from oil and oil + dispersant WAF in 4 d than larvae did in 1 d, resulting in significantly lower tPAH concentrations in tissue at similar aqueous exposure concentrations (Fig. 5). Differences between bioaccumulation in oil-only and oil + dispersant exposures were not significant (p > 0.1), and across WAF treatments, the mean bioconcentration factor in larvae was 249 ± 59 (n = 4) and 76 ± 17 (n = 2) in eggs. Eggs in clean seawater depurated 19 to 32% of the accumulated tPAH after 15.5 h. Alkane concentrations in tissue were poorly correlated with initial aqueous alkane concentrations (0.03 ≤ r2 ≤ 0.63, log-log transform).
Effects on herring embryos
Exposure of herring eggs to oil significantly increased the incidence of yolk sac edema, but other responses did not differ statistically; exposure to UVA did not increase toxicity (Fig. 6). Under control lighting, mean incidence of edema (52%, n = 3) in the highest oil-only treatment (17 μg/L) was significantly greater than in controls (8%), and similar trends of increased edema at higher tPAH concentrations were evident in the UVA treatments. Because exposure to UVA clearly did not increase the incidence of edema (p = 0.877), all light treatments were combined in the final analysis; edema incidence (34%, n = 8) was significantly elevated at 17 μg/L (p = 0.002). Edema effects observed in the oil-only treatments were consistent with observations by Carls et al.  of increased yolk sac edema in 4-d exposures (15 ≤ n ≤ 21), and the 96-h NOEC for yolk sac edema under control lighting was 0.4 μg/g tPAH (Fig. 6). There were other general trends of increasing adverse biological responses of embryos (impaired hatching, survival, spine condition, and swimming) with increasing oil exposure, but values in the highest oil treatments were never significantly different than in controls (data not illustrated). No significant embryo effects were observed in the oil + dispersant treatments, but corresponding tPAH concentrations (maximum 7.8 ± 0.9 μg/L, n = 4) were lower than in oil-only WAF (maximum 17.3 ± 1.2 μg/L, n = 4) and relatively fewer high molecular weight PAHs were present.
In tests designed to discriminate between photosensitization and photomodification, exposure to sunlight caused significant larval mortality and morbidity only when oil was present in larval tissue (Fig. 7). When larvae were preexposed to oil, 4 h of sunlight treatment caused 100% larval mortality within 3 to 5 d in both oil-only and oil + dispersant treatments. There was no significant mortality in any other treatment (2-12% mean mortality and morbidity; Fig. 7).
The interaction of aqueous-phase oil, chemical dispersant (Corexit 9527), and UV resulted in lethal and sublethal effects in Pacific herring larvae and embryos. The ANS was significantly phototoxic to herring larvae, and photoenhanced toxicity only occurred when oil residues were present in tissues rather than through photomodification. Brief exposure to sunlight (∼2.5 h/d for 2 d) was sufficient to significantly increase toxicity to herring larvae and was more potent than laboratory exposures using only UVA. In the sunlight treatment, LC50s, EC50s, and LOECs in larvae were 1.5- to 48-fold lower than in control lighting with minimal UV. In the sunlight + UVA treatment, LC50s, EC50s, and LOECs in larvae were 18- to 450-fold lower than in the control light treatment. Photoenhanced toxicity was not observed in eggs, but only control and UVA exposures were tested, and the bioaccumulated PAHs were lower than in larvae. The toxicity of chemically dispersed oil was similar to oil-only toxicity in control and UVA treatments. In sunlight treatments, oil + dispersant was significantly more toxic than oil-only exposures, and mortality and morbidity occurred more rapidly. The chemical dispersant appeared to accelerate PAH dissolution into the aqueous phase, thus increasing bioavailability. These results are consistent with the hypothesis that weathered ANS is phototoxic and that UV can be a significant and causative factor in the mortality of early life stages of herring exposed to oil and chemically dispersed oil.
