Effects of elevated CO₂ on predator avoidance behaviour by reef fishes is not altered by experimental test water

Introduction

Rising concentrations of atmospheric carbon dioxide (CO₂) have caused an increased uptake of CO₂ by the ocean, leading to a decline in seawater pH and changes in the relative concentration of carbonate and bicarbonate ions, a process called ocean acidification (Sabine et al., 2004; Doney, 2010). The partial pressure of carbon dioxide (pCO₂) at the ocean surface is increasing at the same rate at the atmosphere (Doney, 2010) and thus marine species will need to deal with rising CO₂ levels in addition to declining pH and other changes in ocean chemistry (Portner, Langenbuch & Reipschlager, 2004). Recent models suggest an amplification of seasonal cycles in ocean pCO₂ as atmospheric CO₂ continues to rise, such that surface ocean pCO₂ will reach 1,000 μatm in summer for atmospheric CO₂ concentrations that exceed 650 parts per million. Hypercapnic conditions (>1,000 μatm CO₂) are projected to occur in up to half the surface ocean by 2100 on the current CO₂ emissions trajectory (McNeil & Sasse, 2016).

Elevated CO₂ levels can have a variety of effects on the physiology, life history and behaviour of marine organisms (Portner, Langenbuch & Reipschlager, 2004). Recent studies have demonstrated that high CO₂ levels can fundamentally alter the behaviour of marine fishes and some invertebrates (reviewed by Briffa et al., 2012; Clements & Hunt, 2015; Heuer & Grosell, 2014; Nagelkerken & Munday, 2016). Of particular concern is an impaired response to the threat of predation, because it may increase mortality rates in natural populations (Munday et al., 2010; Ferrari et al., 2011a; Chivers et al., 2014). Larval and juvenile reef fish can exhibit an innate aversion to chemical cues released by predators and chemical alarm cues from injured conspecifics, because they indicate a heightened risk of predation (Holmes & McCormick, 2010; Dixson, Pratchett & Munday, 2012). Individuals typically respond to these chemical cues by reducing activity and seeking shelter (Ferrari, Wisenden & Chivers, 2010). However, larval reef fish that have been reared at near-future CO₂ levels do not reduce activity or stop feeding in the presence of chemical alarm cues (Ferrari et al., 2011a; Lönnstedt et al., 2013) and even become attracted to chemical cues from a predator (called predator odour from herein) and chemical alarm cues when presented in a two channel flume (Dixson, Munday & Jones, 2010; Welch et al., 2014). Furthermore, juvenile fish exposed to high CO₂ lose their ability for associative learning (Ferrari et al., 2012), a process that enables fine-tuning of risk assessment by the association of chemical alarms cues with the identity of specific predators. Finally, high CO₂ affects the kinematics of predator–prey interactions, with juvenile prey exposed to elevated CO₂ allowing predators to get closer before responding (Allan et al., 2013) and exhibiting reduced escape speeds and distances compared with fish reared at current-day CO₂ levels (Allan et al., 2013; Allan et al., 2014). These changes in behaviour alter the outcome of predator–prey interactions, leading to significantly increased rates of mortality of small juveniles in mesocosm experiments (Ferrari et al., 2011b) and in fish transplanted to natural coral reef habitat (Munday et al., 2010; Munday et al., 2012; Chivers et al., 2014).

