Chronic warm exposure impairs growth performance and reduces thermal safety margins in the common triplefin fish (Forsterygion lapillum)

Understanding the physiological responses of organisms to temperature change and the limits of thermal tolerance will play a key role in determining a species’ susceptibility or resilience to climate change (Somero, 2012). It has recently been suggested that studies investigating the effects of climate warming may underestimate the impacts on populations if physiological tolerances to environmental extremes are not included, and if only resting physiological functions are considered (Sandblom et al., 2016). The need to examine the impacts of warming on chronic and acute thermal tolerance collectively may also be particularly important in species that occupy habitats that are exposed to large acute temperature fluctuations, such as those within marine intertidal zones (Nilsson and Lefevre, 2016). As the climate warms, these species have to tolerate not only chronic increases in mean ambient sea temperature, but also greater extremes of acute temperature change owing to the forecasted increase in the frequency and severity of transient heat waves (Meehl and Tebaldi, 2004; Perkins et al., 2012). Despite the universal occurrence of rock pools along coastlines, and that predicted changes in climate will likely impact these environments, there is a surprising lack of research regarding both the chronic and acute thermal tolerance of intertidal biota such as fish.

A potential impact of climate change is that chronic warming will demand changes in aerobic metabolism and cause adjustments to essential life history traits such as growth, reproduction, foraging and locomotion, which will lead to changes in fitness (Pörtner and Knust, 2007; Pörtner and Farrell, 2008). This theory, known as oxygen and capacity limited thermal tolerance (OCLTT), has emerged as a prominent hypothesis linking the thermal limits of metabolism to impacts of climate warming on multiple levels of biological organisation in aquatic ectotherms (Clark et al., 2013; Nilsson and Lefevre, 2016; Pörtner, 2010). Under the OCLTT hypothesis, the availability of aerobic metabolic scope (MS; the degree to which an organism can increase oxygen consumption above basal requirement) is the primary determinant and window of a species’ thermal tolerance. This thermal window encompasses an optimum temperature where MS peaks (T_optMS), upper and lower pejus temperatures that mark the onset of MS decline, and more severe upper and lower critical temperatures where MS is abolished and survival time is limited (Pörtner and Farrell, 2008). In the case of warming, MS is thought to decline at supra-optimal temperatures, because the capacity of the cardiorespiratory system is not capable of increasing maximum oxygen uptake (Ṁ_O2,max) to a rate that keeps pace with increasing basal energetic requirements (Pörtner and Farrell, 2008; Pörtner, 2010). Crucially, it is commonly stated that MS and performance measures such as growth, reproduction and locomotion are causally linked (Claireaux and Lefrancois, 2007; Clark et al., 2013; Lefevre, 2016; Pörtner and Knust, 2007; Pörtner and Farrell, 2008; Pörtner, 2010; Schulte, 2015). Despite initial enthusiasm for the OCLTT hypothesis, its general applicability in explaining thermal tolerance and its value in evaluating the potential impacts of climate change have become increasingly questioned (Clark et al., 2013; Nilsson and Lefevre, 2016). A recent meta-analysis (Lefevre, 2016) has demonstrated that not all marine ectotherms display the expected ‘increase–optimum–decrease’ temperature response of MS that is a central pillar of the OCLTT hypothesis and that species of interest require assessment on a case-by-case basis to examine the temperature dependence of organismal performance and fitness (Nilsson and Lefevre, 2016). Other studies have also found that the optimum temperature for performance and MS are not always matched (Grans et al., 2014; Norin et al., 2014) and, overall, there are actually few studies linking key performance metrics such as growth to MS under chronic thermal exposures (Clark et al., 2013).

