Juvenile rockfish show resilience to CO2-acidification and hypoxia across multiple biological scales

The present study showed short-term elevations in cellular metabolism, alterations in behavior and susceptibility to predation in juvenile rockfish after acute acclimation to CO2-acidification and hypoxic conditions. Physiological and behavioral alterations were restored after 3 weeks suggesting that rockfish possess mechanisms to defend rapid changes in PCO2 and oxygen conditions.


Introduction
Climate change is predicted to have profound implications for marine ecosystems, with adverse ramifications expected for species abundance and diversity, and the sustainability of commercial fisheries (Brander, 2007;Wood and McDonald, 1997). Shifts in multiple environmental factors such as elevated dissolved carbon dioxide (CO 2 , hereafter CO 2 -acidification), low dissolved oxygen (DO, hereafter hypoxia) and increased temperature (Crain et al., 2008;Keeling et al., 2010;Doney et al., 2011) have the potential to threaten individual species and communities by negatively affecting animal physiology and behavior (see reviews Heuer and Grosell, 2014;Nagelkerken and Munday, 2016;Tresguerres and Hamilton, 2017). While the organismal responses to CO 2 -acidification and hypoxia have individually been explored, in nature these factors can change in concert. In particular, in highly productive, mid-latitude coastal systems with seasonal upwelling, seawater can be both high in partial pressure of CO 2 (PCO 2 ) and low in DO due to expansion of oxygen minimum zones (Chan et al., 2008;Largier et al., 2015). This results in dynamic changes in both pH and oxygen in coastal habitats (Feely et al., 2008).
Environmental stressors can directly impact the physiology of individual animals (Hochachka and Somero, 2002), and physiology is the mechanistic driver of an animal's behavior (Kane et al., 2005). The effects of abiotic environmental stressors on organismal physiology and behavior can then lead to alterations in trophic interactions, highlighting the need to scale up responses to understand ecological implications of environmental change. To accurately explore the effects of multiple stressors (e.g. CO 2 -acidification and hypoxia), and link organismal mechanisms to community function, it is vital to examine interactive effects across multiple biological scales, from cellular metabolism to species interactions. For example, physiological costs that result in reduced growth or altered locomotion may increase predation risk (Briffa et al., 2012;Allan et al., 2013;Miller et al., 2016). This is a process through which both physiological and behavioral impacts can have fitness consequences and influence population and community dynamics Munday et al., 2013).
Fish are acid-base regulators, and hence, the effect of CO 2 -acidification on the physiology and behavior of fishes has been projected to be less severe than on other marine species (Kroeker et al., 2010;Heuer and Grosell, 2014). While adult fish are generally less sensitive to CO 2 -acidified seawater, there is accumulating evidence to suggest negative physiological and behavioral costs on larval and juvenile life stages (Munday et al., 2009;Ferrari et al., 2011aFerrari et al., , b, 2012Simpson et al., 2011;Baumann et al., 2012;Allan et al., 2013;Chivers et al., 2014;Chung et al., 2014;Davis et al., 2017). Early life stages of fish have also been shown to be sensitive to hypoxia (Pichavant et al., 2000;Taylor and Miller, 2001;Depasquale et al., 2015). Juvenile fish energetic demands are driven by development and growth such that multiple environmental stressors may create trade-offs, both physiological (e.g. energy allocation for stress response mechanisms and growth and metabolism [Pörtner et al., 2005;Ishimatsu et al., 2008;Pankhurst and Munday, 2011;Sokolova et al., 2012]) and behavioral (e.g. foraging activity and avoiding predation [Lima and Dill, 1990;Houston et al., 1993]). Specifically, CO 2 -acidification and hypoxia can exert sub-lethal physiological impacts that influence performance of animals (Ishimatsu et al., 2004;Sokolova, 2013;Todgham and Stillman, 2013;Steckbauer et al., 2015;Depasquale et al., 2015;Gobler and Baumann, 2016) through increased energetic costs  or a heightened stress response (Hamilton et al., 2014). It is also likely that many species of juvenile fish have a lower capacity to behaviorally avoid changing conditions compared to adult fishes, as early life stages are often more constrained to particular microhabitats due to interactions between abiotic factors and predation risk (Blaber and Blaber, 1980;Laegdsgaard and Johnson, 2001). Therefore, with an inability to avoid adverse environmental conditions, stressors may cause suboptimal performance resulting in slower growth (Gobler et al., 2014;Depasquale et al., 2015), reduced feeding (Appelhans et al., 2012;Thorarensen et al., 2017) and an increased susceptibility to predation (Briffa et al., 2012;Allan et al., 2013;Miller et al., 2016).
Although both predators and prey in nature will likely be exposed to similar environments, most studies to date have only exposed prey species to various projected environmental changes and then tested their capacity to evade predation by predators that were held under control conditions (but see Ferrari et al., 2015). However, if predation efficiency is also impacted by CO 2 -acidification or hypoxia, predator-prey dynamics would be very different than outcomes from studies where only the prey has been exposed. Therefore, the impacts of CO 2 -acidification and hypoxia on biotic interactions can be more robustly evaluated with multi-species exposure studies and will provide insights for communitylevel impacts if multiple stressor impacts are mediated through food webs (Pörtner and Peck, 2010).
Fishes living in seagrass habitats are regularly exposed to highly variable pH and DO, due to respiration and photosynthesis in tandem with other factors such as upwelling  (Diaz and Rosenberg, 2008;Diaz and Breitburg, 2009;Torres et al., 2012); however, these habitats are often overlooked when examining the influence of environmental stressors on organismal physiology and behavior. Previous studies on closely related juvenile rockfish (genus Sebastes) showed that physiological and behavioral responses to elevated PCO 2 were dependent on life-history strategies . For example, juvenile blue rockfish (S. mystinus) that inhabit deeper, low pH habitats were relatively robust to laboratory-elevated PCO 2 conditions. In contrast, juvenile copper rockfish (S. caurinus) that inhabit shallower-surface waters of kelp forests with dynamic fluctuations in pH (and DO levels) showed increased sensitivities to elevated PCO 2 exposures, with alterations in swimming, metabolism and gene expression . To date, physiological and behavioral responses to elevated PCO 2 and hypoxia of juvenile rockfish (genus Sebastes) that develop in a highly fluctuating, variable environment such as seagrass beds have not been explored. Rockfish that inhabit seagrass beds may be particularly robust to environmental perturbation with an enhanced ability to compensate for alterations in PCO 2 and DO compared to species living in more stable environments. Additionally, no studies have assessed the effects of multiple stressors on the interactions between juvenile rockfish and their predators, juvenile cabezon (Scorpaenichthys marmoratus) that share the same fluctuating PCO 2 and DO environment.
Juvenile rockfish and cabezon provide a good study system to explore the interacting effects of multiple environmental stressors (CO 2 -acidification and hypoxia) at multiple biological scales (i.e. cellular physiology and predator-prey dynamics), and to assess whether these fishes have an ability to compensate physiologically or behaviorally to these stressors. The objectives of this study were therefore; (1) to determine if there are energetic costs associated with acute (1 week) and chronic (3 week) exposures to elevated PCO 2 and low DO (hypoxic) conditions in juvenile rockfish and cabezon, (2) to characterize alterations in rockfish movement and antipredator behaviors in response to alarm cue exposure, and finally (3) to assess if actual predator-prey relationships may be altered under exposures to high PCO 2 and hypoxic environments. We predicted that there would be alterations in aerobic and anaerobic metabolic pathways that would affect the capacity to generate ATP after 1 and 3 weeks of acclimation to the two stressors (Windisch et al., 2011;Hamilton et al., 2017), and that movement behavior would also be altered to conserve energy (Herbert and Steffensen, 2005;Davis et al., 2017). Predictions for alterations in behavior, and in turn, effects on predation rates may not be as straight forward with potential interactions between multiple stressors. Elevated PCO 2 may reduce antipredator responses (Ferrari et al., 2012), but hypoxia might also reduce activity which tends to reduce encounter rates with predators. Additionally, either high PCO 2 or hypoxia might reduce swimming and escape ability in rockfish, but also have negative effects on predator activity or attack ability. Therefore, overall effects on predation rates of rockfish under multiple stressors of high PCO 2 and hypoxia are likely complex. By examining the effects of multiple stressors across biological scales of two naturally-interacting species, this study will help our understanding of potential 'trade-offs' in physiological mechanisms and behavioral strategies of prey, and how that might alter predator-prey dynamics of two recreationally and commercially important species (Parker et al., 2000).

