An urban stormwater runoff mortality syndrome in juvenile coho salmon

Aquatic Toxicology 214 (2019) 105231

Contents lists available at ScienceDirect

Aquatic Toxicology journal homepage: www.elsevier.com/locate/aqtox

An urban stormwater runoff mortality syndrome in juvenile coho salmon a

b

c

d

a

Michelle I. Chow , Jessica I. Lundin , Chelsea J. Mitchell , Jay W. Davis , Graham Young , ⁎ Nathaniel L. Scholze, Jenifer K. McIntyrec,

T

a

University of Washington, School of Aquatic and Fisheries Sciences, 1122 Boat St., Seattle, WA 98105, USA National Research Council Research Associateship Program, Under contract to Northwest Fisheries Science Center, National Marine Fisheries Service, NOAA, 2725 Montlake Blvd. E., Seattle, WA 98112, USA c Washington State University, Puyallup Research and Extension Center, 2606 W. Pioneer Ave., Puyallup, WA 98371, USA d U.S. Fish and Wildlife Service, Washington Fish and Wildlife Office, 510 Desmond Dr. S.E., Lacey, WA 98503, USA e Environmental and Fisheries Science Division, Northwest Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 2725 Montlake Blvd. E., Seattle, WA 98112, USA b

ARTICLE INFO

ABSTRACT

Keywords: Non-point source pollution Water quality Stormwater Road runoff Urban stream syndrome

Untreated urban runoff poses significant water quality threats to aquatic organisms. In northwestern North America, ongoing development in coastal watersheds is increasing the transport of toxic chemical contaminants to river and stream networks that provide spawning and rearing habitats for several species of Pacific salmon. Adult coho (Oncorhynchus kisutch) are particularly vulnerable to a stormwater-driven mortality syndrome. The phenomenon may prematurely kill more than half of the coho that return each fall to spawn in catchments with a high degree of imperviousness. Here we evaluate the coho mortality syndrome at the juvenile life stage. Freshwater-stage juveniles were exposed to stormwater collected from a high traffic volume urban arterial roadway. Symptoms characteristic of the mortality syndrome were evaluated using digital image analysis, and discrete stages of abnormal behavior were characterized as the syndrome progressed. At a subset of these stages, blood was analyzed for ion homeostasis, hematocrit, pH, glucose, and lactate. Several of these blood chemistry parameters were significantly dysregulated in symptomatic juvenile coho. Affected fish did not recover when transferred to clean water, suggesting a single runoff event to stream habitats could be lethal if resident coho become overtly symptomatic. Among coho life stages, our findings indicate the urban runoff mortality syndrome is not unique to adult spawners. Therefore, the consequences for wild coho populations in developing watersheds are likely to be greater than previously anticipated.

1. Introduction As the global human population grows, the concomitant expansion of cities and extended metropolitan areas is profoundly changing the structure and function of freshwater ecosystems. This phenomenon is known as the urban stream syndrome (Paul and Meyer, 2001; Meyer et al., 2005; Walsh et al., 2005) and is primarily driven by changes in land cover and land use. For example, increasing imperviousness leads to well-known modifications of physical habitat processes, including more extreme (flashier) instream flows, streambed scour, bank erosion, and increased sedimentation. In northwestern North America, the negative influences of urbanization on physical watershed characteristics have been studied for decades (e.g., Booth et al., 2002). Urban stormwater runoff is a major driver for these changes, and past studies have primarily focused on macroinvertebrate abundance and diversity as in

⁎

situ indicators of biological integrity in streams known to be hydrologically altered by urbanization (e.g., Morley and Karr, 2002; Alberti et al., 2007). Although less studied, changes in stream flows and habitat structure have also been associated with declines of Pacific salmon (Oncorhynchus sp.), including Chinook (O. tshawytscha; Moscrip and Montgomery, 1997; Regetz, 2003) and coho (O. kisutch; Pess et al., 2002; Bilby and Mollot, 2008). Stormwater is also the primary driver for the loading of non-point source pollution to urban waterways (Scholz and McIntyre, 2016). Runoff originating from roadways with dense traffic volumes has long been known to be toxic to aquatic species (Marsalek et al., 1999). Motor vehicles release a broad diversity of chemicals into the environment as a consequence of tire and brake pad wear, leaking of crankcase oil and transmission fluid, and tailpipe exhaust. These contaminants, most of which have not been previously studied, accumulate on highways,

Corresponding author. E-mail address: [email protected] (J.K. McIntyre).

https://doi.org/10.1016/j.aquatox.2019.105231 Received 8 March 2019; Received in revised form 12 May 2019; Accepted 19 June 2019 Available online 20 June 2019 0166-445X/ © 2019 Published by Elsevier B.V.

Aquatic Toxicology 214 (2019) 105231

M.I. Chow, et al.

streets, and other impervious surfaces and are then mobilized into stormwater runoff during rain events. Absent definitive identification, their toxicity to aquatic organisms remains largely unknown. This perspective is changing, however, with the application of novel analytical methods based on high resolution time-of-flight mass spectrometry. For example, Du et al. (2017) and Peter et al. (2018) recently used these methods to characterize hundreds of distinct chemical features in highway runoff and Seattle-area urban streams. Over the past two decades, coho salmon have served as a particularly important sentinel species for stormwater-driven impacts on surface water quality. This is exemplified by the recurring and premature deaths of adult coho when they return each autumn to spawn in urban watersheds (Scholz et al., 2011). Prior to dying, affected fish show a progression of behavioral symptoms including circular surface swimming and gaping, a loss of equilibrium, and immobility. This phenomenon has been variously termed “coho pre-spawn mortality (Scholz et al., 2011)”, the “coho urban runoff mortality syndrome (McIntyre et al., 2018)”, or the “coho acute spawner mortality syndrome (Peter et al., 2018)”. Spawner deaths may amount to as much as 90% of a fall run, and the syndrome has been consistently observed in field surveys across many years (Scholz et al., 2011) and many urban watersheds (Feist et al., 2017). There are multiple lines of evidence for toxic runoff as the cause of the syndrome. First, forensic analyses have largely ruled out conventional water quality parameters, spawner condition, disease, and other alternate hypotheses (Scholz et al., 2011; Spromberg et al., 2016). Second, the mortality syndrome, including symptoms, can be reproduced in experiments wherein adult coho are exposed to road runoff under controlled conditions (Spromberg et al., 2016; McIntyre et al., 2018). Third, adult coho deaths can be prevented by pre-filtering the same roadway runoff through experimental soil bioretention columns to remove pollutants (Spromberg et al., 2016). Finally, landscape modeling has shown that the severity of coho spawner mortality scales with the extent of imperviousness within a watershed (Feist et al., 2011) and, more specifically, the density of motor vehicle traffic near spawning habitats (Feist et al., 2017). The precise cause of death is not yet known, but blood chemistry in symptomatic fish indicates acidosis and a disruption of ion homeostasis (McIntyre et al., 2018). Given that coho salmon are semelparous (reproducing once before dying), the loss of large numbers of fish at the critical spawner stage is likely to pose a significant conservation threat to wild coho. The Puget Sound coho population segment is a species of concern under the U.S. Endangered Species Act, and coho are listed as threatened in the Lower Columbia River. These wild population segments rely on watersheds that are under increasing pressure from urban and suburban development, including the large metropolitan areas of Seattle in the Puget Sound basin and Portland in the Lower Columbia River basin. As might be expected, initial modeling indicates that future urbanization, increased toxic runoff, and increased spawner mortality has the potential to drive rapid local coho extinctions (Spromberg and Scholz, 2011). Retrospectively, the mortality syndrome may partially explain relatively low abundances of coho in developed watersheds, as determined from past freshwater habitat assessments (Pess et al., 2002; Bilby and Mollot, 2008). The population-scale threats of toxic runoff to coho will be greater if other life stages are impacted. Symptomatic juveniles have been observed in urban creeks during spawner surveys (Video S1 and S2). Also, runoff collected from a high traffic arterial roadway was acutely lethal and, similar to adults, this mortality was prevented with bioinfiltration pre-treatment (McIntyre et al., 2015). In terms of discovering the underlying cause of the syndrome, adult coho pose several important logistical challenges – e.g., unpredictable return rates and a narrow yearly window of availability. If the syndrome affects juveniles and adults alike, juveniles might offer a much more tractable experimental framework. Studies on juveniles could also shed light on why coho appear to be much more vulnerable relative to other species of Pacific salmon that also spawn in urban habitats, including chum (O. keta; McIntyre