Many of the pioneering studies on the effects of elevated CO₂ on reef fish behaviour (e.g., Munday et al., 2009; Dixson, Munday & Jones, 2010; Ferrari et al., 2011a; Ferrari et al.,2011b; Cripps, Munday & McCormick, 2011) involved rearing fishes in control and elevated CO₂ conditions and then testing their responses to chemical cues in control water only. This method was chosen in these early experiments owing to logistical constraints and because pilot experiments showed that the response of high-CO₂ exposed fish to predator cue did not differ if the cue was presented in either control or high CO₂ water (Dixson, Munday & Jones, 2010; Munday et al., 2010). Furthermore, studies with freshwater fishes had shown that a pH reduction of 0.5 units in freshwater (<pH 6.5) irreversibly alters the structure of chemical alarm cues and can dramatically affect prey antipredator responses (Brown et al., 2002; Leduc et al., 2004; Leduc, Kelly & Brown, 2004; reviewed in Leduc et al., 2013). Consequently, there was concern that testing marine fish in CO₂-acidified water (albeit still above pH 7.5) could confound the interpretation of any effects of high CO₂ on the behavioural response of the fish with a diminished efficacy of the chemical cue itself. Testing in control water prevented this potential problem. Finally, small-scale laboratory experiments often have limited ecological relevance and it is challenging to extrapolate their findings to natural communities and ecologically relevant spatial scales (Nagelkerken & Munday, 2016). Consequently, it was critical to test the effects of exposure to high CO₂ in natural habitats in the field. This involved transplanting larval fish that had been reared in either control or high CO₂ seawater back into natural coral reef habitat and monitoring their behaviour (Munday et al., 2010; Munday et al., 2012; McCormick, Watson & Munday, 2013).

Since these initial studies, a number of other experiments have been conducted in which the respective treatment seawater in which the fish were reared (control or high CO₂) has been used in the experimental trials, confirming the results of earlier studies (e.g., Allan et al., 2014; Dixson et al., 2015; Welch et al., 2014; Nagelkerken et al., 2016). Nevertheless, a more thorough validation study of the test water used in previous antipredator experiments is warranted. Here, we compared the use of control versus high CO₂ seawater (called test water from hereafter) in experimental trials designed to test the antipredator responses of larval reef fishes reared in high CO₂. We conducted antipredator experiments for two species that have been widely used in previous high CO₂ experiments, the clownfish Amphiprion percula and the ambon damselfish Pomacentrus amboinensis. Specifically, we tested if: (1) using control or high CO₂ water in a two channel flume influenced the response of larval clownfish to predator odour in clownfish reared in either control or high CO₂; and (2) using control or high CO₂ water in the test arena influenced the kinematic response of larval damselfish to a startle stimulus in fish reared in either control or high CO₂. Finally, (3) because the effects of high CO₂ on fish behaviour appear to be due to altered function of GABA-A neurotransmitter receptors (Nilsson et al., 2012; Hamilton, Holcombe & Tresguerres, 2014; Heuer & Grosell, 2014) we tested if antipredator behaviours were restored in clownfish treated with a GABA antagonist (gabazine) in high CO₂ water. Previous studies have demonstrated that abnormal antipredator behaviour of clownfish and damselfish reared in high CO₂ conditions is reversed following treatment with gabazine (Nilsson et al., 2012; Chivers et al., 2014). However, these previous studies involved high CO₂ fish that were treated with gabazine in control water and then behaviourally tested in control water. In this experiment we therefore conducted the same gabazine treatments as previously used, except the drug was administered and behaviour was tested in high CO₂ water.

Materials and Methods

The experiment with larval clownfish was conducted at James Cook University’s (JCU) experimental aquarium facility in Townsville, Australia, where previous high CO₂ experiments with this species have been conducted. The experiment with larval ambon damselfish was conducted at Lizard Island Research Station (LIRS) on the northern Great Barrier Reef, where previous high CO₂ experiments with this species have been conducted. Both experiments were conducted during February 2016. Because our goal was to assess the use of control seawater in experimental trials of previous experiments (e.g., Dixson, Munday & Jones, 2010; Munday et al., 2010; Munday et al., 2012) we used the same methods as previous experiments, except that we applied a fully factorial design where fish reared in control and high CO₂ conditions were behaviourally tested in both control and high CO₂ water. Control water was the ambient in the experimental facility at JCU (496 ± 55 S.D. μatm CO₂) and at LIRS where water is pumped directly from the Lizard Island lagoon (464 ± 32 μatm CO₂). The high CO₂ conditions (1,022 ± 37 and 1,051 ± 18 μatm CO₂, respectively) matched projections for CO₂ levels in open ocean by the end of this century (McNeil & Sasse, 2016).