Phototoxic compounds in oil
The results of this study are consistent with previous observations that oil low in known phototoxic PAHs can cause significant photoenhanced toxicity [2,3,5]. Based on quantitative structure-activity relationship modeling , the phototoxic compounds in oil are specific three- to five-ring polycyclic aromatic compounds containing either carbon (i.e., PAHs) or heteroatom substitutions within the conjugated rings (i.e., heterocycles). The WAFs prepared from weathered ANS had low concentrations of known phototoxic PAHs, including anthracene, fluoranthene, pyrene, benzo[a]anthracene, and benzo[a]pyrene. Dibenzothiophene and alkyl homologs of the known phototoxic PAHs present in ANS are also expected to be phototoxic based on quantitative structure-activity modeling [19–21], but concentrations of these compounds were also low in WAFs. To date, only a limited number of potential phototoxic compounds that occur in oil have been characterized in single-compound tests. For example, heterocyclic aromatic compounds (e.g., acridine) that are known to be both phototoxic and to occur in oil are not quantified in the typical PAH analyses used in petroleum science . In the absence of definitive information on the specific phototoxic components of oil, we recommend tPAH as the best available surrogate for quantifying oil exposures in phototoxicity studies. A polyaromatic structure is necessary to confer phototoxic properties, and because of similar or greater water solubility and partitioning, concentrations of phototoxic PAHs and heterocycles are likely to covary with tPAH.
Mechanism of oil phototoxicity
Previous studies on the toxicity of individual PAHs to fish and aquatic invertebrates exposed to UV have shown that PAHs primarily act through a photosensitization mechanism, where bioaccumulated PAHs rather than aqueous-phase PAHs are activated by UV . The photoenhanced toxicity of a weathered middle distillate oil also appeared to act through a photosensitization rather than photomodification of aqueous-phase oil . In the current study, a comprehensive set of toxicity tests with herring larvae were performed with oil-only and oil + dispersant exposures to evaluate the mechanism of action of weathered ANS. In these tests, significant toxicity was only observed in larvae that first bioaccumulated oil, then were exposed to sunlight. There was no significant toxicity in the photomodification test where aqueous-phase oil or oil + dispersant was exposed to sunlight for 4 h prior to organism exposures. We therefore conclude that the toxicity observed in the presence of both oil and UV in the main larval experiments was due to bioaccumulated petroleum compounds rather than photomodification of the WAF.
Light spectra causing photoenhanced toxicity
The majority of photoenhanced toxicity studies with PAHs and all previous studies with oil have been performed with simultaneous UVA and UVB exposures; thus, the relative contribution of UV regions of the light spectrum has been unknown [4,8]. We used limited sunlight treatments (∼2.5 h/d for 2 d) to provide environmentally realistic UVA and UVB exposures. We also included light treatments with only UVA because of recent research demonstrating that several PAHs that are components of oil were photoactivated in the absence of UVB . Significantly greater toxicity was observed in sunlight-only exposures than in UVA-only tests, even under marginal environmental lighting conditions (rain, clouds, haze, intermittent solar disk visibility) and low doses of UV (UVA, 2,200-3,600 μW·h/cm2; UVB, 19-41 μW·h/cm2). The sunlight + UVA treatment had nearly identical UVB and approximately 50% more UVA than the sunlight-only treatment and resulted in greater toxicity. Sunlight has a greater representation of longer UV wavelengths that may be important in PAH photoactivation . Larger PAHs such as chrysene strongly absorb UVA wavelengths that may be more abundant in sunlight than in our UVA source and provide a potential explanation for the greater toxicity in sunlight and oil + dispersant treatments through dispersant solubilization of larger PAHs. Alternatively, Huovinen et al.  showed that increasing UVB exposure increased the toxicity of the photoactive PAHs pyrene and anthracene, but had no effect on phenanthrene toxicity, which was not expected to be phototoxic based on its structural conformation . These results are consistent with UVB photoactivation of pyrene and anthracene and support the observations in herring larvae that UVB may be important in determining the magnitude of photoenhanced toxicity. Additional research is needed on the interaction of UVA and UVB because estimation of PAH risks based only on UVA may not predict environmental phototoxicity . The possibility that UVB increases tissue susceptibility to damage by photoactivated PAHs should also be considered in future studies.