et al., 2018). However, to date, the mortality syndrome has not been rigorously evaluated in juvenile coho. In the present study, we conducted a detailed assessment of behavior and physiology in juvenile coho exposed to roadway runoff relative to controls held in clean water. Video analyses were used to define discrete categories of symptomology that corresponded very closely to symptoms previously documented in adults. These stages (normal behavior, surface swimming, and loss of equilibrium) then served as phenotypic anchors for a range of measured blood chemistry and hematological parameters. We also assessed whether visibly affected fish could recover upon a transfer to clean water. Our results are discussed in the context of managing imperiled coho populations, as well as ongoing efforts to identify the causal chemical agents that cause the mortality syndrome. 2. Methods Several previous studies have shown that untreated roadway runoff from a densely travelled urban arterial is nearly always acutely lethal to juvenile (McIntyre et al., 2015) and adult (McIntyre et al., 2018; Spromberg et al., 2016) coho salmon. Seasonal differences in antecedent rainfall patterns (and, by extension, measured chemical constituents) have little influence on the mortality syndrome (Spromberg et al., 2016), and coho are similarly vulnerable at different seasonal timepoints throughout a given year (McIntyre et al., 2015). Therefore, the goal of the present study was not to determine the extent to which juvenile coho were affected by stormwater exposures per se (we anticipated close to 100%), but rather to 1) carefully characterize the progression and nature of the symptoms, and 2) determine if affected fish would recover if returned to clean water. 2.1. Urban stormwater collection Urban stormwater runoff was collected from downspouts draining an elevated urban principal arterial; specifically, the westbound onramp to State Route 520 from Montlake Boulevard in Seattle, WA. The ramp receives approximately 15,000 average daily vehicle trips (ADTs) and is paved with Portland cement concrete, a conventional urban impervious surface material. During rain events, the highway runoff was pre-filtered through a fiberglass screen to remove coarse debris and then collected in a 900-L stainless steel tote (Custom Metalcraft Inc., Springfield, MO). The stormwater was subsequently transported to Washington State University Puyallup Research and Extension Center (WSU-Puyallup; Puyallup, WA) and used for juvenile salmon exposures within 24 h postcollection. Samples for analytical chemistry were also collected within 24 h, prior to experimental exposures. 2.2. Juvenile coho salmon Juvenile coho salmon (aged 1+ year) were obtained from the Northwest Fisheries Science Center hatchery facility (Seattle, WA) and maintained at WSU-Puyallup in circular fiberglass tanks supplied with dechlorinated municipal water (fish lab water) at 13 °C. Fish were held on a 12:12 light:dark regime and fed daily with commercial pellet food (Bio-Oregon, Warrenton, OR). At the time of the behavioral trials, fish averaged (mean ± SD) 53 ± 29 g in weight and 163 ± 31 mm in length. Fish exposed for analyses of blood parameters, approximately two months later, averaged (mean ± SD) 208 ± 98 g in weight and 255 ± 44 mm in length. Experiments were conducted in accordance with Experimental Protocol #04860-002, as approved by Washington State University’s Institutional Animal Care and Use Committee (IACUC). 2

Aquatic Toxicology 214 (2019) 105231

M.I. Chow, et al.

Table 1 Behavioral sequence for the urban runoff-elicited mortality syndrome in juvenile coho salmon. Behavior Characteristics a

Stage 1 Stage Stage Stage Stage Stage a b

a

2 3a 4a 5b 6b

Asymptomatic. Discrete surfacing events shorter than 0.25 s. Fish swimming within 1 cm of the surface of the exposure tank were scored. Events that only included the tail were not counted. Short episodes of surface swimming ranging from 0.25-2 s. Sustained continuous surface swimming. All durations were greater than 2 s, included linear and circular swimming. Loss of equilibrium including side swimming, inability to stay parallel to the surface, and spiraling. Loss of buoyancy, as indicated by settlement to the bottom of the tank. Moribund (near death), exhibiting changes in ventilation rates (i.e., gilling), along with spasms and gaping. Moribund coho were also unresponsive to touch.

visible at the water surface. visible from the tank bottom.

2.3. Synthetic water exposures

levels of unexposed (control) coho were monitored by video over a 24-h period to determine an appropriate acclimation period, prior to the onset of behavioral experiments. Normal activity levels resumed 2 h after transfer into the observation tank. This 2 h acclimation interval was sufficient even when the same fish were subsequently transferred to an adjacent tank. Therefore, on each experimental day, six pairs of juvenile coho were transferred from the rearing tanks to each of the six acrylic holding tanks containing static control water, allowed to acclimate for 2 h, after which each pair was moved by net to one of the six recirculating exposure tanks containing control or runoff water (Figure S2). Fish were temporarily exposed to air (< 2 s) during the brief transfer to the adjacent exposure tank. In all cases, mirrors, cameras and blinds were positioned to block a direct line of sight between juvenile coho and human observers. On each experimental day, for each of the six recirculating exposure tanks, the pair of fish was divided into two parallel experiments (Figure S2). In the first experiment, one fish of the pair remained in each tank to monitor the entire progression of behavioral symptoms during exposure to runoff or control water. The goal of the second, parallel, experiment was to determine whether symptomatic coho in the runoff exposure would recover upon transfer to clean water. Accordingly, the other fish in the stormwater-exposed pair was moved into a static clean water tank at a point of intermediate distress (surface swimming; see Table 1). At the same time, one fish from a control exposure pair was similarly moved from the recirculating control exposure tank back into a clean water static tank. Runoff-exposed fish transferred to clean water were observed until they reached a moribund state, which we defined as immobile on the bottom of the tank. To control for the additional transfer stress, the remaining fish in each exposure tank was also lifted out of the water in a net for a few seconds. All exposures were terminated after eight hours. Departures from normal swimming behaviors were quantified from video collected using three overhead cameras (Dahua Technology Lite Series 1.3 M P; Irvine, CA) connected to SecuritySpy video capture software (Ben Software, London, United Kingdom). Each camera was positioned to monitor four tanks, allowing all tanks to be simultaneously observed. Video recordings were analyzed using Timestamped Field Notes (Neukadye, Golden, CO). Due to the near opacity of the stormwater, the cameras only captured behaviors near the surface of each tank in runoff-exposed fish. Accordingly, control behaviors were also only quantified when fish were near the surface of the water. Total time spent at the surface in recirculating control or runoff water was quantified for each individual throughout the continuous exposure. Discrete surfacing events were defined as 0.125 s. Individuals remaining in recirculating continuous control and runoff exposures were monitored until moribund. For fish transferred at the onset of intermediate distress (surface swimming), behavior was only documented prior to transferring. Fish that lost buoyancy and eventually became moribund were visible through the clear acrylic bottom of the tank and were documented by periodic visual observations. Overall, for the categorization of symptomology, a set of three