Fish collection and maintenance
Juvenile rockfish (Sebastes spp.) and cabezon (S. marmoratus) were collected from seagrass beds in Campbell Cove, CA, USA (38°18′19″N, 123°03′28″W) using beach seine nets. Rockfish and cabezon were collected twice for two successive experiments. Fish for the first physiology experiment (Phase I) were collected late April-May 2015 (rockfish: 28.96 ± 2.54 mm and 0.41 ± 0.14 g and cabezon: 83.92 ± 10.97 mm and 12.69 ± 5.44 g, mean ± SD); whereas fish for the second behavior and predation experiment (Phase II) were collected July-August 2015 (rockfish: 31.81 ± 3.18 mm and 0.71 ± 0.22 g and cabezon: 83.21 ± 10.72 mm and 13.77 ± 4.60 g, mean ± SD). Fish were immediately transported to the Bodega Marine Laboratory, University of California Davis, within 2 h of collection. Rockfish and cabezon were held in separate flow-through seawater tanks (32 ± 0.1 ppt, 14°C ± 1°C) for four (Phase II) or six (Phase I) weeks until the start of experiments. Fish were fed finely chopped frozen squid ad libitum daily during the laboratory acclimation period. This research project was conducted in accordance with animal welfare laws approved by the University of California Davis Institutional Animal Care and Use Committee.
All experiments were conducted at the subgenera level, using Pteropodus rockfishes. Post-experimental DNA barcoding of the cytochrome c oxidase loci (COI) conducted on only a small subset of rockfish confirmed that the study included a mix of species in the complex of seagrass rockfishes including gopher rockfish (S. caratus), copper rockfish (S. caurinus), and black-and-yellow rockfish (S. chrysomelas), which all exhibit similar ecological and genetic traits. The genetic similarities of rockfish collected in this study made it impossible to DNA barcode rockfish that were consumed by cabezon; however, given the shared ecological and genetic traits of these species (Echeverria, 1987;Love et al., 2002;Li et al., 2006;Wheeler et al., 2017), we assume negligible differences in the physiological and behavioral responses of Pteropodus rockfishes. For these reasons, and to maintain consistency with other studies (Caselle et al., 2010a, b;Jones and Mulligan, 2014;Wheeler et al., 2017), all experiments were conducted at the subgenera level (Pteropodus).