The critical thermal maximum (CT_max) is the temperature at which an organism loses equilibrium during acute thermal ramping and is used as a measure of upper thermal tolerance in fish (Beitinger et al., 2000; Mora and Maya, 2006). Fish occupying rock pools may already experience acute exposure to temperatures exceeding critical limits, and climate change will likely increase temperature extremes further as heat waves are predicted to become more extreme (Perkins et al., 2012). In fish, warm acclimation can elevate upper thermal tolerance (Beitinger et al., 2000; Drost et al., 2016; Fangue et al., 2006), which indicates that species’ physiological plasticity may buffer thermal limits as the climate warms. However, evidence suggests that this capacity to elevate the upper thermal tolerance point is likely limited in the context of predicted temperature rises associated with climate change (Gunderson and Stillman, 2015). Moreover, the increase in CT_max with warm acclimation so far has always been less than the difference between two acclimation temperatures (Gunderson and Stillman, 2015). Consequently, thermal safety margins (the difference between acclimation temperature and CT_max) are reduced with warming (Sandblom et al., 2016). Although rock pools are among the most likely habitats where the CT_max of fish is exceeded (and this risk increases with climate change), we are unaware of any study that specifically addresses the acclimation capacity of CT_max in intertidal fish in the context of climate change.

The upper critical temperature (T_critMS), not to be confused with CT_max, is incorporated into the OCLTT hypothesis as the temperature at which MS becomes zero (Pörtner, 2010). Because warm acclimation increases the upper thermal tolerance of fish, this should also be associated with metabolic adjustments, which increase or maintain MS at high temperatures. Indeed, under the OCLTT hypothesis, improved upper thermal tolerance in warm-acclimated organisms is proposed to result from downregulation of mitochondrial function, which decreases resting oxygen demand and protects MS during warming (Pörtner, 2001). Conversely, cold-acclimated organisms have higher mitochondrial densities to maintain capacity for MS and adenosine triphosphate (ATP) production under cold temperatures, but this also increases oxygen demand during warming and decreases upper thermal tolerance (Fangue et al., 2009; Pörtner, 2001). Thus, at an equivalent elevated temperature, a thermally tolerant warm-acclimated ectotherm would have a lower metabolic rate than a less tolerant individual acclimated to cooler temperatures (Fangue et al., 2009).

Forsterygion lapillum Hardy 1989 is a small (20–60 mm) benthic marine blennioid fish that occupies shallow subtidal reefs, harbours and intertidal rock pools of New Zealand. This species maintains MS across a thermal acclimation range of 15–25°C and shows a surprisingly high T_optMS of 24°C (Khan et al., 2014a). Under the OCLTT framework, a T_optMS of 24°C would indicate that whole-animal performance may be optimised at a temperature beyond the present-day range, but this seems questionable because significant mortality (50%) occurs upon acclimation to 25°C, and skeletal muscle mitochondrial ATP production capacity shows signs of compromise at 24°C (Khan et al., 2014a). As an in-depth understanding of temperature-specific performance is not known for this species, the possibility of optimised performance at temperatures beyond present-day conditions remains unclear.

As F. lapillum occupies intertidal rock pools and shallow subtidal habitats, this study set out to critically evaluate the chronic and acute thermal tolerance of F. lapillum after 12 weeks of thermal acclimation to 15, 18, 21 and 24°C. To critically evaluate the growth performance and resilience of F. lapillum to climate change, the mass-specific growth rate (SGR) was first measured at the four acclimation temperatures. It was hypothesised that the fitness performance of the 24°C group [i.e. fish chronically exposed to temperatures ∼3°C above the current mean peak summer sea surface temperature (SST) in northern New Zealand] would be negatively impacted. As part of this testing, MS at the four temperatures was also measured and compared against growth to determine whether growth performance and metabolic performance correlated across an extended time frame as expected under the OCLTT framework. Combining archival SST records with the temperature–growth data of F. lapillum then provided the opportunity to model the seasonal growth impact of future warming scenarios within the northern geographical range of its distribution. Following the growth trial, the acute thermal tolerance (CT_max) of thermally acclimated F. lapillum (15, 18, 21 and 24°C) was then assessed to gauge whether plasticity in upper thermal tolerance exerts any influence over the ability of F. lapillum to withstand short-term extremes in temperature during a thermal ramping test. We hypothesised that higher acclimation temperatures would increase upper thermal tolerance (Beitinger et al., 2000) and that this would be associated with a lower metabolic rate on thermal ramping, as predicted by the OCLTT hypothesis (Fangue et al., 2009).