Relative to the fish husbandry (laboratory) water used for control exposures, urban runoff typically has lower pH, alkalinity, and hardness. Each parameter could potentially affect the osmolality of exposed coho salmon, independent of toxic chemical contaminants. To eliminate the possibility that differences in conventional water quality alone induced changes in the blood chemistry of exposed fish, we constituted synthetic water representative of runoff chemistry values previously measured for storm events in 2012–2013 (Spromberg et al., 2016; Table S1). Synthetic waters were created to approximate the conventional chemistry of the fish lab water, as well as low and intermediate ion concentrations previously measured in runoff from the same source as used in this study. Synthetic waters were constituted by filtering deionized water through a carbon filter (Brita LP, Oakland, California) and a Synergy water purification system (EMD Millipore, Darmstadt, Germany) and then adding salts (CaCl2·2H2O, Na·HCO3, MgCl2·6H2O, and KCl) to produce three synthetic water types; synthetic fish lab water, synthetic “low ion” water, and synthetic “intermediate ion” water (Table S1). The pH of the synthetic waters was adjusted using HCl and NaOH as appropriate to match the pH of fish lab water or stormwater. Five juvenile coho salmon were exposed to fish lab water or one of the synthetic waters in 35-L glass aquaria supplied with air stones. Temperature was maintained at 13 °C by placing aquaria in flowthrough circular fiberglass tanks. Beginning at 4.5 h after the onset of each exposure, fish were sampled for arterial blood. Collected blood was analyzed on the iSTAT point-of-care blood analyzer using the methods outlined in Section 2.5.2. 2.4. Characterization of behavioral symptoms in response to stormwater Experiments to assess behavioral symptomology among runoff-exposed juvenile coho were conducted in twelve 56.6-L acrylic tanks (61 cm × 46 cm × 37 cm) with shared walls in an isolated wet laboratory (Figure S1). Mirrors, blinds, and cameras were used to prevent fish from seeing human experimenters. Whereas the sides of the tanks were opaque, the bottoms were clear plexiglass. The tanks were elevated on steel stands to allow observations from below when it was necessary to confirm the presence of a moribund fish. Urban runoff was collected from two separate rainfall events on April 10 and 23, 2017. For each storm, exposures were conducted on two consecutive days. Runoff to be used on the second day was stored overnight in the stainless steel transport tote at ambient room temperature (18 °C). Six of the acrylic tanks were filled with clean fish lab water (control) and maintained under static conditions with air stones to provide supplementary oxygenation. The remaining six tanks, in two rows of three, were plumbed to supply recirculating urban runoff or control water with flow rates set at 6 L/min to ensure sufficient exchange. Dissolved oxygen, temperature, and pH were monitored throughout each of the recirculating exposures (Table S2). In pilot trials to assess handling stress (data not shown), the activity 3

Aquatic Toxicology 214 (2019) 105231

M.I. Chow, et al.

stormwater-exposed and three control fish were evaluated on each of two days, for an n = 6 for each treatment per storm. The behavioral progression of the runoff mortality syndrome (Table 1) was used to define distinct stages of symptomology, ranging from Stage 1 (asymptomatic) to Stage 6 (moribund). In subsequent exposures to collect blood, fish were sampled at Stage 1, 3, and 4 (see below).

following previously described methods for iSTAT usage in fishes (Gallagher et al., 2010). Our results, therefore, represent differences between control and runoff-exposed fish and should not be directly compared to published values for salmonid blood chemistry using traditional measurement methods. When the iSTAT analysis was complete, a 50 μL aliquot of whole blood was collected and stored at 4 °C for quantification of red blood cell indices. The remaining whole blood of each sample was centrifuged for 5 min at 12,000 x g to separate plasma and red blood cell pellets, followed by storage at −80 °C. Red blood cell indices were determined for fish sampled at Stage 4 (loss of equilibrium) within 24 h of blood collection. Briefly, 10 μL of blood was diluted 200-fold with saline (2.4 mM CaCl2, 4.3 mM MgCl2, 10 mM KCl, 140 mM NaCl) and Trypan Blue (0.4%). The blood solution was then loaded in 10 μL volumes into each side of a Double Neubauer Hemocytometer (Hausser Scientific, Horsham, PA). Viable and non-viable cells were quantified across five 0.04-mm2 squares on each side of the hemocytometer. Cells that were permeable to Trypan Blue, as indicated by blue cytoplasm, were characterized as non-viable (Strober, 2001). If duplicate readings of a sample were significantly different (ttest, α = 0.05), the sample was reanalyzed. Total red blood cell counts were adjusted for dilution and reported as 1012 cells/L of whole blood. Mean corpuscular volume of the blood cells (MCV in femtoliters) was calculated from HCT (%) and RBC count as MCV = HCT*10/RBC. Plasma lactate and total protein were measured on a VetTest Chemistry Analyzer (IDEXX Laboratories, Westbrook, ME). After thawing on ice, a 50 μL aliquot of plasma was warmed to room temperature and diluted 1:1 with saline (154 mM NaCl, pH 7.0) for plasma analysis; the remaining plasma was refrozen. Within a few days, plasma was rethawed and quantified from duplicate 10 μL samples on a vapor pressure osmometer (VAPRO 5520, Wescor, Logan, Utah). Fish were measured for length and weight immediately after blood sampling. Condition factor was calculated using Fulton's condition factor (K = [weight (g)/length (cm)3]x100; Anderson and Neumann, 1996). Exposure water samples (fish lab water and all three storms) were collected prior to the onset of juvenile coho exposure trials and analyzed for conventional water quality parameters by a commercial laboratory (Analytical Resources Inc., Tukwila, WA) using US EPA approved methods. Targeted parameters included ammonia, total suspended solids, and dissolved organic carbon. In addition, total and dissolved concentrations of arsenic, cadmium, copper, lead, nickel, and zinc were measured by the same laboratory using inductively coupled plasma mass spectrometry (ICP-MS).