Experimental conditions
After four (Phase II) and six (Phase I) weeks of laboratory acclimation, rockfish and cabezon were divided among four PCO 2 /DO treatments (two PCO 2 levels × two DO levels) and exposed for a three-week period to assess both main and interactive effects of two environmental factors that change during a typical upwelling event. CO 2 -acidified treatments consisted of two levels, one representing Ambient PCO 2 conditions typically experienced off the California coastline (600 μatm PCO 2 ), and a second based on upwelling conditions and predicted High PCO 2 levels for the coast (1600 μatm PCO 2 ). The ambient PCO 2 -treatment was selected based on data from Central and Northern California Ocean Observing System (CeNCOOS) showing that ambient PCO 2 levels along Bodega Bay, CA can remain relatively high compared to other areas along the coast (450-650 μatm PCO 2 ), with spikes above 700 μatm, whereas the predicted High PCO 2 levels were selected based on the worst-case scenario Representative Concentration Pathway (RCP8.5) for climate change (IPCC, 2013). DO treatments included Normoxic (8 mg l −1 O 2 ) and moderately Hypoxic (4.5 mg l −1 O 2 ) conditions consistent with nearshore waters off the coast of California during upwelling (Frieder et al., 2012;Largier et al., 2015; data downloaded from Bodega Ocean Observing Node (BOON) at http://boon.ucdavis.edu.). Each PCO 2 /DO treatment combination was attained using a PCO 2 , air and N 2 gas-delivery system (see details in Supplementary Fig. S1; modified from Fangue et al., 2010). Mixed gas was continuously delivered to four 150-l reservoir containers (one per PCO 2 /DO treatment) where it was mixed vigorously with seawater until equilibrated. Reservoir PCO 2 / DO treatment water was continuously dripped at a rate of 16 l h −1 into three 38-l replicate tanks (per treatment) with n = 25 rockfish in each. For cabezon, PCO 2 /DO treatments were replicated twice with n = 4 fish in each 38-l replicate tank. In addition, PCO 2 /DO treatment gas was directly bubbled into each replicate tank. All fishes were measured for mass and standard length prior to being put in each tank to acquire a baseline body condition (estimated by Fulton's K; 100*[mass in g/standard length in cm 3 ]). During the experiment, rockfish tanks were fed 1 g of frozen, finely chopped squid daily, whereas cabezon tanks were fed 1 g every other day. All tanks were surrounded by opaque blinds and with separate flow-through, circulating water (16 l h −1 ) to ensure rockfish received no olfactory or visual cues from the cabezon.

Experimental seawater chemistry
Temperature (°C), salinity (ppt), and DO (mg l −1 ) of each replicate tank for rockfish and cabezon were recorded once daily using a handheld YSI handheld metre (YSI 85, Yellow Springs, OH USA). Total pH was measured from each replicate tank and PCO 2 /DO gas mixing reservoir container once every 48 h using m-cresol dye (Sigma-Aldrich, St. Louis, MO, USA) in a UV spectrophotometer (SOP 6b, Shimadzu, Columbia, MD, USA). Every four days, total alkalinity was measured using open-cell titration (T50 titrator, Mettler Toledo Inc., Columbus, OH, USA) of the PCO 2 /DO gas mixing reservoir containers (titrant was acquired from the Dickson Laboratory, Scripps Institute, La Jolla, CA, USA). Salinity was additionally measured in the titration samples using a conductivity metre (YSI 3100, Yellow Springs, OH, USA). PCO 2 values were calculated for each replicate tank using temperature, salinity, total pH and alkalinity in R (R Development Core Team, 2013; package seacarb, Gattuso et al., 2015) and are summarized in Table 1 for each experimental phase.

Enzyme activity assays
After 1 and 3 weeks of acclimation, three rockfish were sacrificed from each replicate tank (n = 9 per PCO 2 /DO treatment). One cabezon was sampled from each replicate only after 3 weeks in each exposure (n = 3 per PCO 2 /DO treatment). We euthanized fishes in tricaine methanesulfonate (MS-222), recorded length and weight to calculate body condition and isolated a section of skeletal muscle tissue that was immediately frozen in liquid nitrogen. Due to the small size of the rockfish (~3 cm), a fillet of muscle, containing white and red muscle was isolated for biochemical assays. Following the experiment, all fish samples were transported to the main campus of University of California Davis, Davis, CA for enzyme quantification. We measured ATP-generating enzyme activity of citrate synthase (CS), cytochrome c oxidase (COX) and lactate dehydrogenase (LDH) to determine if PCO 2 /DO treatments altered cellular metabolism and/or increased energetic costs. CS and COX are both enzymes involved in aerobic metabolism in mitochondria, whereas LDH found in the cytosol is an indicator of glycolytic anaerobic metabolism (Jayasundara et al., 2013).
Sample preparation and assays were conducted as described in Davis et al. (2017). Briefly, while COX was assayed fresh, CS and LDH were assayed on frozen supernatants. Each assay sample was run in triplicate, background activity rates were determined (CS and LDH) and enzyme activity rates are presented as μmol min −1 g fresh muscle weight (FW) −1 . All assays were conducted in a spectrophotometer (Synergy HT, BioTek, Winooski, VT, USA) at 14°C (temperature at the collection site) and 24°C, a 10-°C increase for later Q 10 calculations. Temperature of 14°C was maintained using an Isotemp BOD refrigerator chamber (Fisher Scientific, Waltham, MA, USA). was used to evaluate changes in mitochondrial plasticity and oxidative capacity (i.e. aerobic ATP-energy production or limitation) in response to high PCO 2 and low DO (Strobel et al., 2013), whereas LDH:CS was calculated to evaluate metabolic potential and efficiency (anaerobic and aerobic; Hamilton et al., 2017).