Data Analysis

Two-way ANOVA was used to compare percent time that larval clownfish spent in predator cue in the flume. Percentage data was logit-transformed for analysis (Warton & Hui, 2011). Treatment water (control or high CO₂) and test water (control or high CO₂) were the fixed factors. ANOVA was conducted with Statistica (version 13). For the gabazine experiment, the percentage time that larval clownfish treated with gabazine spent in predator cue in the flume was compared with control fish using a t-test on logit-transformed data. Linear mixed effects models (LME) were used to compare escape responses of larval damselfish. Latency, escape distance, escape speed and maximum speed were each tested separately. Treatment water (control or high CO₂), test water (control or high CO₂) and day (1 or 2) were fixed factors in the model. The distance of each fish from the point the stimulus hit the water was also included as a fixed effect in the model to account for any variation associated with differences in the location of fish in the chamber when the stimulus occurred. Tank was included in each model as a random effect to account for the subsampling of fish from replicate tanks within each CO₂ treatment. Best-fit model selection based on AIC values was automated using ‘MuMIn’ (Bartoñ, 2015) within the ‘nlme’ package in the R software package (Pinheiro et al., 2013). The normality, linearity and homoscedasticity of residuals of the models were verified by visual inspection of residual-fit plots. Latency data was log10 transformed before analysis to improve the distribution of the data and a heterogeneous variance structure was used in the model for latency (but not other variables). R code is available in Supplementary Information 1.

This research was conducted in accordance with James Cook University animal ethics guidelines and permits A1961, A1973 and A2080. Fish were collected under permit G12/35117.1 from the Great Barrier Reef Marine Park Authority.

Results

There was a highly significant effect of treatment water on the percentage time that larval clownfish spent in the water stream containing predator odour (F_1,57 = 125.91, p < 0.001), but no effect of test water (F_1,57 = 0.20, p = 0.66) and no interaction (F_1,57 = 0.06, p = 0.81). Laval clownfish reared in control water strongly avoided predator odour, spending <9% of the time in that water stream (Fig. 1). In contrast, larval clownfish reared in high CO₂ water were strongly attracted to predator odour, spending nearly 80% of the time in that water stream (Fig. 1).

As with the response to predator odour, treatment water had a significant effect on escape responses of larval damselfish, and there was no effect of test water. Distance to stimulus did not affect any of the response variables and was not retained in any of the final models. Latency to respond significantly increased in fish reared in elevated CO₂ (F_1,6 = 7.76, p = 0.032), with fish displaying latencies that were 30% longer compared with fish reared in control water (Fig.2A). There was no effect of test water or day of testing on latency to respond and these factors were not retained in the final model. Escape distance following stimulation was not significantly different for fish reared in high CO₂ compared with control water (F_1,5 = 6.15, p = 0.056) although there was a clear tendency for a shorter escape distance in high CO₂ fish (Fig. 2B). Day was retained in the final model, but it was not statistically significant (F_1,5 = 5.15, p = 0.07). There was no effect of test water on escape distance and it was not retained in the final model. Escape speed and maximum speed responded similarly with both kinematic variables declining significantly in fish reared in high CO₂ compared to fish reared in control water (Fig. 2C; F_1,5 = 6.81, p = 0.048 and Fig. 2D; F_1,6 = 9.016, p = 0.0239). Day was retained in the final model for escape speed, but it was not statistically significant (F_1,5 = 3.64, p = 0.115). Day was not retained in the final model for maximum speed. Again, there was no effect of test water for either escape speed of maximum speed and it was not retained in the final model.

Gabazine reversed the effects of rearing in high CO₂ (Fig. 1). Fish reared in high CO₂ and then treated with gabazine exhibited a strong avoidance of predator cue, with the response almost identical to fish reared and tested in control water (t₂₂ = 0.04, p = 0.96).

