The listing of several runs of Pacific salmon as threatened or endangered and federal, state, and local efforts to restore and enhance salmon habitat in the Pacific Northwest make it imperative that the factors associated with these population declines are understood. Currently, 29 out of 52 designated evolutionarily significant units for salmonids on the West Coast of the United States are listed under the Endangered Species Act, including the Chinook salmon (Oncorhynchus tshawytscha), steelhead trout (Oncorhynchus mykiss), and bull trout (Salvelinus confluentus) in Puget Sound. Puget Sound coho salmon (Oncorhynchus kisutch) are currently listed as a species of concern (http://www.nwr.noaa.gov/ESA-Salmon-Listings/Salmon-Populations/Coho/Index.cfm; last accessed October 24, 2012).
Recent federal, state, and local efforts to document pesticide concentrations in surface waters have generated a significant amount of information, which for the first time allows for the study of the hazards known pesticide mixtures pose to aquatic resources 1–3. In most cases levels reported are low (<1.0 µg/L); but the presence of these chemicals has generated concerns, particularly in terms of their potential effects on salmonids 4. For example, Voss et al. 1 detected 23 pesticides at 12 sites within 10 urban/suburban watersheds following rainfall events and related the detections to retail sales of formulated products for garden and yard care by homeowners. In a more intensive study of many of the same creeks, Frans 3 detected the following 37 pesticides: 20 herbicides, 9 insecticides, 2 fungicides, and 6 transformation products. Although pesticide inputs to surface waters are predominantly considered to be an agricultural concern, concentrations of many of the pesticides detected in urban streams in western Washington are frequently greater than their agricultural (eastern Washington) counterparts, and the mixtures are different. For example, maximum concentrations of the 12 pesticides detected in urban streams in western Washington with a frequency greater than 30% 3 were, with few exceptions, one to two orders of magnitude greater than those for the same pesticides associated with drainages in agricultural areas in eastern Washington; and the frequency of detection of the latter was less than 5% 2. Five herbicides and one insecticide frequently detected within the agricultural drainages (>30%) were not among those most frequently detected in the urban streams.
Previously, we tested the hypothesis that prespawn mortality of coho salmon was a result of adult exposure to pesticide mixtures (“cocktails”) reported in urban streams and that reproductive success of exposed adults that did spawn would be adversely affected. Longevity, ripening in females, and brain acetylcholinesterase (AChE) were not significantly affected by continuous exposure to the maximum reported concentrations of the pesticides. Fertilization, hatching success, and growth of fry were also not affected when green adults were exposed to these concentrations for 96 h. Although exposure to the pesticide cocktail did not appear to be associated with prespawn mortality in coho salmon 5, concerns remained regarding effects on early life stages. The early life stages of fish have been found to be more sensitive to chemical exposure than later life stages 6, and salmonid fry have been demonstrated to be highly sensitive to chemical pollutants. For example, Elson 7 determined that aerial applications of dichlorodiphenyltrichloroethane (DDT) in the 1960s to control spruce budworm (Choristoneura fumiferana) caused a reduction of 40% in Atlantic (Salmo salar) salmon parr, while reductions in young of the year were nearly 100%. Lower and Moore 8 studied the effects of insecticides used in sheep dip treatments in the United Kingdom on Atlantic salmon. Cypermethrin, a synthetic pyrethroid, and diazinon, an AChE inhibitor, were the two primary active ingredients. Environmentally relevant concentrations of the insecticides resulted in a reduction of fertilization rates, a disruption of emergence timing, and reduced fitness and survival of the fry. The authors also noted that these effects were greater when each insecticide was tested separately and noted an antagonistic effect when applied in a binary combination. In samples collected before and after a spill of azinphos-methyl, also an AChE-inhibiting insecticide, on Prince Edward Island, Canada, young of the year rainbow (O. mykiss) and brook (Salvelinus fontinalis) trout showed a higher degree of impact than older age classes 9. In a study with rainbow trout, eggs, sac fry, and fry exposed to a number of pesticides, including cadmium chloride, manganese ethylenebisdithiocarbamate, pentachlorophenol, parathion-ethyl (an AChE inhibitor), 1,2,4,5-tetrachlorobenzene, and dieldrin, the early fry stage was found to be the most sensitive life stage 10.
Although additive toxicity has been assumed among chemicals with similar modes of action11, the nonadditive (synergistic) interactions among some AChE inhibitors have heightened concerns over the presence of these pesticides in surface waters and their effects on salmonids 12. Furthermore, the potential for nonadditive interactions among other chemicals within surface waters is relatively unknown 11. Few studies have examined effects associated with simultaneous exposure to multiple pesticides with different modes of action. For example, the addition of triazine herbicides to mixtures of some organophosphate pesticides also resulted in synergistic toxicity 11. Triazine herbicides are also frequently detected in urban streams, likely associated with the use of lawn treatments 1–3.
In the present study, we exposed gametes and early life stages of coho salmon to a pesticide cocktail representative of the pesticides most frequently reported in urban streams in western Washington 3. The cocktail consisted of eight herbicides (including two triazines), two insecticides (both AChE inhibitors), a wood preservative, and a pesticide-breakdown product. End points included fertilization, hatching success, growth, and survival to five weeks post–swim-up.
MATERIALS AND METHODS
Our three-year study consisted of a series of experiments conducted in three facilities at the University of Washington, Seattle, Washington, USA. Because of differences and limitations within their physical plants, methods for housing and exposing specific life stages to the pesticide cocktail varied (Table 1). All animal procedures were approved by the Institutional Animal Care and Use Committee at the University of Washington under Protocol 2185-35. Details in addition to those provided below can be found in King 5. Mention of trade names, products, or suppliers does not constitute endorsement by the federal or state agencies with which the authors are affiliated.
Table 1. Description of facilities at the University of Washington housing each of the life stages in the three-year study
Year and life stage
Tank size (L)
H = hatchery; FTR = Fisheries Teaching Research Building-room; CoS = City of Seattle dechlorinated; LWA = Lake Washington; PR = partial recirculation; F = flow-through; C = controlled; NC = not controlled; NA = not applicable.
A and C
Pair and group
Coho salmon eggs and milt were obtained from the Washington Department of Fish and Wildlife's hatchery in Issaquah, Washington, USA, in November of each year. Each egg lot (in an individual 3.8-L Ziploc freezer bag) and milt from each male (in an individual whirlpack) was topped off with oxygen, placed on cardboard over wet ice in an insulated cooler, and then transported to the University of Washington for processing.