Rockfish behavior assays
Anti-predator responses including exploration and activity behaviors were assayed using modified methodology from Ferrari et al. (2011a). The main difference in methodology was Ferrari et al. (2011a) examined changes in fish antipredator responses in the presence of food, whereas the current study did not. Here, each fish was assayed individually in one of six, static 38 l (50 cm length × 25 cm width × 30 cm height) experimental tanks that were filled to 15 cm water depth of fresh seawater (14°C and 8.0 mg l −1 DO). The six tanks were placed next to each other in a row and surrounded by black plastic sheeting to reduce experimenter influence. An airline tube fixed to the inside of each experimental tank was used to inject cues via three syringes handled from outside of the black plastic sheeting. One syringe had a control cue of 10 ml seawater collected from the arena, and a second contained 10 ml of alarm cue. Nonpredator conspecific alarm cues have been shown to affect a large range of taxa . The alarm cue was freshly prepared as in Cripps et al. (2011) before each assay round by sacrificing a non-experimental rockfish and dorsoventrally scoring it seven times. Care was taken to ensure no organs were lacerated, only skin and muscle. The scored fish was then held over a beaker and rinsed with 10 ml of seawater six times, leaving 60 ml of alarm cue. This was split into six 10-ml batches for each of the six arenas. The third syringe was filled with 20 ml of seawater from the fish's arena before the trial began and was used to flush the airline tubing after injecting the control and alarm cues. At the beginning of the assay the test fish was netted out of its PCO 2 /DO treatment tank and placed into the experimental tank. The assay included three phases: (1) a 15-min acclimatization phase to reduce the effects of handling stress, during which the fish was free to swim throughout the experimental tank and get used to its surroundings, (2) a 10-min baseline behavior phase (i.e. seawater cue) and (3) a 10-min antipredator behavior phase (i.e. alarm cue). At the start of the baseline behavior phase, 10-ml seawater cue was injected into the tank at 5 ml s −1 followed by 10 ml of flush water. The seawater cue was injected to get a baseline behavior for each PCO 2 /DO treatment and to compare alterations in behavior before and after the alarm cue. After the 10-min baseline phase, the prepared 10-ml conspecific alarm cue was injected at 5 ml s −1 followed by another 10 ml of flush water. At the end of the anti-predator phase, each fish was removed with a net and placed into a separate 38-l tank, each aerated with the appropriate PCO 2 /DO treatment gas, in preparation for the predator-prey trials. The experimental tanks were emptied and cleaned multiple times with fresh seawater and set up for the next trial. At both 1 and 3 weeks acclimation time points, six behavior trials (with six individuals in each trial) took place each day for 3 days (n = 9 per PCO 2 /DO treatment per day, and n = 27 per PCO 2 /DO treatment each week) allowing 216 fish to be recorded in total. New fish were used at each of the 1-and 3-week time points. For each assay trial, fish were run at random (with respect to treatment) and baseline and anti-predator behavior phases were recorded using camcorders (520 line resolution QC with onboard IR) mounted above each of the six experimental tanks.
A 5-min period of each of the baseline and anti-predator behavior phases was played back on a computer monitor and analyzed manually by graphically applying a 5 cm × 5 cm squared-grid (50 grid squares total) to each tank to quantify the location of the fish over time. The first two minutes following the cue introduction was not included in this 5-min period to help control for physical disturbances from cue introduction such as airline tubing movement and air bubbles, but also the cue took >1 min to disperse through the arena (observed with dye injections). Each grid square was marked every time the fish entered into it and from this information two parameters were calculated: (1) area explored, defined as the number of novel squares entered (each square could only be counted once), and (2) activity, defined as the total number of times a fish crossed a grid line. To assess where rockfish were active within the arena, space use was quantified as the frequency of visits to each of the 50 arena squares (5 cm × 5 cm per grid square). To visualize potential effects of PCO 2 and DO on space use, the average visits within each grid was calculated and plotted for the baseline and antipredator phases at both 1 and 3 weeks acclimation time using the akima package in R for interpolation of gridded irregular spaced data (Akima et al., 2015). To test for a difference in space use before and after the alarm cue, four post-experimental zones were assigned to the arena including zones nearest the cue (5 × 25 cm 2 ), sidewalls (5 × 40 cm 2 ), centre (15 × 40 cm 2 ) and the farthest distance from the cue (5 × 25 cm 2 ) (see Supplementary Fig. S2 for zonation details).

Predator-prey trials
Six rockfish from each PCO 2 /DO treatment were randomly selected (with respect to the previous behavior trial) and transferred to a 180-l predation arena (91 cm length × 46 cm width × 43 cm height, with 42 cm water depth), with the respective PCO 2 /DO treatment water, and with one side of the arena having a standard volume of green ribbon anchored to a plastic rack to simulate submerged aquatic vegetation (Keller and Brown, 2008). A single cabezon, fasted for 24 h, was transferred to an affixed predator cage within the top corner of the predation arena. Both rockfish and cabezon were allowed 1 h to acclimate to the arena, after which the predator cage was opened and the cabezon were allowed to enter the predation arena with the rockfish. Each arena was covered with a black netted screen allowing for light penetration but minimizing the external disturbances. Predation trials lasted 24 h after which the surviving rockfish were counted, and predation mortality rate was calculated (number of fish eaten/total) as in Ferrari et al. (2015). All surviving fish of both species were removed, euthanized in MS-222 and length and weight were recorded for body condition index calculations. One predation trial for each PCO 2 / DO treatment (n = 4 total trials per day) was conducted each day for three days (n = 3 replicates per PCO 2 /DO treatment and time point). Predation trials were conducted after 1 and 3 weeks in experimental Phase II.