Discussion

Previous studies have demonstrated that CO₂ levels projected to occur in the surface ocean by the end of this century (700–1,000 μatm) impair predator avoidance behaviours in larval reef fishes (reviewed in Nagelkerken & Munday, 2016). However, some of these previous studies have tested the behaviour of high-CO₂ reared fish using control water in the experimental trials, which could potentially influence the results if there is an immediate physiological response associated with the transfer of fish from high CO₂ rearing water to control test water. Here we show that the results of these earlier studies are robust and that antipredator behaviour is not affected by using control water in the experimental trials. Larval fish reared in high CO₂ exhibited the same responses to predator odour in a two channel flume and the same kinematic responses to a startle stimulus regardless of whether high CO₂ or control water was used in the experimental trial. Similarly, there was no difference in the behavioural response of fish reared in control seawater when tested in either high CO₂ or control water. These results demonstrate that the choice of test water has not influenced the results of previous laboratory studies into the effects of high CO₂ on antipredator behaviour.

Our results are consistent with preliminary observations carried out in previous studies showing that the test water (control or high CO₂) used in a two-channel flume did not alter the behavioural response of larval fishes to predator odour (Dixson, Munday & Jones, 2010; Munday et al., 2010). Furthermore, our study extends these findings to the kinematics of predator avoidance, by showing that the type of water used in the test arena does not affect vital components of the escape response, including latency to respond, escape speed and maximum speed. Together these results show that antipredator behaviours of larval fishes that have been reared in high CO₂ are not compromised by a return to control seawater for experimental testing. These findings are also relevant to field-based studies that have tested the effects of high CO₂ on reef fishes in their natural habitat (e.g., Munday et al., 2010; Munday et al., 2012; Devine, Munday & Jones, 2012; Devine & Munday, 2013; Ferrari et al., 2011a; McCormick, Watson & Munday, 2013). Such studies are critical for predicting the impacts of high CO₂ on reef fish populations; however, they involve rearing larval fish for 4–5 days in high CO₂ conditions before transferring them to ambient CO₂ conditions in the field. These studies have demonstrated mortality rates from predation are higher in juvenile fish that have been exposed to high CO₂ compared with fish exposed to ambient control conditions (Munday et al., 2010; Munday et al., 2012; Ferrari et al., 2011a). Our results indicate that such studies are unlikely to have been affected by any immediate effect on antipredator behaviour caused by transplanting fish from high CO₂ conditions in the laboratory to ambient CO₂ conditions in their natural habitat. Nevertheless, fish in the field studies also remained in control conditions for a much longer period of time (several days) compared with the laboratory tests conducted here (minutes) and further experiments of longer duration would be required to ascertain any differences in control and high CO₂ conditions that might accrue over a longer period of time.

In contrast to the absence of a test water effect on antipredator behaviours, we found that rearing treatment water had a highly significant effect on antipredator behaviour. Consistent with previous studies (Dixson, Munday & Jones, 2010; Munday et al., 2010), larval clownfish reared in high CO₂ exhibited a reversal from strongly avoiding predator cue (<10% of time in predator cue in the flume) to a strong attraction to predator cue (nearly 80% of time in predator cue in the flume). Similarly, larval damselfish exposed to high CO₂ for 4 days exhibited a slower response to a threat stimulus and a slower escape speed compared with fish kept in ambient control water. There was also a tendency for a shorter distance travelled. These findings are consistent with changes in escape responses that have has been reported previously (Allan et al., 2013; Allan et al., 2014) and provide further confirmatory evidence for the effects of high CO₂ on antipredator behaviour in larval reef fishes. The differences in escape speed between high CO₂ and control fish were relatively small; however, when combined with a much longer latency to respond (>30% increase in time to respond) these changes in kinematic responses would likely have a detrimental effect on the probability of a prey fish escaping a predator attack.