Chemical cocktail and exposure scenario
Early life stages were continuously exposed to pulses of the cocktail simulating those in urban streams in either fall or winter when coho salmon spawn. The cocktail consisted of those pesticides detected in at least 30% of the urban creeks in western Washington sampled by Frans 3. The herbicides 2,4-dichlorophenoxyacetic acid (2,4-D), dicamba, dichlobenil, 2-methyl-4-chlorophenoxyacetic acid (MCPA), methylchlorophenoxypropionic acid (MCPP), prometon, simazine, and triclopyr; the insecticides carbaryl and diazinon; the fungicide pentachlorophenol; and the breakdown product 4-nitrophenol met this criterion.
Nominal concentrations were either the maximum reported during fall or winter (August–December, in most cases associated with storm flows) or the average of detections during base flows in the same time period. Average concentrations of the pesticides during base flows were approximately 20% of their maximum values. With the exception of pentachorophenol and 4-nitrophenol (technical grade), formulated products containing only one active ingredient were selected based on availability. Liquid formulations were preferred, but some active ingredients were available only as granulars or wettable powders 5.
In an effort to simulate pulses of the cocktail associated with storm water flows, fertilized eggs and fry were exposed to the maximum concentrations (high) for a minimum of 48 h and 20% of the maximum concentrations (low) for a minimum of 72 h each week using a dosing pump. The pump was turned to high for 96 h and to low for 72 h. Actual time of exposures depended on the turnover rate within each test system. For example, the turnover within the partially recirculating incubation system (1.1 L/min) was estimated to be 24 h; comparable times for flow-through incubation systems (∼40 min) or grow-out tanks (∼180 min) were substantially less.
Each year, 1 L of concentrated stock (3 × 106) was mixed in November using distilled water and maintained at 4°C. This concentrate was used to restock the delivery tanks each week by adding the necessary amount of concentrate to distilled water. The target concentration within the delivery tanks was adjusted based on the stroke rate and delivery volume of the dosing pump to inject the appropriate amount of chemical into the water flow.
The dosing systems used among the years had some differences. In 2006/2007 for the incubation system, we used a Pulsatron Series D (LF02SA-VTC1, 6 gal [22.7 L]/d, 150 psi; Idex) pump and a 200-L circular Nalgene delivery tank with a lid that was covered in black plastic to prevent photodegradation. The head in the tank was maintained by a pump (model MD6L; Iwaki America) that circulated the contents of the tank, drawing from the bottom. Water flow (clean) to the grow-out tanks was 38.4 L/min (3.8 L/tank). In 2007/2008, we used a Pulsatron Series D (LF02SA-VTC1, 6 gal [22.7 L]/d, 150 psi or LF04SA-PVC1, 24 gal [90.9 L]/d, 100 psi; Idex) dosing pump. The delivery tanks for each of the dosing systems were 115- or 200-L Nalgene tanks with a lid that was covered in black plastic. The head in the delivery tanks was maintained by a pump (model MD6L; Iwaki America) that circulated the contents of the tank, drawing from the bottom. Water flow to the grow-out tanks receiving the cocktail was 42.2 L/min (3.8 L/tank). In both years, the tank was placed in a water heater pan (7.5 cm deep) with circulating facility water in an effort to cool the concentrate. Freshwater flow was 1.1 L/min within the incubation racks with an internal circulation of 11.5 L/min and, in 2007/2008 in the flow-through system, 11.5 L/min of freshwater.
In 2008/2009, we used a different dosing pump (model EZB10D1-PC, 0.6 gal [2.3 L]/d; Iwaki America) for incubation and grow-out systems and 15- or 20-L Nalgene delivery bottles placed within a small refrigerator (Kenmore model 94683; Sears). In an effort to maintain the delivery stock at ≤4°C, the stock was circulated through a coil of polypropylene tubing within a cooler with ice by a small pump (model 75420-00; SHURflo). Water flow (clean) in the incubation system for the fertilization dose–response tests was similar to that described above.
In each of the systems, the dosing and recirculation pumps were outside of the delivery tanks. Chemical delivery was monitored daily by checking the actual pump delivery (milliliters per minute), recording the drop in the delivery tank volume (millimeters), and calculating the amount delivered per minute.
For the fertilization tests in 2006/2007, treated water was collected from the incubation system 24 h after fresh stock had been added to the stock tank with the dosing pump set at high. In 2007/2008, freshly mixed stock within the delivery tank for the partial recirculation system was mixed with clean incubation water to generate the treated water (maximum concentration). In subsequent tests (2008/2009), stock was freshly mixed specifically for the fertilization tests and serial dilutions with clean incubation water were used to generate the geometrically arranged concentrations for each of the two dose–response tests (A and B). For test A we used five concentrations—0, 0.1, 1, 10, and 100 times the maximum concentrations in the pesticide cocktail; for test B we used 0, 0.3, 1, 3, and 10 times the maximum concentrations.
Water samples were collected from the incubation racks and grow-out tanks to compare nominal and measured concentrations (Table 2). Samples were stored at 4°C and within 24 h of collection shipped overnight on ice packs to the analytical laboratory. In 2006/2007, samples were sent to Edge Analytical Laboratories. Analyses for the chlorophenoxy herbicides were conducted using U.S. Environmental Protection Agency (U.S. EPA) method 515.1 (Edge Analytical Standard Operating Procedure [SOP] “Analysis of Chlorinated Herbicides by Capillary Gas Chromatography Using Diazomethane Methylation”). Collection bottles (0.5 L) contained 40 to 50 mg of sodium sulfite. Analyses for the other pesticides were conducted using U.S. EPA method 525.2 (SOP “Analysis of Semivolatile Organic Compounds in Drinking Water by Solid Phase Extraction and Capillary Column Gas Chromatography/Mass Spectrometry [GC/MS]”). Five milliliters of 6 N HCI was added to the 1-L samples postcollection except for those to be analyzed for diazinon because previous studies had shown that the insecticide was not stable under acidic conditions, with significant losses occurring within 7 d 5.