Statistical analyses
Statistical analyses were conducted in R (version 0.98.1103, R Development Core Team, 2013), with statistical significance set at P < 0.05. All datasets were examined graphically via residuals vs. fitted values, residuals vs. treatment and acclimation time, Q-Q plots for normal distributions, and residuals vs. day and tank to detect any effects of replicates.
A three-way analysis of variance (ANOVA) was used to evaluate the effects of elevated PCO 2 , decreased DO and acclimation time on cellular metabolism metrics (COX, CS, LDH, Q 10 s and enzyme ratios) in juvenile rockfish. Cabezon cellular metabolism metrics were only measured at a single time-point, and hence a two-way ANOVA was conducted to test the effect of PCO 2 and DO level on enzyme activities and Q 10 values. A post hoc Tukey's test ('lsmeans' package, Lenth, 2016)
To determine effects of PCO 2 , DO and acclimation time on fish body condition index a three-way ANOVA was conducted for rockfish (log transformed) and cabezon.
A three-way linear mixed effect model (lmer, nlme package, Pinheiro et al., 2015) was used to test the effects of PCO 2 , DO and the cue (seawater or alarm cue) on area explored and activity. Each lmer model included fish identification for repeated measures accounting for individual differences while also testing for treatment effects. A separate model was conducted for each acclimation time-point (1 and 3 weeks). An ANOVA table was generated for each lmer model, and significant F-statistics were followed by a Tukey's post hoc test (lsmeans) to determine specific differences in treatment, time and the behavioral response before and after the cue. To compare the change in space use (i.e. post-defined arena zones) after presentation of the alarm cue across PCO 2 and DO levels, a generalized least squares (gls) model was conducted to incorporate heterogeneity within the treatments into the model using the variance structure 'varIdent' with arena zone and PCO 2 /DO treatment combination as the fixed factors (nlme package, Pinheiro et al., 2015;Zuur et al., 2009). An ANOVA was then run on the gls model, and post hoc Tukey's test was used to detect differences between PCO 2 /DO treatment and the interaction between zones using the lsmeans package. A separate gls model was conducted for each acclimation time point (1 and 3 weeks).
Predation mortality rate (the proportion of fish consumed) was arcsine-transformed to normalize data and analyzed with a two-way ANOVA as in Ferrari et al. (2011b), followed by a post hoc TukeyHSD, with fixed factors being PCO 2 /DO treatment combination, and acclimation time. All fish (rockfish and cabezon) metrics were originally nested within their replicate tank (n = 3 per tank) and measurement day (n = 3) within the model; however, with no tank or day effect, the models were reduced to simplest form and fish were pooled for a single treatment group.
The ratio of aerobic enzymes COX:CS was significantly altered by PCO 2 (F 1,64 = 5.40, P = 0.02), but the effect depended on the DO treatment (Normoxic or Hypoxic; F 3,64 = 0.10, P = 0.75) with a significant interaction between PCO 2 and DO (F 1,64 = 5.74, P = 0.02, Fig. 2a). There was no effect of acclimation time or significant interactions between PCO 2 and DO treatments and acclimation time (P > 0.05, Table 2). Under Normoxic conditions COX:CS ratios were most similar, independent of PCO 2 treatment (P > 0.05); however, under Hypoxic conditions, the cooccurring High PCO 2 significantly increased COX:CS ratios (P < 0.05) compared to the single stressor of hypoxia alone. LDH:CS (anaerobic to aerobic) was significantly altered by acclimation time (F 1,64 = 15.65, P < 0.001), independent of PCO 2 (F 1,64 = 2.61, P = 0.11) and DO treatments (F 1,64 = 0.24, P = 0.63) (Fig. 2b). While there were no two-way interactions between factors (P > 0.05, Table 2), a three-way interaction at P = 0.068, indicated the main effect of acclimation time had some dependence on PCO 2 and DO treatment. For example, LDH:CS in fish exposed to single stressors (High PCO 2 or Hypoxia) significantly decreased (P < 0.05) after 3 weeks of acclimation time compared to initial levels at 1 week, but this decrease was not significant in the multiple stressor LDH:CS remained similar to 1 week. Temperature sensitivity of enzymes quantified by Q 10 values at assay temperatures of 14°C and 24°C showed only LDH was affected across acclimation time (F 1,64 = 5.67, P = 0.02) with no effects of PCO 2 or DO on Q 10 values for COX, CS, and LDH (P > 0.05, see Table 2 for all model results). Average Q 10 values for each PCO 2 /DO treatment are presented in Fig. 2c. Enzyme activity values measured at 24°C and Q 10 s for each acclimation time can be seen in Supplementary Tables S1 and S2.