2.5. Characterization of physiological responses to stormwater 2.5.1. Runoff exposures Juvenile coho salmon were exposed to urban runoff collected from a storm event on June 8, 2017 (n = 45 fish), or fish lab water (n = 45 fish) at WSU-Puyallup. Exposures were conducted in the twelve acrylic tanks described in Section 2.4 after they were re-plumbed to supply recirculating stormwater or clean fish lab water to each of six tanks per treatment. Flow rates were set to 6 L/min. Dissolved oxygen, temperature, pH, and conductivity were monitored throughout the exposure. Runoff-exposed fish were sampled at discrete behavioral categories of progressing symptomology (Stages), as determined from the behavioral assessments above (see also Table 1). The three discrete behavioral categories were Stage 1 (asymptomatic), Stage 3 (surface swimming), and Stage 4 (loss of equilibrium). The elapsed time of exposure prior to sampling at each Stage (mean ± SD) was 62.2 ± 18.3 min (Stage 1), 125.1 ± 49.5 min (Stage 3), and 199.6 ± 63.2 min (Stage 4). Six tanks, each containing five fish, were used for each Stage (n = 15 fish per treatment). The start of each exposure was staggered to allow adequate sampling time. Behaviors were closely monitored using the overhead camera system described previously. When a runoff-exposed fish reached the targeted behavioral Stage, it was removed from the tank and euthanized by blunt force to the head. A corresponding control fish from the paired clean water tank was then sampled. 2.5.2. Hematology and tissue analyses Immediately after euthanizing an individual fish, up to 1 mL of arterial blood was collected from the dorsal aorta as described in McIntyre et al. (2018) using a 25-gauge, 1-in needle connected to a lithium-heparinized syringe (7 IU/mL ion-balanced lyophilized; ABG Line Draw, Smiths Medical, Dublin, OH). After brief mixing by horizontally rolling the syringe, a drop of blood was discarded and 95 μL of blood was loaded into a CG8+ iSTAT System Test Cartridge (Abaxis, Union City, CA). The cartridge was then inserted into an iSTAT point-ofcare blood analyzer (Abaxis, Union City, CA) for analysis. Blood metrics analyzed included sodium (Na+), ionized calcium (iCa2+), partial pressure of carbon dioxide (pCO2), pH, glucose (GLU), and hematocrit (HCT). Other parameter outputs from the iSTAT, including bicarbonate (HCO3−), total carbon dioxide (TCO2), oxygen saturation (sO2), extracellular base excess, and hemoglobin, were indirectly calculated and therefore not included in subsequent analyses. Partial pressure of oxygen (pO2) and potassium (K+) measured on the iSTAT were also excluded due to unreliability of values identified in a previous study on adult coho and chum salmon (McIntyre et al., 2018). Although the values of blood parameters measured on the iSTAT have been shown to deviate from the values measured using conventional controlled laboratory procedures (Harter et al., 2014; Harrenstien et al., 2005), the utility of the iSTAT has been demonstrated in many field studies when conventional equipment may not be available (Cooke et al., 2008; Meland et al., 2010a; Regan et al., 2016; Forrestal et al., 2017; McIntyre et al., 2018). For example, while the measured value of hematocrit on the iSTAT is underestimated when compared to conventional techniques (McIntyre et al., 2018; Harrenstien et al., 2005), the deviations in hematocrit values are reliably underestimated by 8–10% and therefore reported as measured. Corrections were applied for temperature-sensitive parameters (pH and pCO2) using the average temperature of the exposure water (Table S2)

2.6. Statistical analyses Non-parametric Kruskal-Wallis tests were used to compare blood metrics between synthetic water treatments due to lack of normality across blood metrics. The effect treatment (continuous exposure or transient) and runoff age (Day 0 or 1) on the average amount of time that runoff-exposed fish took to become moribund were each tested by linear mixed-effects models controlling for storm (1 or 2) as a random effect. Time that fish spent at the surface in the continuous behavior experiments was summed across 15-min intervals for each tank. For the first 90 min, two fish were present in each continuous exposure tank (no transfers had yet occurred for the transient exposure). Changes in summed surfacing activity for the six 15-min intervals up to 90 min was assessed by the Friedman test. This non-parametric test was appropriate because the data were non-normally distributed (Shapiro-Wilk: runoff p ≦0.001; control p < 0.001-0.080), and the values for each time bin were dependent (activity of same fish across time bins). Post-hoc analysis used the Wilcoxon Signed Ranks Test for pairwise comparison of dependent samples. Correction of the significance level (α = 0.05) for multiple comparisons was conducted by a modified false discovery rate method (Narum, 2006). 4

Aquatic Toxicology 214 (2019) 105231

M.I. Chow, et al.

Student’s t-tests were applied to compare red blood cell metrics between treatments. Principal component analysis (PCA) was used to identify specific blood profiles associated with each treatment and sampled symptomology Stage. PCA is a linear ordination technique that uses Euclidean distances to describe correlations among measured variables. This reduces the dimensionality of the multivariate matrix into synthetic dimensions (principal components) for easier data interpretation. For each of nine measured blood chemistry metrics, data were collected for 83 individual fish. Of these, n = 3 were omitted as outliers (a Euclidean distance standard deviation > 3), and an additional n = 14 were excluded because we were unable to obtain one or more of the target measures. Consequently, the PCAs were based on blood chemistry data derived from a total of 66 fish. In total, five PCAs were conducted. Analyses were conducted on datasets separated by treatment (ncontrol = 33, nrunoff = 33) and by symptomology Stage (nStage 1 = 25, nStage 3 = 21, nStage 4 = 20). For each dataset, PCA was conducted on a correlation matrix to standardize among variables (mean = 0 and standard deviation = 1). The significance of eigenvalues was tested using the Broken Stick Model and a Monte Carlo randomization test with 1000 permutations. Both tests determine if derived eigenvalues diverge from those predicted by a null hypothesis. Principal components with significant eigenvalues and high variable loadings (> 0.25, absolute value) are considered important for explaining the observed variance and were therefore retained. Eigenvectors were varimax rotated and ordination biplots constructed using the scores of significant principal components. To statistically test if fish loaded differently along principal components depending on behavioral Stage or treatment, linear regression models were applied on the principal component scores (i.e., coordinates of the biplot) for each PCA. All statistical analyses were conducted in R (version 3.3.2; R Core Team, 2016). Significance was set at α = 0.05 in all cases.

The progression of symptoms for runoff-exposed coho closely resembled the behaviors previously reported for adult coho returning to spawn in Puget Sound urban watersheds (Scholz et al., 2011). From behaviors visible from the surface or the bottom of the tank, the sequence was as follows: discrete surfacing events > continuous surface swimming > short bursts of high velocity swimming > loss of equilibrium > loss of buoyancy > unresponsive (moribund). The distinctive characteristics of each of these categories, or Stages, are described in Table 1. While the amount of time spent by each individual fish in a given Stage varied, 11 of the 12 juvenile coho transitioned through the six categories of increasing distress. Control fish exhibited only Stage 1 behavior, characterized by brief episodic surfacing. The amount of surface activity in controls did not change over time (χ2(5) = 6.503; p = 0.260), but that of runoff fish did (χ2(5) = 15.266; p = 0.009; Fig. 3). A significant increase in surface activity over that in the first 15 min occurred for the 60 min time bin (z = -2.249; p = 0.015), which persisted for the subsequent time bins (75 min: z = -2.229, p = 0.026; 90 min: z = -2.668, p = 0.008). Activity in prior time bins was not significantly different than the initial activity level (45 min: z = -1.319, p = 0.187; 30 m in. z = -0.070, p = 0.944). This initial increase in surface activity describes the transition to Stage 2 in runoff-exposed fish. The short episodes of continuous surface swimming during Stage 2 progressed to sustained surface swimming in large circles or around the edges of the tank (Stage 3). Surfacing activity was sustained until the next major shift in symptomology in which fish lost equilibrium (Stage 4). Fish transitioned from surface swimming in a normal orientation (dorsal-up) to surface swimming in an abnormal orientation. In the early phase, this manifested as either a shift from dorsal-up to anteriorup, or the fish rolled onto their sides. Both effects on equilibrium were initially transient, but coho soon began swimming continuously on their side and/or spiraling until they eventually lost buoyancy and sank to the bottom of the exposure tank (Stage 5). Moribund fish on the bottom of the tank were observed ventilating, gaping, and occasionally exhibited spasms (Stage 6). The trial was terminated when (netted) moribund fish were no longer responsive to touch. Several collected blood parameters varied between Stages and treatments (Table 3). Multivariate analyses (PCA) revealed unique blood profiles specific to urban runoff exposure. Across all runoff-exposed fish, the PCA on blood metrics extracted two principle components (PC), both of which varied by Stage (Figure S3). PC1 accounted for 44.11% of the variance in the dataset and PC2 accounted for an additional 26.17% of the variance. A difference between Stages was observed along PC1 (F(1,31) = 5.426; p = 0.027) and PC2 (F (1,31) = 17.55; p < 0.001) indicating a change in blood profiles as the syndrome advanced. The PCA for control fish did not identify a difference between Stages. Important blood metrics for explaining the variation between Stages along PC1 included decreased pH and increased hematocrit, total protein, iCa2+, and pCO2, indicated by high variable loadings associated with PC1 (> 0.25, absolute value). High

3. Results Coho salmon exposed to synthetic water (altered ion content) exhibited none of the behavioral symptoms characteristic of fish exposed to urban runoff. Furthermore, out of the eight measured iSTAT blood variables, only iCa2+ was significantly different between treatments (χ2(3) = 8.584; p = 0.035; Table 2). No other disturbances in blood chemistry were detected. Exposures to collected road runoff were almost universally lethal to juvenile coho salmon (Fig. 1), with 96% of all exposed fish (n = 23 of 24) losing equilibrium and becoming immobilized on the bottom of the exposure tanks (i.e., moribund) within seven hours (mean ± SD: 236 ± 84 min) in both continuous and transient exposures. The linear mixed-effects model showed no difference in the time to moribund between continuous or transient exposures, or the age of stormwater when considering storm as a random effect (p = 0.233; Fig. 2). No mortality was observed in control fish exposed to fish lab water.