Table 2. Nominal and measured concentrations (micrograms per liter active ingredient, means with range) of pesticides in the chemical cocktail in the incubation systems and grow-out tanks and for the fertilization dose–response test A: Concentrations of the amine herbicides are expressed as salts
In 2007 through 2009, samples (2 L) were sent to the Pacific Agricultural Laboratory. For carbaryl, the Pacific Agricultural Laboratory used neutral extraction (3535A), followed by high-performance liquid chromatography–mass spectrometry (HPLC-MS) in selected ion monitoring (SIM) mode (modified U.S. EPA method 8321B). For diazinon, the 3535A extraction was used, followed by gas chromatography (GC)–flame photometric detector analysis (U.S. EPA method 8141B GC-FPD). The 3535A extraction process was also used for dichlobenil, followed by a GC-electron capture detector analysis (U.S. EPA method 8081BA GC-ECD). After the 3535A extraction, the triazine herbicides were analyzed using a GC-mass selective detector in SIM mode (modified U.S. EPA method 8207C). A base catalyzed hydrolysis followed by an acidic extraction was used for the chlorinated herbicides, followed by analysis with GC-mass selective detector in SIM mode (U.S. EPA method 8151B). The extraction for both MCPA and MCPP also involved a base catalyzed hydrolysis, followed by an acidic extraction and then HPLC-MS analysis in SIM mode (U.S. EPA method 8321A [HPLC-MS]).
Extractions were performed within the recommended hold times, but recommendations varied among years: 14 d in 2006/2007, 7 d in 2007/2008 and 2008/2009. Active ingredients within the amine herbicides (2,4-D, MCPA, MCPP, and triclopyr) were expressed as salts. Analytical results were not adjusted for percentage of recovery. In most cases, recovery exceeded 75% (2006/2007: mean = 100.3%, range 65–135%, 3% < 75%; 2007/2008: mean = 78.1%, range 52–101%, 32% < 75%; 2008/2009: mean = 83.4%, range 50–137%, 24% < 75%). Low recoveries were frequently associated with the amine herbicides, dichlobenil, pentachlorophenol, and 4-nitrophenol, which may account for some of the low measured concentrations for these pesticides.
Studies have demonstrated genotype-dependent impacts in animals exposed to toxicants 13. Therefore, each egg lot was halved to provide its own control, and each male was randomly assigned to fertilize both halves of a lot. With the exception of the fertilization dose–response tests in 2008/2009, all eggs were handled the same way. Eggs and ovarian fluid from each female were split volumetrically between the two 7.7-L plastic buckets.
The amount of milt and method of delivery varied among years. In 2006/2007, an eyedropper was used to add 2 to 2.5 ml of milt to both control and treatment buckets. The trigger water (500 ml) for the control and treatment lots was collected from the corresponding incubation system. The dosing pump for the treatment incubation system was set on high during this collection. Controls were processed first. The water and eggs were allowed to stand for at least 2 min, after which each bucket was drained and a 100-mg/L PVP iodine (Western Chemical) solution was added (4× volume of eggs for fertilization-only assessments, 2× for all others) and allowed to sit for 15 min. The buckets were drained again and the eggs placed into a tray in the appropriate incubation system rack.
In 2007/2008, 1-cc syringes were used to standardize the amount of milt used. Using a separate syringe for each whirlpack of milt, 1 ml of milt was added to each bucket in succession at the same time that the trigger water was added. This change was made because it was determined that fertilization can occur in just ovarian fluid 14, potentially masking a treatment effect. While the controls were processed, the syringes were left in the open whirlpacks. Once the control lots were in the incubation racks, the same syringes and milt were used to fertilize the treatment lots. Egg exposure to the milt and use of an iodine solution were the same as in 2006/2007.
In 2008/2009, we were concerned that the apparent treatment effect observed the previous year was due to the methodology we used to fertilize the eggs, specifically potential water contamination of the syringes used to deliver the milt and the delay in processing the treatment lots. Two different syringes filled with milt (1 ml) at the same time were used to fertilize each pair of egg lots (control then treatment). The appropriate trigger water was applied at the same time as the milt to each bucket. Groups of five lots were processed by pair prior to placement in the corresponding incubation system as in previous years.
To further address methodological concerns and variation in measured versus nominal concentrations of some of the constituents of the cocktail, we conducted two fertilization dose–response tests (A and B). In these tests, five egg lots and ovarian fluid were split volumetrically five ways (∼300 eggs per batch) into a 1-L food-grade plastic cup. Five 1-cc syringes were used to administer 0.25 ml of milt from one randomly assigned male per egg lot, one per cup, and then discarded. Trigger water (500 ml) was either clean, dechlorinated City of Seattle water (controls) or the same water spiked with the appropriate amount of cocktail stock (treatment) and was added at the same time as the milt. Controls were processed first, followed by the treatment eggs, moving up in concentration. Each egg lot was processed sequentially in an environmental chamber (12°C), and eggs were placed in individual trays within a clean water flow-through incubation system in which each rack was designated for a specific treatment group.
In the University of Washington hatchery, the water used in each of the two incubation systems was dechlorinated City of Seattle water. The incoming water was charcoal-filtered (model 300; Polymetrics) that fed dual (treated and control) partially recirculating systems, each equipped with a chilling unit (model F3AD-0151-CFV-001; Copeland) to maintain the temperature between 10 and 11°C and a UV filter (model 02065; Emperor Aquatics). In 2007/2008 and 2008/2009, we used an additional incubation system in the university's Fisheries and Teaching Research Building. The water was charcoal-filtered and chilled (8–10°C). The Fisheries and Teaching Research incubation system was flow-through, with each rack receiving either treated or clean water. Water quality (pH, temperature, conductivity [microseconds], and dissolved oxygen [milligrams per liter]) was monitored daily (Table 3). For the fertilization dose–response tests, water quality was assessed at the beginning and end of the 24-h test (test A) and daily until eye-up (∼21 d, test B) (Table 3).
Table 3. Water quality (means, ranges in parentheses) during each test each year
Year and test
Dissolved oxygen (mg/L)
C = control; T = treatment.
Incubation (partial recirculation)
Incubation (partial recirculation)
Fry grow-out A, C
Fry grow-out B
Incubation (partial recirculation)
Dose response, test A
Dose response, test B
A sample of 50 eggs was collected from each lot and cleared with Stockard solution. The two- to four-cell stage was targeted, with the time of sampling dependent on the number of accumulated temperature units needed to hit the targeted stage (Incubwin Version 2.1 software; Canadian Department of Fisheries and Oceans, http://www.pac.dfompo.gc.ca/science/aquaculture/sirp/incubwin-eng.htm, last accessed October 24, 2012). In 2006/2007, the two-cell stage was targeted; however, fertilization was generally indeterminable. The four-cell stage was targeted in subsequent years. Once the eggs cleared, they were individually examined under a light microscope to determine the presence and stage of cell cleavage.