Wild F. lapillum were caught using bait catcher traps and transferred to the Leigh Marine Laboratory (Leigh, New Zealand), where they were housed in 30 l flow-through ambient temperature seawater tanks (14.7±0.16°C, air saturated, 200 μm filtered, 35 ppt salinity) for a period of 14 days to allow laboratory acclimation. These fish had an initial total mean (±s.e.m.) length of 55.3±0.55 mm and mass of 1.53±0.03 g and were sexually mature 1- to 2-year-old fish based on length-at-age determinations for this species (P. E. Caiger, unpublished). During this acclimation period they were fed daily to satiation on a mixture of crushed aquaculture pellets (Skretting, Australia) and chopped pilchard. Animal ethics approval was granted from The University of Auckland Animal Ethics Committee (AEC approval 001801).

where m₁ is the initial mass at the start of the growth period t₁ and m₂ is the final mass at the end of the growth period t₂. The overall growth rate of individual fish was taken as the mean SGR of the three 30 day growth periods. Throughout the growth trial each tank was fed daily to satiation on a mixture of crushed aquaculture feed (Skretting) and chopped pilchard. Ammonia, nitrate and nitrite were regularly measured in each tank and water quality remained high throughout the trial.

At the conclusion of the growth trial, metabolic rate and acute thermal tolerance were then assessed in all individuals. Metabolic and thermal tolerance measurements were carried out in an automated respirometry set-up that could monitor four fish simultaneously in individual chambers (∼20 h per run). All trials were completed over 35 days and, to avoid confounding handling effects on metabolic measurements, growth measurements were not assessed during this period. The temperature treatment for each set of four fish was assessed in a cycled order so that the period between the conclusion of the 12 week growth trial and assessment of metabolic rate and upper thermal tolerance was as similar as possible across treatment groups.

where V is the respirometry chamber minus fish volume, Δ%S/t is is the change in percentage air saturation of oxygen per unit time, α is the solubility coefficient of oxygen (mg O₂ %S⁻¹ l⁻¹) in seawater (35 ppt) and M_b is the body mass of the fish in grams (Schurmann and Steffensen, 1997). The repeated measurement cycles were interspersed by automated flushing periods (1 min) so that the water within the chamber was refreshed and maintained at an O₂ concentration above 80% air saturation.

Metabolic measurements were completed in a sequence where Ṁ_O2,max was assessed after exhaustive exercise; this was then followed by measures of standard metabolic rate (SMR) during an overnight period of respirometry. Ṁ_O2,max was determined following exhaustive exercise where fish were continually chased by hand for a period of 5 min. The chasing was carried out in a 30 l tank filled with air-saturated seawater maintained at the required experimental temperature. The fish were visibly exhausted after chasing and this method was used as it has been previously used to elicit maximal Ṁ_O2 values in F. lapillum (Khan et al., 2014a) and is the most reliable method for obtaining Ṁ_O2,max in benthic species that do not swim continuously in a flume (Clark et al., 2013; Norin and Clark, 2016). The fish were then transferred to a respirometry chamber within 30 s of the conclusion of chasing and repeating 4 min Ṁ_O2 measurement cycles were initiated. Ṁ_O2,max was taken as the highest Ṁ_O2 value recorded in any measurement cycle, which in almost all cases was obtained from the first measurement cycle following exhaustive exercise. When the metabolic rate of fish was clearly declining, the measurement period was extended to 8 min and Ṁ_O2 was measured repeatedly for >16 h. During overnight Ṁ_O2 measurements, fish were left undisturbed in the respirometers and SMR was estimated from the mean of the lowest 10% of Ṁ_O2 over this time (Khan et al., 2014a; Norin et al., 2014). MS was defined as the difference between the body-mass-corrected Ṁ_O2,max and SMR (see below) for each individual. In all estimates of metabolic rate, only Ṁ_O2 values with R²>0.95 for the decline in oxygen per unit of time were used.