Table 2 Morphometric and blood variables for juvenile coho exposed to synthetic waters or control water. sd = standard deviation of the mean. Synthetic waters

Control water

Low ion (n = 4)

Intermediate ion (n = 2)

Fish lab water (n = 2)

Fish lab water (n = 3)

Fish Metric

Units

mean

sd

mean

sd

mean

sd

mean

sd

Length Weight pH pCO2 Na iCa2+ Glucose HCT

mm g

49.25 175.00 7.23 14.32 141.75 1.63 119.75 34.25

26.79 26.05 0.09 3.02 3.50 0.06 21.70 6.08

47.00 176.50 7.26 13.37 145.50 1.56 134.00 31.50

26.87 26.16 0.02 3.14 6.36 0.02 7.07 0.71

32.50 148.00 7.28 12.06 140.00 1.48 142.50 37.00

21.92 28.28 0.00 0.37 7.07 0.05 19.09 11.31

39.33 160.67 7.17 14.53 145.67 1.51 154.33 38.67

10.79 12.86 0.05 1.53 1.15 0.02 25.50 4.04

mmHg mmol/L mmol/L mg/DL %PCV

5

Aquatic Toxicology 214 (2019) 105231

M.I. Chow, et al.

Fig. 1. Juvenile coho exposed transiently to urban runoff and then transferred to clean water (Transient Runoff) became moribund at the same rate as fish exposed to runoff during the entire experiment (Continuous Runoff). There were no time-matched control fish that became moribund from the continuous or transient control exposures. All treatments represent a sample size of 3 fish. Exposure 1 and 2 were conducted using Storm 1 (04/10/2017) and Exposure 3 and 4 conducted using Storm 2 (04/23/2017). Fig. 2. Juvenile coho exposed transiently to urban runoff and then transferred to clean water (Transient) became moribund over the same time course as fish continuously exposed to runoff (Continuous) for each exposure to each storm. Each treatment represents a sample size of 3 fish. Exposure 1 and 2 were conducted using Storm 1 (04/10/2017) and Exposure 3 and 4 conducted using Storm 2 (04/23/2017). Error bars indicate ± one standard error of the mean.

Fig. 3. Runoff-exposed fish spent more time at the surface relative to coho in clean water. Control fish only exhibited episodic surfacing behavior (discrete events < 0.125 s) whereas runoff-exposed fish progressed to periods of continuous surface swimming (discrete events > 2 s). Each point represents the summed surfacing time per tank over a 15-minute interval (Control = white; Runoff = black). The lines represent the mean of the summed surfacing time across tanks per 15-minute interval (dotted = Runoff, solid = Control). Shapes indicate the number of fish in each tank (circles = 1 fish, triangles = 2 fish) during the 15-minute interval. Across the entire experiment, summary statistics for surfacing (median; mean; standard deviation; max) were 0.1; 0.5; 2.0; 25.1 s/15 min for control fish and 12.5; 67.0; 123.2; 720.0 s/15 min for runoff-exposed fish.

loadings for PC2 included a decrease in Na+, lactate, and osmolality and an increase in pH. The PCA for Stage 1 did not identify specific blood profiles associated with runoff exposure, as no difference in principal component scores was observed between treatments (Table 4; Fig. 4). For fish sampled at Stage 3 and Stage 4, however, PCA did identify specific blood profiles corresponding with runoff exposure (Table 4; Fig. 4). For Stage 3, two principal components were extracted; PC1, explaining 35.36% of the variance, and PC2, explaining 26.00% of the variance. Treatment did not explain the difference in principle component scores for PC1 (F(1,19) = 0.032; p = 0.860), but a significant difference between treatments was identified along PC2 for principal component scores (F(1,12) = 12.43; p = 0.002), associated with a decrease in Na+ and osmolality, and an increase in pCO2, HCT, and total protein. For fish sampled at Stage 4, PC1 explained 59.57% of the variance, while PC2 explained 13.45% of the variance. Along PC1, there was a significant difference in principal component scores between treatments (F

(1,18) = 13.96; p = 0.002). High variable loadings included a decrease in Na+ and osmolality and an increase in glucose, hematocrit, lactate, and total protein. Treatment did not explain the loadings for PC2 (F (1,18) = 1.69; p = 0.210). Red blood cell viability at Stage 4 was greater than 99% for both control (mean ± SD: 99.5% ± 0.004) and runoff-exposed fish (mean ± SD: 99.2% ± 0.006). Red blood cell count was not significantly different between treatments in fish sampled at Stage 4 (t(20) = -0.875; p = 0.392). However, a significant difference between treatments was observed for the mean corpuscular volume of red blood cells at Stage 4 (t(20) = -3.678; p = 0.001; Fig. 5). Conventional water chemistry parameters and metals measured in runoff and control waters collected throughout the study are summarized in Table S2 and S3.

6

Aquatic Toxicology 214 (2019) 105231

M.I. Chow, et al.

Table 3 Morphometric and blood variables for juvenile coho sampled at specific behavioral stages during runoff exposure along with time-matched controls. Abbreviations include: CF – condition factor, iCa2+ – ionized calcium, HCT – hematocrit, TP – total protein, and Osmo – osmolality. Stage 1: Asymptomatic Control

Length Weight CF pH pCO2 Na+ iCa2+ Glucose HCT Lactate TP Osmo

Runoff

Stage 3: Surface swimming

Stage 4: Loss of equilibrium

Control

Control

Runoff

Runoff

Units

n

mean

sd

n

mean

sd

n

mean

sd

n

mean

sd

n

mean

sd

n

mean

sd

mm g kg/cm3

15 15 15 14 14 14 14 13 14 13 13 13

256.4 204.5 1.17 7.31 1.3 139.7 1.45 5.7 33.8 11.63 5.74 304

27.9 67 0.12 0.08 0.1 4.75 0.08 0.8 4.54 5.06 0.78 15.5

15 15 15 14 14 14 14 14 13 13 13 13

251.9 203.8 1.15 7.31 1.2 143.5 1.48 5.7 34.3 10.44 5.38 310

50.8 105.6 0.13 0.11 0.2 3.55 0.09 1.0 8.86 4.26 1.47 17.4

15 15 15 11 11 11 10 11 11 11 11 12

259.5 204.5 1.14 7.33 1.4 139 1.44 6.2 31.2 11.68 5.18 309

28.3 67 0.13 0.11 0.2 5.04 0.09 1.2 4.96 6.5 0.91 11.2

15 15 15 15 15 15 15 14 14 12 12 12

259.5 203.4 1.07 7.32 1.5 137.6 1.49 5.7 38.9 7.31 6.35 294

65.6 105.6 0.26 0.09 0.3 6.16 0.13 1.5 7.34 5.87 0.91 17.4

15 15 15 14 13 13 13 13 14 10 10 10

256.8 219 1.32 7.37 1.4 138.2 1.43 5.7 30 3.91 5.16 298

33.4 78.5 0.65 0.09 0.4 3.56 0.12 0.8 7.93 2.26 0.76 7.9

15 15 15 15 13 15 14 14 15 12 12 12

261.2 224.2 1.16 7.27 1.5 131.9 1.51 8.0 46.9 5.8 6.78 285

45.1 119.1 0.24 0.12 0.5 5.82 0.21 1.7 12.76 4.33 1.73 17.7

kPa mmol/L mmol/L mmol/L %PCV mmol/L mmol/L mOsm/kg

Table 4 Principal component analysis (PCA) revealed specific blood profiles associated with urban runoff exposures. Stage 1 (n = 25)