Incubation trays were first checked for dead eggs at 24 h (2006/2007 and 2008/2009) and 96 h (2007/2008) postfertilization (initial egg pick-off). Due to egg sensitivity, with the exception of the initial egg pick-off, trays were not opened again until they were checked for eyes beginning on approximately day 21. Lots were then checked every 3 to 4 d until hatching, and dead eggs were counted and removed. In order to calculate hatching success, the total number of eggs was calculated as the number of live (hatched) plus dead at sorting plus the dead eggs removed prior to sorting. The percentage of eggs that hatched was calculated by dividing the number live by the total number of eggs and then multiplying by 100.
Fifty eggs were also sampled from each tray for the fertilization dose–response tests. For test A (0–100 times), all remaining eggs were discarded the morning following sampling. In test B (0–10 times), the eggs were allowed to develop through eye-up as a second fertilization assessment. At eye-up, live and dead eggs were counted, and all treatment groups except the controls and one-times group were discarded. The latter were used in subsequent experiments 5.
In 2006/2007, lots were transferred at swim-up to a randomly assigned 125-L grow-out tank (model DM35; Aquatic Ecosystems) where all fish were counted and numbers standardized. For the first 7 d, fry were kept near the surface by a semirigid screen placed 10 to 15 cm below the water's surface to facilitate feeding behavior and culling of deformed fry. The water in the grow-out tanks was unfiltered Portage Bay water, temperature controlled (10–11°C), and with the exception of the first week, gently aerated (model LPH45; JEHM).
In 2007/2008 an objective was to release as many smolts as possible to assess return rates 5. Therefore, numbers at swim-up were not standardized, and all lots were counted and transferred. In order to maximize the fish available and due to logistical constraints, some treated and control lots were placed in clean water within 266-L high-density polyethylene tanks (Polytank) in the main hatchery for a 35-d growth assessment. Using a modified standpipe, these fish were kept in 25-cm-deep water during the first week. A 35-d growth assessment was not conducted in 2008/2009.
In both years, starting at time 0, 30 fish were randomly selected and killed with MS-222 and their weights and lengths recorded. Groups of 30 fish were sampled 7 d apart for 35 d, for a total of six sample periods. After the first 7 d, the water depth was increased to normal tank depth. Fish were fed 1.5% (2006/2007) and 3% (2007/2008) of body weight. Food was adjusted daily to account for mortality and weekly to account for growth and sampling. Water quality was monitored in each tank at least every other day (Table 3).
All percentage data were arcsine-transformed prior to analyses 15. Since the lots were split into control and treated halves, paired t tests were conducted to test for differences between control and chemical-exposed groups in fertilization and hatching success, survival, deformities, and condition factor. For the dose–response tests, the fertilization rates (samples of 50 eggs) for both tests and at the eye-up stage in the second dose–response test were analyzed using a one-way analysis of variance. For all analyses, differences were considered significant if p ≤ 0.05. All analyses were performed using SPSS Version 19 for Mac and Excel (Microsoft Office 2008 for Mac) software.
Hatching success and survival from hatching to swim-up were similar between (1) eggs fertilized within the pesticide cocktail or clean water and then incubated in clean water (Fig. 1) and (2) eggs fertilized in clean water and then incubated within the pesticide cocktail or clean water (Fig. 2). Survival to 7 d post–swim-up and from 8 to 35 d post–swim-up were also unaffected in embryos exposed postfertilization (Fig. 2). No significant differences (p = 0.948, 0.519) were observed in percentage of deformities between the treatment (average = 0.49%, 0.28%) and control (average = 0.50%, 0.22%) groups in the experiments described above (1 and 2, respectively). Growth as measured by condition factor (weight/length3) did not differ significantly between fry exposed to the cocktail postfertilization through swim-up and clean water controls at any of the six sample points (Fig. 3). Fertilization success at the two-cell stage in both experiments was indeterminable.
Hatching success of pesticide-exposed embryos was reduced 19 to 25% (p < 0.003) in each of the three fertilization assessments (Fig. 4). Survival from hatching to swim-up was unaffected in the three groups (data not shown, see King 5). In each of these assessments, there were significant (p < 0.035) differences in fertilization between the treated and control lots as measured by cleavage at the two- to four-cell stage. Fertilization (data not shown, see King 5) and hatching success (Fig. 4) were essentially equal, suggesting that the difference in hatching success was due to an effect on fertilization.
After swim-up, two of the three groups of egg lots used for the fertilization assessment (groups 2 and 3) were used to evaluate effects on growth. Lots within group 2 were separated into two groups: the five largest split lots were transferred to grow-out tanks where the exposure of treated fish to the pesticide cocktail continued (grow-out A) and the remainder were transferred to grow-out tanks and exposed to clean water (grow-out B). Swim-up fry from the six split lots within group 3 were transferred to grow-out tanks where the exposure of treated fish to the pesticide cocktail continued (grow-out C). No significant differences were observed in survival from swim-up to 7 d post–swim-up or 8 to 35 d post–swim-up in any of the three experiments (data not shown, see King 5). Nor were there significant differences (p = 0.760, 0.512, 0.383) in percentage of deformities between the treatment (average = 0.15%, 0.32%, 0.31%) and control (average = 0.17%, 0.37%, 0.18%) groups in the three fertilization experiments, respectively. Growth (condition factor) of fry exposed to the pesticide cocktail during the 35-day grow-out was equal to or significantly greater (p ≤ 0.05) than that of controls at each of the time points within the three experiments (Fig. 5).
In response to the apparent treatment effect on fertilization in 2007/2008, three experiments were conducted. In the first, effects on fertilization were examined in 15 split lots, half fertilized in the pesticide cocktail and half in clean water, with five paired (treated and control) split lots processed at a time. No significant differences were observed in hatching success or survival to swim-up between lots exposed to the pesticide cocktail (treatment) or clean water (control) at fertilization and continuing through swim-up (Fig. 1, 2008/09). No significant difference (p = 0.556) was observed in percentage of deformities between the treatment (average = 0.09%) and control (average = 0.16%) groups for release. Results of the two fertilization dose–response tests indicated that there were no significant effects when assessed by cleavage either at the two- to four-cell stage or at eye-up (Fig. 6). Assessments based on cell cleavage appeared to underestimate fertilization (Fig. 6B).