Cabezon enzyme activity
Mean (± SEM) enzyme activities for cabezon predators at 3 weeks only are presented in Table 3. Overall, cellular metabolism in cabezon were unaffected by both single or multiple-stressors of CO 2 -acidifcation and hypoxia. Enzyme activities of COX, CS and LDH showed no effect of PCO 2 or DO treatments (P > 0.05, see model results in Table 3). Furthermore, COX:CS or LDH:CS ratios were unaltered by PCO 2 or DO treatments, nor Q 10 values CS and LDH enzymes; however, Q 10 values of COX were significantly decreased by elevated PCO 2 (P < 0.05). All model results and interactions are presented in Table 3.
(a) (b) Figure 3: Response change of juvenile rockfish exposed to PCO 2 and DO conditions for 1 and 3 weeks acclimation time. Mean ± SEM of (a) area explored and (b) total activity are shown after exposure to a seawater cue (baseline control) and a conspecific alarm cue (n = 17-24 per point). Coloured lines represent the behavior trend pre-and -post-alarm cue in Ambient PCO 2 /Normoxic (green circle), Ambient PCO 2 / Hypoxic (blue diamond), High PCO 2 /Normoxic (yellow square) and High PCO 2 /Hypoxic (red triangle). Lowercase letters indicate a significant difference in PCO 2 and DO treatments within each cue, bracketed letters indicate a significant effect of DO level (Hypoxia) across cues, and asterisks signify a change in behavior between the seawater and alarm cue within a specific treatment (P < 0.05). Alterations in space use following an alarm cue were observed in rockfish after acclimation to PCO 2 and DO treatments (Fig. 4). At 1 week of acclimation, a significant interaction (F 9,184 = 7.47, P < 0.0001) between arena zones and PCO 2 /DO treatment combination indicated activity differed by arena zone (F 1,184 = 15.73, P < 0.0001) depending on PCO 2 /DO treatment combination (F 3,184 = 16.40, P < 0.0001). Rockfish under Ambient PCO 2 (independent of DO level) increased activity in the zone farthest from the cue site, whereas rockfish under High PCO 2 , Normoxic and Hypoxic conditions showed little change in zone activity (lsmeans Tukey, P < 0.05, see zone assignments in Supplementary  Fig. S2). After 3 weeks of acclimation, a change in space use following the alarm cue occurred in all four treatments (F 3,184 = 8.51, P < 0.0001), characterized by a significant increase in activity in the zone (F 3,184 = 92.65, P < 0.0001) farthest from the cue site. We detected significant trends in zone alterations, in particular increased preference for cells at the wall farthest from the cue site (Supplementary  Table S3).

Discussion
In this study, we characterized the interacting effects of CO 2acidification (high PCO 2 ) and low DO (hypoxia) on cellular metabolism, behavior and predator-prey dynamics to assess whether rockfish living in seagrass beds have the ability to compensate physiologically or behaviorally to these naturally co-occurring stressors. Acute negative effects of elevated PCO 2 and hypoxia were detected in juvenile rockfish after a one-week acclimation. Predation mortality increased with elevated PCO 2 and hypoxia exposures, correlating with metabolic and behavioral adjustments. Following three weeks of acclimation, evidence of physiological and behavioral compensation was observed suggesting juvenile rockfish have some capacity to acclimate to elevated PCO 2 and hypoxia. The observed resilience of juvenile rockfish may be due

12
to their evolutionary history in an upwelling system, characterized by relatively frequent fluctuations in PCO 2 and DO.

Cellular metabolism
Rockfish showed trends of elevated metabolic costs under conditions of elevated PCO 2 and hypoxia through increases in LDH activity, an index of cellular anaerobic capacity (Fig. 1c). In contrast, CS and cytochrome c oxidase (COX) activity, indices of cellular aerobic capacity, remained relatively unaffected. Similar metabolic adjustments have been observed in the spot, a common estuarine fish (Leiostomus xanthurus) exposed to hypoxic conditions (0.8, 2 and 4 mg l −1 , Cooper et al., 2002), and juvenile Antarctic rockcod (Trematomus bernacchii) exposed to high levels of PCO 2 (~1200 μatm PCO 2 , Davis et al., 2017) where increases in anaerobic LDH activity was observed, independent of changes in CS activity. Although non-significant, several trends of rockfish enzyme activity may warrant further discussion. Subtle elevations in LDH and COX enzyme activity in response to the multiple-stressor treatment were demonstrated by the highest mean enzyme activities from fish in the High PCO 2 and low DO treatment at 1 week (Fig. 1a,c). These correlated increases in glycolytic (LDH) and respiratory chain activity (COX) may indicate cellular adjustments to maximize energy producing pathways and under the CO 2acidified and hypoxic conditions. Second, upward trends of COX and CS activities under hypoxic conditions (independent of PCO 2 level) were demonstrated by an increase in mean activities from 1 to 3 weeks acclimation (Fig. 1a,b). Increased aerobic activity in response to hypoxia may indicate rockfish have increased the quantity of CS and/or COX enzymes to increase aerobic energy production under low oxygen conditions. In contrast to rockfish, cabezon predators in the present study exhibited no alterations in activity of any of the three metabolic enzymes (i.e. COX, CS, or LDH) after acclimation to single and multiple stressor treatments of elevated PCO 2 and low DO (see Table 3).
Although CS activity in juvenile rockfish was unaffected by elevated PCO 2 and hypoxia, functionally the ratio of CS to other key enzymes has been shown to be more revealing of metabolic adjustments and cellular reorganization (Windisch et al., 2011). The synergistic interaction of increased PCO 2 and hypoxia on rockfish enzyme ratios supports the hypothesis that the ability to cope with a single stressor may be compromised if combined with another stressor, an idea that is supported by previous work (see review Pörtner, 2008;Davis et al., 2017). The highest mitochondrial (COX:CS) and glycolytic capacities (LDH:CS) were observed in rockfish under the multiple stressor treatment (high PCO 2 and hypoxia) and suggest that metabolic costs were likely greatest under this exposure. Stress-induced metabolic costs could include acid-base regulation and protection mechanisms of proteins (Sokolova et al., 2012(Sokolova et al., , 2013. Potentially, in efforts to provide sufficient energy supply (i.e. cellular ATP production) for increased metabolic demands, mitochondrial and glycolytic adjustments were made in juvenile rockfish after 1 week of acclimation to treatment exposures. Metabolic adjustments by altered LDH:CS have been described in juvenile rockfish that inhabit kelp forests; however, in contrast to the present study, adjustments followed a single stressor of extremely high PCO 2 (2800 μatm) and were observed after 5 months exposure . The overall reestablishment of cellular energetic pathways after 3 weeks of acclimation with no obvious impacts on body condition ( Supplementary  Fig. S3) suggests that juvenile rockfish in the present study have the capacity to compensate if given sufficient time.