Eigenvalue % Variance Monte Carlo pH pCO2 Na+ iCa2+ Glu HCT Lactate Protein Osmo Difference between treatments F-value df p-value

Stage 3 (n = 21)

Stage 4 (n = 20)

PC1

PC2

PC1

PC2

PC1

PC2

3.25 36.06 < 0.001* 0.483

2.13 23.71 < 0.001*

3.18 35.36 0.002* 0.465 −0.378 −0.348 −0.359 −0.211

2.34 26.00 < 0.001* −0.190 0.272 −0.337

5.36 59.57 < 0.001*

1.21 13.45 1.000 0.449 −0.521 0.133 −0.554 0.152 −0.270

−0.436 −0.319 −0.334 −0.436 −0.409 0.508 1,23 0.483

−0.517 0.218

−0.620

−0.424 −0.113 −0.383

0.566 −0.147 0.535 −0.381

−0.401 −0.123 0.565 0.284 0.255 0.265 −0.528

0.530 1,23 0.474

0.032 1,19 0.860

12.43 1,19 0.002*

13.96 1,18 0.002*

−0.524

−0.276 −0.162 1.69 1,18 0.210

Bold indicates high variable loadings for each principal component (> 0.25 absolute value) indicating a high correlation of the blood metric to the principal component. PC scores between treatments indicate the significance (p-value) of principal component scores between treatments determined by linear regression. Abbreviations include: Monte Carlo – the significance of each eigenvalue obtained from Monte Carlo randomization test; iCa2+ – ionized calcium; HCT – hematocrit; TP – total protein; Osmo – osmolality; * –indicates significance.

4. Discussion

loss of equilibrium) was remarkably similar to behaviors widely observed among distressed coho spawners in Puget Sound urban streams (Scholz et al., 2011). The dysregulation of blood chemistry was also similar for the two life history stages, in both cases culminating in severe ion and osmoregulatory impairment as the syndrome progressed (for adult blood chemistry, see McIntyre et al., 2018). Despite a somewhat small sample size in the present study, the manifestation and characteristics of the syndrome were highly consistent across 23 of 24 fish (96%) exposed to runoff. This extends previous studies showing that untreated urban roadway runoff is almost always acutely lethal to coho (McIntyre et al., 2018, 2015). Although urbanization and associated changes in land cover and land use are an ongoing and increasing global phenomenon, there have been relatively few studies on the impacts of roadway runoff on fish physiology, behavior, and survival. Our current results extend earlier findings for brown trout (Salmo trutta) experimentally exposed to stormwater discharged from a highway near Oslo, Norway (Meland et al., 2010a&b). While not lethal to brown trout, runoff exposures dysregulated blood chemistry over a time course of a few hours. Similar to coho, PCA showed a shift in brown trout blood parameters driven by a decrease in Na+ concentration and an increase in glucose and

Coastal areas of the contiguous United States, including lowland stream and river networks in the Pacific Northwest, are increasingly under pressure from human population growth and associated increases in development, impervious area, stormwater runoff, and degraded freshwater habitat quality. In Puget Sound, adult coho salmon have proven to be an important sentinel species for both toxic stormwater impacts (Scholz et al., 2011; Spromberg et al., 2016; McIntyre et al., 2018), as well as the effectiveness of clean water mitigation strategies in the form of green stormwater infrastructure (Spromberg et al., 2016). However, the corresponding impacts of untreated runoff on other coho life stages, in particular freshwater-phase juveniles, are poorly understood. This study provides the first detailed analysis of the urban mortality syndrome in juvenile coho, including the temporal progression of behavioral and physiological symptoms in response to experimental exposures to stormwater runoff collected from a high traffic volume arterial in Seattle. As with adults, runoff from this location was almost universally acutely lethal to juvenile coho, and the sequence of behavioral symptoms (e.g., surfacing and gaping > spiral swimming > 7

Aquatic Toxicology 214 (2019) 105231

M.I. Chow, et al.

4.1. Behavioral indications of coho distress During fall spawner surveys in the field, there are essentially two methods for positively confirming the coho urban stream mortality syndrome. The first is a forensic examination of carcasses. In the absence of predation, dead females with most or all of their eggs retained (i.e., unspawned) are considered pre-spawn mortalities (Scholz et al., 2011). The second method is more elusive, as it requires video documentation of live but symptomatic fish (both males and females). Direct evidence of symptomology in the field is more difficult to obtain because encounter rates with live fish in urban streams can be relatively low. Also, affected coho usually die on a timescale of a few hours, and thus field surveys often miss the window of overt symptomology. For these reasons, and the logistical difficulties of controlled experiments involving large volumes of stormwater and coho returning to spawn at hatcheries (Spromberg et al., 2016; McIntyre et al., 2018), the progression of behavioral symptoms associated with the mortality syndrome has not been carefully studied in adults. The sequence of surface- or bottom-oriented behaviors leading up to juvenile mortality was remarkably consistent, both within and between storms. This allowed for a categorization of severity, or designation of Stages, that was used as reference points for evaluating physiological parameters in blood and tissues. The behavioral reference points also provided a basis for answering, at least provisionally, two key questions. The first is whether stormwater toxicity to coho is irreversible. This is important because stormwater runoff events are transient in space and time. If individual coho can survive a short-term influx of contaminants to freshwater habitats, the conservation threats to wild populations might be diminished. Accordingly, the transfer of intermediately distressed fish in stormwater (i.e., surface swimming) to clean water had one of two likely outcomes: 1) a reversal of the symptomology sequence and a return to normal swimming behavior, or 2) a continuation along the sequence, culminating in death. Our results suggest the mortality syndrome is irreversible, and that coho are likely to die if a single storm degrades surface water quality to the extent fish become visibly symptomatic. The second key question is a differentiation between toxic chemical contaminants in stormwater and the substantive differences in ion content between highway runoff and clean reference water. The lack of a behavioral response to synthetic water amended with conventional ions at the lower and intermediate range of measured levels in collected runoff (Spromberg et al., 2016) suggests that coho are dying in response to one or more toxic contaminants in roadway runoff and not from changes in conventional water chemistry. This reinforces the need to characterize the full chemical complexity of urban runoff (Peter et al., 2018; Du et al., 2017) as a basis for fractionation strategies to identify the as-yet unknown causal agent(s).