Across the three-year study, there were no consistent treatment effects. In 2006/2007, hatching success was not affected when gametes were exposed to the maximum concentrations of the pesticide cocktail and subsequently exposed to clean water (Fig. 1, 2006/07); survival and growth of fry were also not affected (Figs. 2, –4). Fertilization and hatching success were reduced 19 to 25% in 2007/2008, but survival and growth of fry continuously exposed to pulses of the pesticide cocktail did not differ from controls (Fig. 4). In 2008/2009, we could not reproduce the effect on fertilization and hatching success observed the previous year or in dose–response tests containing concentrations up to 100 times the maximum concentrations used in previous tests (Fig. 6). Hatching success and survival and growth of fry were also not negatively affected when eggs or fry were continuously exposed to pulses of the pesticides. The proportion of deformed fry did not vary statistically between the treatment and control groups and was comparable to that in the University of Washington hatchery (∼0.5%; J. Wittouck, University of Washington, Seattle, WA, USA, personal communication). Results suggest that, for the end points examined, continuous exposure to pulses of the maximum concentrations of the pesticides most frequently detected in urban streams in western Washington poses little direct hazard to coho salmon reproduction.
We do not believe that differences between years in the genetic stock of the adults from which we collected gametes influenced our results as all adults were from the same hatchery stock. The adults were reared in the same environment and expected to follow similar routes as outmigrating smolts and returning adults and, therefore, were likely to be exposed to similar stressors including contaminants. Our results complement those of a previous study in which we did not observe effects on longevity and ripening in adult coho salmon when continuously exposed to the maximum concentrations of the same chemical cocktail or on their ability to produce viable offspring as assessed by hatching success and survival and growth of fry 5.
Measured versus nominal concentrations
The maximum nominal concentrations and the continuous pulsed exposure we employed likely represent a worst-case scenario for salmon gametes and fry within urban streams in western Washington. We do not know the actual duration of pulses of pesticides in these urban streams; however, studies have indicated that storm runoff events in these areas are likely to be considerably less in duration than those that were tested as part of the present study. Urban streams in the Puget Sound lowlands have been described as “flashy” 16 and are characterized by a “faster rise and recession of streamflow, higher peak rates, and increased storm flow” 17. Burges and colleagues 18 indicate that the peak high flows in Puget Sound urban streams are less than half of the peak (∼48 h) in our study, and McCarthy and colleagues 19 indicate that peak storm flows may last only a matter of hours.
Existing data were based on single-grab samples coinciding with storm or base flows 3. The duration of the maximum concentrations (96 h > high pulse ≥ 48 h) was largely based on that associated with standardized toxicity tests (e.g., 96-h median lethal concentration [LC50]) 20. A number of recent studies of the effects of pesticides on salmonids have utilized a comparable duration of exposure 12, 21–24 or less 22, 25. The maximum nominal concentrations were on average nearly an order of magnitude (mean = 8.4 times, 0.4–28.7 times) greater than geometric means for all detections within urban streams for the pesticides in the cocktail during the fall and winter 5. Baseline concentrations were based on the average of maximum values during base flows and were 20% of maximal values during storm flows and approximately three times the geometric means.
In some cases, we were unable to maintain the desired duration of the maximum concentrations. Differences between measured and nominal concentrations, particularly for 2,4-D, MCPA, and 4-nitrophenol, among years appeared to be primarily associated with the conditions and concentrations within the chemical delivery tanks. The apparent loss of these three chemicals by the time the water samples were collected 3 to 4 d after new chemical cocktail concentrate was added in 2006 to 2008 was likely because the contents of the delivery tank were not adequately chilled and the entire contents of the tank were not replaced each time fresh concentrate was added. Elevated temperatures (∼15–16°C) within the delivery tank, aeration from the recirculation pump, and the presence of “old” cocktail all likely facilitated bacterial growth and chemical degradation. Analysis of the concentrate within the 1 L of stock in spring 2007, 3.5 months after mixing, indicated that all of the constituents, including 2,4-D and MCPA, were present (2,4-D 93.7% of nominal, MCPA 160.3%, 4-nitrophenol 22.8%). Therefore, with the exception of 4-nitrophenol, degradation in the stock maintained at ≤4°C during the study was likely minimal.
In contrast, concentrations of the three chemicals in the delivery tanks for the flow-through systems (incubation and five-week grow-out) approached nominal, with the exception of 4-nitrophenol in the flow-through incubation system, suggesting that the loss of the chemicals was also a function of target concentration. Nominal concentrations of the three chemicals were 9.9 to 46.6 times greater in the two flow-through dosing systems, reflecting differences in freshwater inputs. The greater concentrations within the flow-through systems may have inhibited bacterial growth in the delivery tanks, irrespective of tank temperature.
All three chemicals were present within the samples collected from the partial recirculation system in 2008/2009, and concentrations were consistently closer to nominal yet not as close as the flow-through systems. In this year, the delivery tanks were placed in a small refrigerator, the contents were circulated through a cooler filled with ice, and the stock was completely replaced each week. Although temperatures in the delivery tank rarely exceeded 4°C (range 2–6°C, 14% of daily values >4°C) in the partial recirculation system, heat created by the recirculating pumps necessary to maintain the head in the delivery tanks made it difficult to lower the temperature of the pesticide concentrate further.
With the exception of 2,4-D, MCPA, and 4-nitrophenol, however, measured concentrations were consistently greater than the geometric means calculated from measurements in urban streams in western Washington during the fall and winter. With few exceptions (2,4-D, MCPA, and 4-nitrophenol primarily during incubation), the minimum concentrations to which coho salmon eggs and fry were exposed, based on the analytical chemistry data, were 1.4 to 26 times greater than the geometric means, with averages among the three years between 6.8 and 8.9 times. Concerns associated with failure to achieve target concentrations were addressed through replication of tests across years and, in the case of fertilization, additionally through the dose–response tests. Furthermore, although degradation of some of the constituents of the cocktail occurred by the time water samples were collected 3 to 4 d after fresh stock was added to the delivery tanks, eggs in the incubation racks with partial recirculation likely still experienced maximum concentrations but for a time period shorter than that desired (24–48 vs 72–96 h).
Amine herbicides: Acid equivalents versus salts
We assumed that the concentrations reported for the amine herbicides in Frans 3 were expressed as salts, the actual active ingredient, and not as acid equivalents. This distinction was not provided in the publication, but it appears that the values given in the publication were expressed as acid equivalents (S. Magoon, Washington Department of Ecology, Manchester, WA, USA, personal communication). Therefore, the nominal values for the amine herbicides used in the present study may have been lower than those reported in the U.S. Geological Survey study: 2,4-D (–17%), dicamba (–32%), MCPA (–18%), MCPP (–15%), and triclopyr (–28%). However, nominal values were still significantly greater than the geometric means (2,4-D six times, dicamba nine times, MCPA three times, MCPP three times, and triclopyr eight times 5).