Based on tissue residue levels, herring embryos and larvae appeared to have similar sensitivity to oil exposure. For example, embryos developed elevated yolk sac edema at tPAH concentrations of 1.6 μg/g tPAH (oil-only exposure), but no effect on mortality or morbidity was observed. Consistent with this observation, mortality and morbidity were not elevated in our experiments with larvae at 1.2 μg/g tPAH. The apparent lower sensitivity of eggs than larvae based on aqueous concentrations of oil is explained by lower tPAH bioaccumulation by eggs (4-d bioconcentration factor = 76) than by larvae (1-d bioconcentration factor = 249). The UVA treatment did not have appreciable UVB and had no effect on embryo survival or yolk sac edema at the highest tested oil concentrations. Ultraviolet treatments with sunlight were only performed in the larval experiments. We hypothesize that significant photoenhanced toxicity in herring eggs would occur in the presence of elevated oil residues in eggs and sunlight exposure.
Polycyclic aromatic compounds are likely the toxic components of weathered ANS because of their known toxicity  and the observed dose-response relationships in both eggs and herring in the current study. Egg results were consistent with previous observations by Carls et al.  using WAF prepared by passing water through an oiled gravel column instead of the high-energy WAF prepared in the current study. In the column WAF experiments, Carls et al.  observed significantly elevated yolk sac edema at tissue residues at 0.4 μg/g in exposures to more weathered oil. Mortality, morbidity, yolk sac edema, and other sublethal effects increased with the duration of embryo exposure and accumulated tPAH, and the effects were more severe with more weathered oil . Although Neff et al.  suggest that bacterial metabolites may have been responsible for the high toxicity observed by Carls et al. , the consistency of results with the current study indicates that petroleum compounds were directly responsible for toxicity rather than metabolites because the short WAF preparation and exposure period precluded significant microbial growth. Alkanes and the unresolved complex mixture present in WAF generally did not accumulate in embryo and larval tissue. Additionally, a polyaromatic structure is necessary to confer phototoxic properties , and photoenhanced toxicity studies with herring larvae clearly indicated a causal relationship between tPAH exposure and UV treatment.
The differences in PAH composition between oil-only and dispersed-oil WAF preparations were caused by the interacting effects of dispersant and oil volume on the surface-to-volume ratio (S:V)oil of the oil under conditions of constant mixing energy. The (S:V)oil is simply the amount of surface area a unit of mass of oil has. The ratio increases linearly with decreasing radius of oil droplets and is the single most important factor governing the rates at which soluble components such as PAH dissolve into aqueous solution . In the high-oil (3,000 μl) WAF preparation without chemical dispersant, oil droplets created by the mixing process may coalesce relatively easily compared with the other preparations. This could lead to a relatively low (S:V)oil and hence to comparatively slow PAH dissolution rates, resulting in a PAH composition almost like crude oil. Addition of dispersant promotes formation of smaller oil droplets with a much higher (S:V)oil and hence to faster PAH dissolution rates. This can then result in a higher tPAH concentration and to greater losses of the most volatile PAH to the atmosphere as was observed in our experiments.
In the low-oil (300 μl) WAF preparations, the smaller amount of oil intrinsically decreased the probability of droplet coalescence by simple dilution, leading to a lower (S:V)oil and attendant effects on PAH composition. The addition of dispersant had less effect on PAH composition, most likely because of greater dilution by water. The dispersant was added following the oil to simulate conditions more likely in the field. However, only 12 μl of dispersant was added to 32 L of sea-water in the 300-μl oil WAF preparations compared with 10 times that amount in the 3,000-μl oil preparations, so the amount of dispersant added in the low-oil preparations was likely to be correspondingly less effective. The chemical dispersant did increase the proportion of dissolved PAH as indicated by the lower phyt:tPAH and increased the proportions of the less-substituted PAH homologs. The increased concentrations of PAH in true solution caused by dispersant likely led to more rapid bioaccumulation of PAH by larvae, which might account for the more rapid appearance of toxic effects in these larvae when exposed to sunlight.