Fig. 4. Ordination biplots from principal component analysis (PCA) on fish sampled at Stage 3 (surfacing) and Stage 4 (loss of equilibrium) indicate a unique blood profile for runoff-exposed juveniles: a) principal component scores (i.e., location of individuals along principle components) of fish sampled at Stage 3 were significantly different by treatment along PC2 (linear regression; p = 0.002), but not PC1; b) principal component scores for fish sampled at Stage 4 were significantly different by treatment along PC1 (linear regression; p = 0.002), but not PC2. Ellipsoids represent the standard deviation of principal component scores by treatment group.

hematocrit (Meland et al., 2010b). It is becoming increasingly evident that salmonid species vary in their sensitivity to untreated urban stormwater (e.g., McIntyre et al., 2018). However, much more work is needed to understand the chemical composition of runoff and the etiology of lethal and sublethal toxicity across fish species.

4.2. Physiological correlates of behavioral stress Our analyses of blood chemistry revealed a significant effect of treatment on blood ions and hematology. Stormwater-exposed fish sampled at Stage 3 (surface swimming) showed evidence of an osmotic Fig. 5. Runoff-exposed fish sampled at loss of equilibrium had a significantly higher hematocrit and red blood cell volume compared to time-matched controls: a) Hematocrit as percent packed cell volume (%PCV); b) Total density of red blood cells (RBC); c) Mean corpuscular volume (MCV in femtolitres (fL)). Asterisks indicate statistical difference among treatments.

8

Aquatic Toxicology 214 (2019) 105231

M.I. Chow, et al.

imbalance in the form of a decrease in osmolality, Na+, total protein concentration, and red blood cell swelling – the latter contributing to an increased hematocrit. The specific cause of the osmotic imbalance is unknown, but could be due to increased respiratory demands, impairment of osmoregulatory function at either the gill or kidney, or another unidentified mechanism. By Stage 4 (loss of equilibrium), along with an osmotic imbalance, runoff-exposed fish developed elevated glucose and lactate. Changes in surface activity occur much earlier than these changes in hematology. Future studies should look for physiological changes during Stage 2 in order to better understand the etiology of the syndrome. The observed effects on the blood chemistry of juvenile coho exposed to urban runoff are comparable to the effects recently reported for symptomatic coho adults (McIntyre et al., 2018). At the point of equilibrium loss (Stage 4 in this study), both life history stages had a measurable increase in blood glucose, lactate, and hematocrit, and a decrease in blood ion content. One notable difference between studies, however, was a characteristic blood acidification in adults that was not detected in the juveniles assessed here. This may be due to a difference in sampling time between life history stages rather than physiological response. Specifically, blood was previously collected from adults that were more severely symptomatic, suggesting a progressive increase in acidosis. Nevertheless, the behavioral aspects of the mortality syndrome, the dysregulation of specific blood parameters, and the rapid onset of the phenomenon are very similar for juvenile and adult coho. The above core finding suggests a common toxic mechanism across life stages. If so, juveniles represent an expanded platform for understanding and eventually mitigating stormwater toxicity to wild coho. Over the past two decades, the logistical challenges associated with studies on adults have slowed the pace of research and limited the range of experimental questions that can be addressed in a given year. By contrast, juveniles are convenient to handle, amenable to small volume exposures, and abundantly available year-round. In the future, these considerations will be particularly important in the context of high-throughput screens of chemical sub-mixtures obtained via targeted fractionation. From a management perspective, the similar effect of urban runoff exposures on juvenile coho also raises the question of susceptibility of juvenile coho salmon in the field. Juveniles spend a considerably longer time in urban creeks than most other Pacific salmonids where they may be exposed to urban stormwater runoff. Compared to adult spawners, juvenile mortality is more difficult to monitor, as juvenile carcasses are smaller, would likely be eaten by predators, and mortality could be occurring year-round. However, lower abundances of juvenile coho have been observed in urban watersheds compared to non-urban ones (Scott et al., 1986; May et al., 1997) that could potentially be related to the lethal effects of urban runoff observed in this study. If juveniles are similarly susceptible to urban runoff in the field, long-term coho conservation in urbanizing watersheds may be even more challenging than previously predicted (Spromberg and Scholz, 2011). In conclusion, our results confirm the urban runoff mortality syndrome in juvenile coho, consistent with past field observations (Video S1 and S2) and laboratory bioassays (McIntyre et al., 2015). However, representative rates of juvenile mortality in urban streams remain unknown. Adult salmon are easy to locate in spawning habitats, and fish lost to the syndrome are readily confirmed by the retention of eggs in dead females. This is not the case for juveniles, and there are practically no in situ observational data for juvenile losses in urban watersheds, particularly during the fall and winter residence (rearing) months when storm events are common. Although numerous studies have shown that juvenile coho abundances are lower in urban stream networks relative to rural or forested habitats (Scott et al., 1986; May et al., 1997), the role of degraded water quality in this trend remains unclear. More work is needed to define juvenile losses, in part because initial population models – which forecast rapid localized extinctions – are based exclusively on simulated mortality rates for adults (Spromberg and

Scholz, 2011). Accurate toxicity information for other life stages is needed, because the stormwater threat to imperiled coho in the Pacific Northwest, now and in the future, is likely greater than previously anticipated. Our findings also further reinforce the need for green stormwater infrastructure and other development strategies that limit polluted runoff, as these methods have been shown to remove contaminants and protect both coho juveniles (McIntyre et al., 2015) and adults (Spromberg et al., 2016) from the mortality syndrome. Funding This research was funded in part by the U.S. Environmental Protection Agency, Region 10’s Puget Sound National Estuary Program, and Seattle Public Utilities. This study received agency support and funding from the National Marine Fisheries Service, the U.S. Fish & Wildlife Service, the University of Washington, and Washington State University. Acknowledgements We appreciate a critical review of the draft manuscript by David Baldwin (Northwest Fisheries Science Center) and the indispensable assistance provided by K. King (U.S. Fish & Wildlife Service), E. Mudrock, J. Prat, and J. Wetzel (Washington State University), and J. Cameron, C. Laetz, and D. Baldwin at NWFSC. Findings and conclusions herein are those of the authors and do not necessarily represent the views of the sponsoring organizations. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.aquatox.2019.105231. References Alberti, M., Booth, D., Hill, K., Coburn, B., Avolio, C., Coe, S., Spirandelli, D., 2007. The impact of urban patterns on aquatic ecosystems: an empirical analysis in Puget lowland sub-basins. Landsc. Urban Plan. 80, 345–361. Anderson, R.O., Neumann, R.M., 1996. Length, weight, and associated structural indices. In: Murphy, B.R., Willis, D.W. (Eds.), Fisheries Techniques. Fisheries Society, Bethesda, pp. 447–481. Bilby, R.E., Mollot, L.A., 2008. Effect of changing land use patterns on the distribution of coho salmon (Oncorhynchus kisutch) in the Puget Sound region. Can. J. Fish. Aquat. Sci. 65, 2138–2148. Booth, D.B., Hartlet, D., Jackson, R., 2002. Forest cover, impervious-surface area, and the mitigation of stormwater impacts. J. Am. Water Resour. Assoc. 38, 835–845. Cooke, S.J., Suski, C.D., Danylchuk, S.E., Danylchuk, A.J., Donaldson, M.R., Pullen, C., Bulte, G., O’Toole, A., Murchie, K.J., Koppelman, J.B., Shultz, A.D., Brooks, E., Goldberg, T.L., 2008. Effects of different capture techniques on the physiological condition of bonefish Albula vulpes evaluated using field diagnostic tools. J. Fish Biol. 73, 1351–1375. Du, B., Lofton, J.M., Peter, K.T., Gipe, A.D., James, C.A., McIntyre, J.K., Scholz, N.L., Baker, J.E., Kolodziej, E.P., 2017. Development of suspect and non-target screening methods for detection of organic contaminants in highway runoff and fish tissue with high-resolution time-of-flight mass spectrometry. Environ. Sci. Process. Impacts 19, 1185–1196. Feist, B.E., Buhle, E.R., Baldwin, D.H., Spromberg, J.A., Damm, S.E., Davis, J.W., Scholz, N.L., 2017. Roads to ruin: conservation threats to a sentinel species across an urban gradient. Ecol. Appl. 27, 2382–2396. Feist, B.E., Buhle, E.R., Arnold, P., Davis, J.W., Scholz, N.L., 2011. Landscape ecotoxicology of coho salmon spawner mortality in urban streams. PLoS One 6, e23424. Forrestal, F.C., McDonald, M.D., Burress, G., Die, D.J., 2017. Reflex impairment and physiology as predictors of delayed mortality in recreationally caught yellowtail snapper (Ocyurus chrysurus). Conserv. Physiol. 5, cox035. Gallagher, A.J., Frick, L.H., Bushnell, P.G., Brill, R.W., Mandelman, J.W., 2010. Blood gas, oxygen saturation, pH, and lactate values in elasmobranch blood measured with a commercially available portable clinical analyzer and standard laboratory instruments. J. Aquat. Anim. Health 22, 229–234. Harrenstien, L., Tornquist, S., Miller-Morgan, T., Fodness, B., Clifford, K., 2005. Evaluation of a point-of-care blood analyzer and determination of reference ranges for blood parameters in rockfish. J. Am. Vet. Med. Assoc. 226, 255–265. Harter, T.S., Shartau, R.B., Baruner, C.J., Farrell, A.P., 2014. Validation of the i-STAT system for the analysis of blood parameters in fish. Conser. Physiol. 2, 1–12. Marsalek, J., Rochfort, Q., Brownlee, B., T, M, Servos, M., 1999. An exploratory study of urban runoff toxicity. Water Sci. Technol. 39, 33–39.