In 2006/2007, while there were no significant differences in any of the end points tested, hatching success in the fertilization assessment was low (∼45%) in both treatment and control lots (Fig. 1) in comparison to those fertilized in clean water, chemically exposed during incubation, and then placed in clean water during grow-out (∼75–80%, Fig. 2). The latter is more consistent with hatching rates within the University of Washington hatchery. However, the hatching success associated with gametes obtained from the Issaquah hatchery by University of Washington hatchery staff to supplement the University of Washington run at the same time that we collected our gametes was also low and comparable to what we observed in our fertilization assessment.
In 2007/2008, we observed a 19 to 25% reduction in fertilization success across three replicate experiments conducted over three different days, a result we did not observe in 2006/2007. We were concerned that the apparent treatment effect was due to water contamination of the syringes used to deliver the milt and the delay in processing the treatment lots (control lots were processed first to reduce the potential for cross-contamination). The following year, we took steps to correct this (see Methods for 2008/2009) and did not observe a treatment effect. Also, we did not detect a reduction in fertilization success in our dose–response tests. We also repeated the methods we used in 2007and 2008 with five egg lots (unique pairings) but could not reproduce the effect 5. The reasons for the differences among years are not known.
To our knowledge, no data exist on actual sperm to egg ratios for coho salmon in a natural (wild) setting. In a study with African catfish (Clarias gariepinus) and mercury, the optimal ratio of sperm to egg was determined to be 15,000 to 1 26. These authors also noted that a surplus of milt can inhibit fertilization as well as overwhelm any potential treatment effect 26. The milt to egg ratio that should be used for an experiment like ours was unknown. In our tests, we used approximately 2.5 ml undiluted milt to 500 ml of trigger water for a large number of eggs (average 1,700), with the amount of milt reduced to 1 ml in 2007/2008 and in 2008/2009. In our fertilization dose–response tests, we used 0.25 ml of undiluted milt to approximately 300 eggs. Stekoll et al. 6 used a similar ratio of milt to solution. For groups of 20 to 300 eggs, they used 0.2 ml of undiluted milt to an initial 100 ml of trigger water, followed quickly by another 100 ml (M. Stekoll, University of Alaska, Juneau, AK, USA, personal communication). Stekoll et al. 6 selected 0.2 ml not as an estimate of the amount in a natural ejaculum but because previous experience indicated that this amount was adequate to achieve complete or high fertilization rates in controls, would not input much excess organic material in the solution, and was a way to standardize the procedure (W. Smoker, University of Alaska, Juneau, AK, USA, personal communication). It is likely that tens of milliliters of milt may be involved in one fertilization event of a mature male salmon in the wild (W. Smoker, personal communication). However, dilution and movement of milt within flowing natural waters may be significant, making it difficult to determine normal milt to egg ratios. In a recent study with sockeye salmon (O. nerka), Macfarlane and colleagues 27 determined that 20 µl of milt in 500 ml of river water was sufficient to ensure 100% fertilization of approximately 150 eggs. If this was also true for coho salmon, the smallest volumes of milt we used were more than adequate to ensure fertilization. The extent to which milt to egg ratios may have influenced our results is not known.
Water and chemical delivery
In the flow-through incubation system in 2007/2008, two daily consecutive spikes in the temperature of the dechlorinated freshwater entering the racks were detected (group 3). Temperature spikes were short-lived (∼60–75 min), with temperatures climbing an average of 2.5°C during this time period and then dropping back to previous temperatures over approximately 10 min. An inadvertent bypass of the dechlorination system when the problem was being diagnosed resulted in a spike of chlorinated water (≤1 mg/L chlorine) that lasted approximately 30 min in the incubation racks. No effect was apparent as the hatching success of controls was similar among the three groups (Fig. 4). In addition, the dosing pump on the incubation rack (partial recirculation) in the hatchery malfunctioned (group 2). The pump reading indicated the correct pump rate (19%, high), but the pump was actually in prime mode (100%). The malfunction occurred after all the egg lots were almost completely hatched and lasted approximately 21.5 h. We estimate that the exposure to the pesticide cocktail was approximately 5.26 times the maximum desired for approximately 12 h. No increase was observed in egg or hatchling mortality in comparison to the controls.
Comparisons to other studies of direct effects
Comparable studies exposing different life stages of salmonids continuously to a pesticide cocktail representative of that reported in the surface waters they inhabit are lacking. Nominal concentrations for the weekly “high” pulses were the maximum reported within urban streams in western Washington and likely represented a worst-case scenario. Although maximum concentrations were orders of magnitude lower than levels known to be lethal 5, concerns over the possibility of interactions within the mixture resulting in increased toxicity remained. Those studies that have examined the effects of pesticides singly or in combination on Pacific salmon have focused on organophosphate and carbamate insecticides because of their mode of action (AChE inhibition), the importance of acetylcholine in nerve function, and/or reported synergism among different AChE inhibitors in vivo 12. However, effect levels in Laetz et al. 12 for the AChE inhibitors in our cocktail (carbaryl and diazinon) singly or in combination exceed those within urban streams in western Washington and, in most cases, salmon-bearing surface waters along the West Coast (C. Grue and K. King, unpublished data).
In a continuation of the present study, we sampled coho salmon smolts that had been continuously exposed to pulses of the pesticide cocktail since fertilization and did not detect any inhibition of brain AChE activity 5. Laetz et al. 12 observed slight synergism (<10% increase in AChE inhibition) in juvenile coho salmon exposed to carbaryl + diazinon for 96 h, each at a concentration of 7.3 µg/L. Nominal concentrations in our cocktail were 0.06 and 0.58 µg/L. Baldwin et al. 21 predicted that growth in juvenile Chinook would be reduced if brain AChE inhibition were reduced 10% continuously or at least by 50% once for 96 h beginning at 30 days post–swim-up. Concentrations necessary to achieve these levels of AChE inhibition in coho salmon with carbaryl and diazinon would be 29 µg/L (effective concentration 10% [EC10]) and 145 µg/L (median effective concentration [EC50]) when exposed singly, or 7.3 µg/L (EC10) or 29 µg/L (EC50) when combined 12. Concentrations of the two insecticides in U.S. Geological Survey monitoring studies of salmon-bearing watersheds in eastern Washington, Oregon, and northern California rarely exceeded 1 µg/L. For example, on the Columbia Plateau in eastern Washington, carbaryl and diazinon rarely exceeded 0.1 µg/L 28. Only in the San Joaquin-Tulare basin in California did diazinon concentrations frequently exceed 1 µg/L 29.