Baseline behavioral alterations
The present study showed that absolute activity of rockfish was not affected by elevated PCO 2 and hypoxia (Fig. 3b). This could indicate that there was sufficient aerobic capacity to support any elevated energetic demands caused by either of the stressors, such that activity was not compromised. Unaltered swimming activity in response to elevated PCO 2 was similar to activity responses of number of other fishes exposed to high PCO 2 including juvenile splitnose rockfish (1125 μatm, Hamilton et al., 2014), juvenile spiny chromis (1000 μatm, Sundin et al., 2017), anemonefishes (Nowicki et al., 2011) and damselfish (Ferrari et al., 2012), but contrasts with previous findings that showed significantly increased activity in juvenile coral trout (960 μatm PCO 2 , Munday et al., 2013). Hamilton et al. (2017) showed negative effects of elevated PCO 2 on rockfish performance (behavioral lateralization, Ucrit, and aerobic scope), but only at extreme PCO 2 levels of >2000 μatm. Exploratory behavior in the current experiment, however, showed unique responses dependent on acclimation time and the interaction of DO level and PCO 2 treatment. After a 1 week exposure, fish under elevated PCO 2 (normoxia) explored more area compared to control fish, similar to exploratory and boldness behaviors documented in previous studies Ou et al., 2015). It is possible that increased exploration was a result of elevated PCO 2 reducing anxiety levels in fish (Hamilton et al., 2014). The increased exploration under elevated PCO 2 at 1 week was no longer present after 3 weeks such that exploration levels were more similar to the control. Therefore, there may be behavioral compensatory mechanisms that provide resilience to elevated PCO 2 levels (Clements and Hunt, 2015;Duteil et al., 2016). These rockfish may have adaptive mechanisms already in place to deal with high levels of PCO 2 indicative of their natural seagrass bed environment, which exhibit large fluctuations on PCO 2 . After 3 weeks, fish exploration was most affected by hypoxia (independent of PCO 2 level). Increased fish exploration under hypoxic conditions differed from our initial hypothesis that rockfish under hypoxic conditions would experience reduced activity and exploration behavior as an energetic conservation mechanism to reduce energy demand (Herbert and Steffensen, 2005). Increased exploration could be due to a motivation to escape low DO conditions (Herbert and Steffensen, 2005;Domenici et al., 2000); however, it is important to note that behavioral assays were conducted in normal seawater (not PCO 2 /DO treatments), and hence, elevated exploratory behavior could have been a response to normoxic conditions.

Anti-predator responses
The lack of change in fish activity following a conspecific alarm cue in any of the PCO 2 /hypoxic treatment conditions could suggest these juvenile rockfish lack the typical antipredator response. Following a conspecific alarm cue it would be expected for prey fish to decrease activity in the presence of a predator (freezing behavior) or possibly rapidly increase movement (for improved predator confusion or evasion, Couzin and Laidre, 2009;Kelley et al., 2011). In the current study, rockfish activity remained constant. In nature, these fish might respond to an alarm cue by actively seeking refuge (and then becoming inactive once in refuge), but the fact that they stayed active might reflect the lack of structural refuge in our test aquaria. While anti-predator responses were not evident in activity, exploratory behavior did reflect anti-predator responses. Rockfish exhibited a significant reduction in exploration when acclimated to single stressors of elevated PCO 2 and hypoxic conditions, which contrasts with some previous work showing elevated PCO 2 diminished predator recognition ability and thus prey maintained exploration levels after alarm cue exposure (Ferrari et al., 2011a;Lönnstedt et al., 2013). Importantly, however, fish exposed to a single stressor of elevated PCO 2 or hypoxic conditions showed an altered response compared to control fish, consistent with the GABA hypothesis (Tresguerres and Hamilton, 2017). In the control treatment, the lack of a reduction in exploration following the alarm cue may be related to the fact that the conspecific alarm cue was injected after the baseline behavioral phase, and therefore was confounded with time spent in the arena. Fish are expected to decrease exploration with time, and rockfish had already explored area during the baseline phase in addition to a 15min period provided after transferring the fish to the arena. However, it may also be the case, that rockfishes simply do not have a strong response to conspecific alarm cue.
The reduction in exploration following the alarm cue and the absence of any change in activity is better explained by looking at the space use analysis, detailing where the rockfish were active within the arena. Space use analyses showed there may be some short-term impairments to anti-predator responses after exposures to elevated PCO 2 with fish under control treatments shifting the location of activity to areas farthest from the cue site (Fig. 4). This same shift in spatial activity (i.e. away from the cue site) was also seen in fish acclimated to hypoxic conditions. This spatial shift under hypoxia following the alarm cue further explains the reduction in exploration but not activity under hypoxia. While fish may have increased activity in only one section of the tank, the number of new squares they explored would naturally decrease. In contrast, rockfish in the high PCO 2 (independent of DO treatment) after 1 week did not show a significant shift in space use following the alarm cue, which could be a result of impaired chemical detection mechanisms due to a malfunctioning of GABA A receptors (reviewed in Tresguerres and Hamilton, 2017). After 3 weeks of acclimation, fish in high PCO 2 showed a typical anti-predator response in spatial usage, as all treatment responses showed the expected pattern of an increase in activity in the area of the arena furthest from the alarm cue injection point. The return to a normal anti-predator response after 3 weeks of acclimation suggests rockfish have the capacity to compensate to some degree for elevated PCO 2 if given adequate acclimation time. More research is warranted to uncover the mechanisms driving acute behavioral changes of rockfish in response to environmental stressors and whether elevated PCO 2 interferes with brain ion-regulation, olfaction or vision shown in previous studies Nilsson et al., 2012, Heuer et al., 2016reviewed in Tresguerres and Hamilton, 2017). Assessments of behavior at finer spatial and temporal scales than traditional methods could allow for a more comprehensive understanding of how environmental stressors affect movement behavior (Ferrari et al., 2011a;Duteil et al., 2016).