Altered behaviour of fish in a high CO₂ environment appears to be due to the effects of acid-base regulation on the function of GABA-A receptors (Nilsson et al., 2012; Hamilton, Holcombe & Tresguerres, 2014; Heuer & Grosell, 2014). Fish prevent plasma and tissue acidosis in a high CO₂ environment by accumulating HCO${}_3^{-}$ in exchange for Cl⁻ (Brauner & Baker, 2009). The GABA-A receptor is a gated ion channel that conducts HCO₃⁻ and Cl⁻ when activated by the neurotransmitter GABA. The altered concentration of HCO${}_3^{-}$ and Cl⁻ in fish exposed to high CO₂ could decrease or reverse the flow of these ions through the GABA-A receptor, causing a reversal in receptor polarization (Heuer & Grosell, 2014). Evidence for the role of GABA-A receptors in behavioural impairment of fish in high CO₂ has come from studies showing that treatment with gabazine, a GABA antagonist, reverses the effects of high CO₂ on behaviour (Nilsson et al., 2012; Chivers et al., 2014; Chung et al., 2014; Lai, Jutfelt & Nilsson, 2015; Ou et al., 2015; Regan et al., 2016). In the first of these studies, Nilsson et al. (2012) found that gabazine reversed the impaired response to predator odour exhibited by clownfish that had been reared in high CO₂. However, in that study gabazine was administered in control seawater and high-CO₂ reared fish were tested in control water. Therefore, the effects of gabazine in restoring the response of larval fishes to predator odour has not been previously tested in high-CO₂ conditions. Here we show that treatment with 4 mg l⁻¹ gabazine completely reverses the abnormal behavioural of high-CO₂ fish to predator odour, restoring their natural aversion to predator odour. This contributes to a large number of independent studies pointing to a key role of GABA-A receptors in the behavioural impairment of fish exposed to high CO₂.

While we found no effects of using control or high CO₂ water in experimental trials to test the effects of high CO₂ on fish behaviour, Sundin & Jutfelt (2016) reported differences in relative lateralization of juvenile goldsinny wrasse when tested in control water versus high CO₂ water after 21 days in high CO₂. Their study did not include the reverse treatment, where control reared fish were tested in high CO₂ water, and there is the potential for learning effects in control fish because the same fish were tested on three different occasions, at days 9, 19 and 21. Nevertheless, that study does suggest that the response of some behavioural traits could be sensitive to differences in the test water used in behavioural trials. Further experiments with a broader range of behavioural traits and using experiments that are not complicated by the repeated testing of the same individuals are required to test this possibility. Furthermore, there are likely to be substantial species-specific differences in the response of fish to control or high CO₂ water. For example, Jutfelt & Hedgärde (2013) found that juvenile Atlantic cod were able to distinguish between low and high CO₂ water and exhibited a strong preference for low CO₂ water. Here, we only tested two species of damselfish, but other species might exhibit different responses. Indeed, there is considerable interspecific variation in the sensitivity of damselfishes to high CO₂ (Ferrari et al., 2011a), therefore, it is possible that there may also be species-specific differences in their antipredator responses depending on the test water used in the experimental arena.

Pioneering studies into the effects of high CO₂ on fish behaviour used control seawater in experimental trials due to logistical constraints and concerns about the possible effects of low pH on olfactory cues. Our results confirm the results of those initial studies by showing that test water (control of high CO₂) had no effect on the antipredator responses exhibited by fish reared in either control or high CO₂ conditions. Nevertheless, we do not advocate the continued use of control seawater for testing the behavioural effects of high CO₂ in marine organisms. Methods for treating seawater to replicate future pCO₂ levels are now sufficiently validated that it is possible to treat most experimental arenas with the respective treatment water that the animals have been reared in, and there may be some behavioural traits that are sensitive to different test water in the experimental trials. The most reliable and unambiguous results will likely come from using the respective treatment water in experimental trials testing the effects of high CO₂ on the behaviour of marine organisms. Future studies also need to consider the potential for adaptation of behavioural responses to projected future CO₂ levels (Sunday et al., 2014). Individual variation in behavioural tolerance to higher CO₂ levels (Munday et al., 2012) could provide the opportunity for selection of more tolerant genotypes and improved performance in a future high CO₂ environment.

Supplemental Information