Levels of exposure to AChE inhibitors associated with changes in behavior also exceed those within our cocktail. For example, swimming stamina was reduced 23 to 44% in three species of salmonids suffering approximately 50% AChE inhibition 30. Sandahl et al. 23 reported a statistically significant reduction in spontaneous swimming rate (–27%) in juvenile coho salmon with brain AChE inhibition of 23%, whereas feeding, swimming rate, time to first feeding strike, and total feeding strikes were not impaired until enzyme inhibition reached 51%. Similarly, data presented by Beauvais et al. (22, Fig. 3A) suggest a similar threshold for swimming speed in juvenile rainbow trout (O. mykiss) exposed to carbaryl. More recently, Tierney et al. 31 reported impaired swimming performance after a threshold of approximately 32 or 50% AChE inhibition was reached, depending on the type of performance test utilized.
Although we did not observe changes in behavior or reduced growth or survival in coho salmon fry exposed to the pesticide cocktail, more subtle effects may have occurred. Olfaction in salmonids appears to be more sensitive to pesticide exposure than the overt effects discussed above. Tierney et al. 24 exposed juvenile rainbow trout to a mixture of the 10 most frequently detected pesticides within the Nicomekl River in British Columbia in 2004 (1.0 times, actual = 1.2 times) for 96 h and quantified their olfactory response to L-serine (avoidant) using electro-olfactograms. Responses were also quantified following exposure to 0.1 times (actual = 0.2 times) and 10 times (actual = 16 times) concentrations with and without background odorant concentrations and a subsequent single acute (5 min) 20 times exposure. Electro-olfactograms in response to L-serine did not vary significantly among pesticide-exposed and unexposed fish. However, the magnitude of olfactory responses in the presence of background odorant concentrations was reduced by the 1 and 10 times exposures. Previous exposure to the pesticide mixture at the 0.1 and 1.0 times concentrations did not affect the magnitude of response to the subsequent 20 times exposure, but exposure to the 10 times concentration caused a smaller decrease compared to the other two treatments. The cocktail included six organophosphate insecticides, three herbicides (two triazines), and one organochlorine insecticide. Total actual concentrations of the organophosphates were 0.18, 0.93, and 13.09 µg/L across the three treatment groups. Both the 1 and 10 times total concentrations of AChE inhibitors exceeded that in our pesticide cocktail (0.64 µg/L).
Whether or not the effects on olfaction observed at the high concentration by Tierney and colleagues 24 were associated with brain AChE inhibition is not known because the authors did not measure activity of the enzyme. However, diazinon is known to be primarily responsible for the significant synergism among AChE inhibitors observed by Laetz et al. 12 through its inhibition of carboxylesterases, enzymes essential for detoxification processes (K. King et al., unpublished data). Given the magnitude of the concentrations of the organophosphates dimethoate (6.6 µg/L), diazinon (1.8 µg/L), malathion (0.93 µg/L), and parathion (3.5 µg/L), synergism at the high concentration likely was resulting in significant brain AChE inhibition. Assuming that the olfactory effects observed by Tierney et al. 24 were associated with the concentration of AChE inhibitors, it is unlikely that our fry were similarly affected. A recent study found that the avoidance response of juvenile rainbow trout to skin extract from conspecifics was not impaired following a 96-h exposure to the maximum concentration of our pesticide cocktail (V. Babchanik, University of Washington, Seattle, WA, USA, unpublished data).
With respect to the potential for olfactory effects, the concentrations of 2,4-D and triclopyr in our cocktail can be compared to those used in tests with juvenile rainbow trout (C. Curran, unpublished data, University of Washington, Seattle, WA, USA). Trout were exposed to the maximum label rates used for the control of aquatic weeds (2,4-D = 4.0 mg/L, triclopyr = 2.50 mg/L) for 96 h. The avoidance response of these trout to a conspecific skin extract was similar to that of controls. The levels tested were more than three orders of magnitude greater than those within our cocktail.
Changes in pesticide inputs to urban streams
Our pesticide cocktail was based on data for urban streams in western Washington collected between 1998 and 2003 3. Comparison of these data with the results of monitoring in 1997 1, 2003 2, and 2007 32 suggests that the pesticides most frequently detected have changed and concentrations have declined. For example, data specific to Thornton Creek were reported for 2003 and 2007 by Anderson et al. 2 and Anderson and Dugger 32. In 2003 pesticides with a frequency of detection greater than 10% were pentachlorophenol (89%, maximum concentration = 0.08 µg/L), dichlobenil (72%, 0.34), triclopyr (56%, 0.19), diazinon (46%, 0.21), MCPP (35%, 0.15), 2,4-D (33%, 0.16), and prometon (22%, 0.03). Comparable data for 2007 were dichlobenil (63%, 0.07), prometon (15%, 0.03), carbaryl (13%, 0.05), 2,4-D (11%, 0.22), and MCPP (11%, 0.08). Anderson et al. 2 noted that maximum concentrations of dichlobenil, diazinon, and prometon detected in 2003 were lower than those reported for Thornton Creek between 1995 and 1998 33. In 1997, 2,4-D, diazinon, dichlobenil, MCPP, prometon, and pentachlorophenol were detected in 100% of the water samples collected during storm flows 1 from the same creeks monitored by Frans 3 between 1998 and 2003. In the study by Frans 3, frequencies of detection for these same pesticides in spring and autumn storm flows were 2,4-D 87.5 and 85%, diazinon 87.5 and 65%, dichlobenil 83.3 and 60%, MCPP 91.7 and 85%, prometon 87.5 and 90%, and pentachlorophenol 87.5 and 90%, respectively. The most significant change was the absence of diazinon in the most recent samples from Thornton Creek. Diazinon was no longer registered for homeowner use as of 2003 3.