9

Aquatic Toxicology 214 (2019) 105231

M.I. Chow, et al. May, C.W., Horner, R.R., Karr, J.R., Mar, B.W., Welch, E.B., 1997. Effects of urbanization on small streams in the Puget Sound ecoregion. Watershed Prot. Tech. 2, 483–494. McIntyre, J.K., Davis, J.W., Hinman, C., Macneale, K.H., Anulacion, B.F., Scholz, N.L., Stark, J.D., 2015. Soil bioretention protects juvenile salmon and their prey from the toxic impacts of urban stormwater runoff. Chemosphere 132, 213–219. McIntyre, J.K., Lundin, J.I., Cameron, J.R., Chow, M.I., Davis, J.W., Incardona, J.P., Scholz, N.L., 2018. Interspecies variation in the susceptibility of adult Pacific salmon to toxic urban stormwater runoff. Environ. Pollut. 238, 196–203. Meland, S., Heier, L.S., Salbu, B., Tollefsen, K.E., Farmen, E., Rosseland, B.O., 2010a. Exposure of brown trout (Salmo trutta L.) to tunnel wash water runoff - chemical characterisation and biological impact. Sci. Total Environ. 408, 2646–2656. Meland, S., Salbu, B., Rosseland, B.O., 2010b. Ecotoxicological impact of highway runoff using brown trout (Salmo trutta L.) as an indicator model. J. Environ. Monit. 12, 654–664. Meyer, J.L., Paul, M.J., Taulbee, W.K., 2005. Stream ecosystem function in urbanizing landscapes. J. North Am. Benthol. Soc. 24, 602–612. Morley, S.A., Karr, J.R., 2002. Assessing and restoring the health of urban streams in the Puget Sound basin. Conserv. Biol. 16, 1498–1509. Moscrip, A.L., Montgomery, D.R., 1997. Urbanization, flood frequency, and salmon abundance in Puget lowland streams. J. Am. Water Resour. Assoc. 33, 1289–1297. Narum, S.R., 2006. Beyond Bonferroni: less conservative analyses for conservation genetics. Conserv. Genet. 7, 783–787. Paul, M.J., Meyer, J.L., 2001. Streams in the urban landscape. Annu. Rev. Ecol. Syst. 32, 333–365. Pess, G.R., Montgomery, D.R., Steel, E.A., Bilby, R.E., Feist, B.E., Greenberg, H.M., 2002. Landscape characteristics, land use, and coho salmon (Oncorhynchus kisutch) abundance, Snohomish River, Wash., U.S.A. Can. J. Fish. Aquat. Sci. 59, 613–623. Peter, K.T., Tian, Z., Wu, C., Lin, P., White, S., Du, B., McIntyre, J.K., Scholz, N.L., Kolodziej, E.P., 2018. Using high-resolution mass spectrometry to identify organic contaminants linked to urban stormwater mortality syndrome in coho salmon. Environ. Sci. Technol. 52, 10317–10327. R Core Team, 2016. R: a Language and Environment for Statistical Computing. R Project

for Statistical Computing, Vienna, Austria. www.r-project.org. Regan, M.D., Turko, A.J., Heras, J., Andersen, M.K., Lefevre, S., Wang, T., Bayley, M., Brauner, C.J., Huong, D.T.T., Phuong, N.T., Nilsson, G.E., 2016. Ambient CO2, fish behaviour and altered GABAergic neurotransmission: exploring the mechanism of CO2-altered behaviour by taking a hypercapnia dweller down to low CO2 levels. J. Exp. Biol. 219, 109. Regetz, J., 2003. Landscape-level constraints on recruitment of chinook salmon (Oncorhynchus tshawytscha) in the Columbia River basin, USA. Aquat. Conserv. Mar. Freshw. Ecosyst. 13, 35–49. Scholz, N.L., McIntyre, J.K., 2016. Chemical pollution. In: Fishes Closs, G.P., Krkosek, M., Olden, J.D. (Eds.), Conservation of Freshwater. Cambridge University Press, Cambridge, pp. 149–177. Scholz, N.L., Myers, M.S., McCarthy, S.G., Labenia, J.S., McIntyre, J.K., Ylitalo, G.M., Rhodes, L.D., Laetz, C.A., Stehr, C.M., French, B.L., McMillan, B., Wilson, D., Reed, L., Lynch, K.D., Damm, S., Davis, J.W., Collier, T.K., 2011. Recurrent die-offs of adult coho salmon returning to spawn in Puget Sound lowland urban streams. PLoS One 6, e29013. Scott, J.B., Steward, C.R., Stober, Q.J., 1986. Effects of urban development on fish population dynamics in Kelsey Creek, Washington. Trans. Am. Fish. Soc. 115, 555–567. Spromberg, J.A., Baldwin, D.H., Damm, S.E., McIntyre, J.K., Huff, M., Sloan, C.A., Anulacion, B.F., Davis, J.W., Scholz, N.L., 2016. Coho salmon spawner mortality in western US urban watersheds: bioinfiltration prevents lethal storm water impacts. J. Appl. Ecol. 53, 398–407. Spromberg, J.A., Scholz, N.L., 2011. Estimating the future decline of wild coho salmon populations resulting from early spawner die-offs in urbanizing watersheds of the Pacific Northwest, USA. Integr. Environ. Assess. Manag. 7, 648–656. Strober, W., 2001. Trypan blue exclusion test of cell viability. Current Protocols in Immunology. pp. A.3B.1–A.3B.2. Walsh, C.J., Roy, A.H., Feminella, J.W., Cottingham, P.D., Groffman, P.M., Morgan, R.P., 2005. The urban stream syndrome: current knowledge and the search for a cure. J. North Am. Benthol. Soc. 24, 706–723.

10