Phototoxicity of chemically dispersed oil
The photoenhanced toxicity of chemically dispersed oil was assessed using Corexit 9527, which is the principle chemical dispersant stockpiled in Prince William Sound for oil spill response. Corexit 9527 contains nonionic and anionic surfactants and the solvent (ethylene glycol monobutyl ether) . These components are not expected to be phototoxic based on structural considerations . Corexit 9527 was not directly toxic to eggs (0.4 μl/L) or larvae (4 μl/L) at the highest concentrations tested. Other studies have shown low direct toxicity of Corexit 9527 to saltwater fish when no oil is present, with 2- to 4-d LC50s greater than 10,000 μl/L [27,28].
Increasing UV exposure significantly elevated mortality and morbidity, and within a light treatment oil-only and oil + dispersant LC50 and EC50 values were generally similar. In contrast, NOECs and LOECs for oil + dispersants were lower in sunlight and sunlight + UVA treatments than in oil-only exposures, and toxicity occurred sooner in the presence of dispersant and sunlight. Addition of dispersant promoted dissolution of the higher molecular weight PAHs, and some of these may be photochemically excited by UV wavelengths that are relatively more abundant in sunlight than in the UVA lamps we used. In any case, our results suggest the need to match the spectra and intensity of light in photoenhanced toxicity studies, as has been recommended by Barron et al. .
This research supports an increasing body of evidence that oil is toxic to aquatic organisms at extremely low PAH concentrations [10–12,22,29,30] and that exposure to UV can significantly increase the toxicity of oil [2,3,5]. In the current study, WAF was acutely lethal in the absence of UV light when prepared with only 0.01 to 0.1 g/L of applied oil, resulting in part per billion concentrations of tPAH. While other hydrocarbons (alkanes, unresolved complex mixture) also entered treatment water, they generally did not accumulate in herring. Previous studies have demonstrated that embryonic exposures of herring and pink salmon to weathered ANS at 0.4 to 5 μg/L tPAH caused malformations, genetic damage, mortality, decreased size, and impaired swimming in herring and salmon larvae and reduced the marine survival of pink salmon [10,11,29]. The observation of significant photoenhanced toxicity for weathered ANS is in agreement with phototoxicity studies of fresh crude oils and refined fuels, including Prudhoe Bay crude  and a weathered middle distillate oil [3,5]. Along with the current work, these studies support the conclusions of Barron and Ka'aihue  that photoenhanced toxicity of spilled oil may occur in Alaskan waters. Existing laboratory studies and toxicity databases on the toxicity of chemically dispersed oil to aquatic organisms [13,27,28] do not assess photoenhanced toxicity and thus may substantially underestimate the hazard of chemically dispersed oil in the environment .
Herring appear to be extremely vulnerable to photoenhanced toxicity because both the eggs and larvae are translucent (allowing UV penetration) and inhabit the photic zone of the water column. Laboratory studies, including those used to develop the toxicity database for Alaskan species , were performed with minimal exposure to UV. Thus, injury and risk to aquatic organisms from an oil spill may be underestimated if based on standard laboratory bioassays and existing toxicity databases. The results of this study also suggest the potential for photoenhanced toxicity as a factor in herring impacts from the Exxon Valdez oil spill because of part per billion concentrations of tPAH in the water column during the 1989 spill, the high phototoxicity of weathered ANS and limited UV required for photoactivation, and the potential for sufficient UV exposures in the water column in Prince William Sound .
We hank E. Brown, L. Holland, P. Johnson, M. Larsen, A. Moles, and other Auke Bay Laboratory staff for assistance. We thank Steve Diamond of the U.S. Environmental Protection Agency for advice, assistance, and peer review. We thank Dave Gordon, Alaska Department of Fish and Game, for help in collecting herring. This study was supported by the Prince William Sound Regional Citizens' Advisory Council.