Other differences are likely in response to the change in the registration for diazinon. Retail sales of carbaryl have increased as homeowners seek an alternative to diazinon 3. Johnson and colleagues 34 report that pyrethroid insecticides have become the principal replacement products in commercial, agricultural, and urban settings; and a number of studies have documented concentrations in sediments in urban waterways that are toxic to aquatic invertebrates (for recent review, see Kuivila et al. 35). With additional restrictions on the registrations of organophosphate and carbamate insecticides being considered by the U.S. EPA, the increases in use of synthetic pyrethroids may pose a greater direct and indirect hazard to salmonids than AChE inhibitors 36. Synthetic pyrethroids are up to 1,000 times more toxic to fish than to mammals and birds, compared with organophosphates and carbamates 37, and can be more persistent in the environment 38.
Relevance to other aquatic systems and salmon species
We compared concentrations in our pesticide cocktail to those reported in surface waters associated with agricultural areas in eastern Washington 2 and in urban streams elsewhere in the United States 39. Maximum concentrations of only three pesticides detected in three creeks within the agricultural Yakima basin exceeded those in our cocktail (2,4-D 1.9 µg/L, dichlobenil 0.5 µg/L, carbaryl 10 µg/L). Only concentrations of 2,4-D in one of the creeks exceeded those in our cocktail, and the frequencies of detection of carbaryl and dichlobenil in the three creeks were low (ND–4.4%). Prometon and triclopyr were the only pesticides in our cocktail that were not detected in the agricultural creeks. The composition of our cocktail differed from that reported for urban streams by Hoffman et al. 39 for other areas of the United States. Only 2,4-D, prometon, and simazine were detected in >10% of the water samples analyzed. Carbaryl and diazinon were the only insecticides detected in >10% of the samples. Maximum concentrations for the pesticides in common were in most cases greater than those in our cocktail (simazine 8.2 µg/L, prometon 2.93, MCPA 1.3, 2,4-D acid 1.2, carbaryl 3.2, diazinon 1.4); exceptions were dichlobenil (0.3) and triclopyr (0.34). Due to concerns associated with the potential for nonadditive toxicity 12, we previously compared concentrations of carbaryl and diazinon in our cocktail with those reported in surface waters inhabited by salmonids on the West Coast (see above); a more thorough comparison was beyond the scope of the present study.
Differences may exist between the sensitivity of coho salmon gametes and fry and those of other species of salmon when exposed to our pesticide cocktail. To partially address this, we repeated our dose–response fertilization test with Chinook gametes and a pulsed exposure from fertilization through five weeks post–swim-up. Similar to our studies with coho salmon, we did not observe any significant effects on fertilization and hatching success or on survival and growth of fry when compared to negative controls (K. King, unpublished data).
Results of the present study based on the end points examined suggest that direct exposure to the maximum concentrations of the most frequently detected pesticides in urban streams in western Washington does not pose a significant hazard to coho salmon reproduction. We did not assess the potential for these pesticides to affect reproductive success indirectly by altering the aquatic environment (e.g., reductions in the growth or survival of emergent fry and juvenile fish as a result of pesticide-induced changes in the availability of food 4). To gauge the potential for indirect effects, we compared the maximum reported concentrations of the pesticides in our cocktail to effects thresholds for aquatic life utilized by Wilson et al. 40 for a subset of the streams in western Washington that formed the basis for the present study 3. Only concentrations of diazinon and simazine exceeded thresholds. Maximum values for diazinon exceeded the threshold (0.09 µg/L) in three of the 11 creeks that formed the basis for our cocktail with concentrations of 0.14, 0.18, and 0.59 µg/L 3. Values for simazine exceeded the threshold (0.12 µg/L) in two creeks (0.14, 0.42 µg/L).
Anderson et al. 2 compared concentrations of pesticides detected in Thornton Creek, one of the urban creeks included in the two previous U.S. Geological Survey studies 1, 3, to existing water quality criteria following more intensive sampling in 2003. Only diazinon exceeded recommended water quality criteria (0.1 µg/L); 4% of samples had concentrations of 0.13 and 0.21 µg/L. In a subsequent study 32, diazinon was not detected in the creek. Stark (J. Stark, Washington State University, Puyallup, WA, USA, unpublished data) exposed Daphnia pulex to concentrations of five pesticides reported by Frans 3 as formulated products singly or as a mixture—2,4-D (1.0 µg/L), carbaryl (0.04 µg/L), chlorpyrifos (0.02 µg/L), diazinon (0.3 µg/L), and malathion (0.1 µg/L)—and compared population growth rates to that of controls. The concentration of diazinon exceeded water quality criteria. The population growth rate was not affected by the concentrations tested singly or as a mixture. However, exposure to 2.5 times the concentrations as a mixture significantly reduced reproduction. Additional studies are needed to evaluate the indirect effects of current levels of pesticides in urban streams in western Washington.
Fertilization, hatching success, survival, and growth of fry were not significantly affected by weekly pulses of the maximum concentrations of the pesticides most frequently detected in urban streams in western Washington. The pesticide exposure employed likely represents a worst-case scenario for early life stages of coho salmon through five weeks post–swim-up in these streams. The results complement those of a previous study in which exposure to the maximum reported concentrations did not affect longevity or ripening in adult coho salmon, subsequent fertilization and hatching success of their eggs, or survival and growth of resulting fry 5. Based on the end points assessed in the present study, toxicity associated with direct exposure of adults and early life stages to pesticides in urban streams in western Washington likely is not a major factor governing the recovery of Pacific salmon populations in Puget Sound. The extent to which pesticide-exposed fry would suffer reduced survival during smoltification, during outmigration, or in the ocean needs to be addressed.
Funding was provided by the U.S. Geological Survey Cooperative Research Units, the School of Aquatic and Fishery Sciences at the University of Washington, the Washington Cooperative Fish and Wildlife Research Unit, the Washington State Commission on Pesticide Registration, and the Washington Friends of Farms and Forests. The unit is financially supported by the U.S. Geological Survey; University of Washington; Washington State University; the Washington Department of Ecology, Fish, and Wildlife; and the Washington Department of Natural Resources. Special thanks to J. Wittouck, N. Hurtado, and D. Rose (University of Washington Hatchery) for technical assistance and to V. Blackhurst, S. Damm, W. Madden, G. Marston, M. Horne-Brine, M. Sternberg, and M. Tamayo for assistance with fish capture, spawning, or data collection. C. O'Toole (Edge Analytical Laboratories) and S. Thun (Pacific Agricultural Laboratory) provided guidance on water sample collection for chemical analyses and interpretation of results from analytical chemistry. L. Conquest, G. Young, and two anonymous reviewers kindly provided comments on earlier versions of the manuscript.