Predator-prey interactions
The effects of elevated PCO 2 and hypoxia on predator-prey relationships appear to vary by species and are dependent on how both the prey and predator are affected (Ferrari et al., 2011b;Couturier et al., 2013;Midway et al., 2017;Cripps et al., 2011). The assessment of predator-prey relationships in the present study suggest prey (rockfish) and predators (cabezon) may be differentially affected by elevated PCO 2 and hypoxic conditions. Rockfish under control conditions (Ambient PCO 2 /Normoxia) appeared to have adequate antipredator escape responses after both 1 and 3 weeks of acclimation as none were consumed by cabezon (Fig. 5). Rockfish exposed to single stressors of elevated PCO 2 and hypoxia experienced increased predation mortality after 1 week and 3 weeks; whereas rockfish predation mortality returned to zero under the multiple stressor treatment (High PCO 2 /Hypoxia). Cabezon stopped feeding on rockfish in the multiple stressor treatment (High PCO 2 /Hypoxia), suggesting cabezon were negatively impacted by exposure conditions. Indeed, other predators exposed to elevated PCO 2 have displayed decreased activity (Rosa and Seibel, 2008), a reduction in their attraction to injured prey (Cripps et al., 2011) and alterations in prey selectivity (Ferrari et al., 2011b(Ferrari et al., , 2015. Our results provide evidence that the rockfish's ability to evade predation under elevated PCO 2 was decreased. Increased predation of rockfish under high PCO 2 may indicate potential 'trade-offs' in physiological and behavioral strategies (Careau and Garland, 2012;Davis et al., 2017). The significant predation mortality at 1 week under high PCO 2 does correlate with the observed adjustments in cellular metabolism (i.e. reliance on anaerobic mechanisms) as well as the trends in the increased exploration and activity data, suggesting increased energetic costs under elevated PCO 2 treatment. Furthermore, rockfish in high PCO 2 showed no shift in space use in response to the alarm cue compared to the control suggesting potential sensory impairments Nilsson et al., 2012;Ferrari et al., 2011a). These correlations in traits imply that physiological and behavioral trade-offs may have occurred, leaving rockfish less capable of avoiding predation. Metabolic and behavioral changes showed a degree of compensation after 3 weeks of acclimation, and predation mortality also decreased providing additional evidence these rockfish have the capacity to acclimate to elevated PCO 2 over 3 weeks of acclimation.
In conclusion, juvenile rockfish and cabezon represent a valuable study system in which to explore the interaction of multiple changes in environmental factors over various biological scales. Predation pressure is very strong in seagrass communities (Hindell et al., 2000), typically with high mortality of juvenile fish (relative to adults). Understanding the factors that determine juvenile rockfish survival during their time in seagrass beds is critical to understanding population dynamics. By combining physiology, behavior and predatorprey interactions this study described how three key interacting factors might be altered by elevated PCO 2 and low DO, influencing juvenile rockfish survival. Juvenile rockfish had the capacity to acclimate to elevated PCO 2 and low DO, such that short-term alterations in physiology and behavior were restored after 3 weeks of acclimation, which correlated with decreased predation mortality. Experiencing diel fluctuations in their natural seagrass bed environment, rockfish likely have adaptive strategies in place to cope with broad ranges of PCO 2 and DO levels, as seen in a range of other species (e.g. blue rockfish , shallow-shore crabs [Pane and Barry, 2007] and Pacific krill [Cooper et al., 2016]). Cabezon predators showed no alterations of cellular metabolism following elevated PCO 2 and low DO acclimation, but predation of rockfish ceased under multiple stressors indicating potential negative impacts on predator behavior. Studies that do not test both prey and predators under environmental stressors may misinterpret predator-prey interactions if differential species-specific sensitivities exist. Understanding how environmental stressors are integrated across biological scales, such as behavior and physiology in fishes, can be a powerful tool to identify trade-offs, assess vulnerability and project how community interactions may be altered under changing ocean conditions.

Supplementary material
Supplementary material is available at Conservation Physiology online.