Following SMR estimation, fish were subjected to thermal ramping to determine CT_max. In all cases, thermal ramping for an individual fish began from the experimental temperature a particular fish had been acclimated to in the growth trial phase of the study. The temperature of the seawater reservoir was increased at a rate of 4°C h⁻¹ and water temperature was continuously measured using a shielded temperature probe connected to a FireSting O₂ system (PyroScience, Aachen, Germany). After every 1°C increment in temperature, the routine Ṁ_O2 (RMR) of fish was measured over a measurement cycle length of 3–5 min. Thermal ramping and Ṁ_O2 measurements continued until the point at which fish lost equilibrium for 10 s or more, which was taken as the CT_max (Beitinger et al., 2000; Brijs et al., 2015; Sandblom et al., 2016)_. The fish was then removed from the respirometer and transferred to air-saturated ambient temperature seawater (∼21°C) for recovery. The maximum routine metabolic rate during thermal ramping (RMR_maxTR) was taken as the highest Ṁ_O2 measurement obtained at any of the 1°C temperature increments prior to reaching the point of CT_max. The thermal safety margin for each fish was also calculated and was defined as the temperature difference between CT_max and the acclimation temperature (Sandblom et al., 2016).

Background oxygen consumption was measured immediately after thermal ramping. Pilot trials showed that background oxygen consumption remained negligible throughout the overnight SMR measurement cycles but developed linearly during thermal ramping. A linear regression was therefore used to back-calculate the background oxygen consumption values throughout thermal ramping and these were subtracted from fish Ṁ_O2. The mean background oxygen consumption at 15, 18, 21 and 24°C was 3.2±0.72, 3.4±0.5, 3.9±0.64 and 4.7±0.6% of fish Ṁ_O2 during thermal ramping, respectively.

where Ṁ_O2,2.5g is the Ṁ_O2 for a fish with the standardised (corrected) new mass of 2.5 g, Ṁ_O2,meas is the measured Ṁ_O2, m is the mass of the fish, m_2.5g is the standardised body mass of fish set to 2.5 g and A is the mass exponent describing the relationship between metabolic rate and body mass (Schurmann and Steffensen, 1997). The mass exponents (A) describing the relationship between body mass and Ṁ_O2 for SMR and Ṁ_O2,max were 0.75 and 0.65, respectively, and were derived from a dataset of triplefin Ṁ_O2 measurements collected across multiple studies (T.J.M., unpublished data). The values for these mass exponents are within the expected range for fish (Clarke and Johnston, 1999).

In all statistical tests, significance was accepted at P<0.05. Mixed-model analysis was used to perform a nested analysis of covariance and assess the effect of acclimation temperature on growth rate (SGR) following the model fitting procedure outlined by Zuur et al. (2009). The final model included initial fish mass as a covariate, temperature as a fixed main effect and, to control for possible tank effects, individual tanks as fixed random effects. The model assumptions of normality and homoscedasticity were satisfied and specific post hoc comparisons were carried out using Tukey contrasts. The relationship between SGR and temperature was also assessed with a second-order (non-linear) polynomial regression. To examine how climate warming may impact annual patterns of growth, the relationship between temperature and SGR derived from the laboratory-based growth trial (Fig. 1B) was used to calculate projected growth rates over an annual present-day temperature regime and annual temperature regimes based on two warming scenarios (present-day temperature +1.5 and +3°C). The annual present-day temperature regime was based on the 7 day temperature means from a 40 year SST record from the Leigh Marine Laboratory, which is in the northeastern region of this species’ geographical range and where the F. lapillum used in this study were sourced. SMR, Ṁ_O2,max, MS, CT_max, thermal safety margin and RMR_maxTR were each compared between acclimation temperatures using one-way ANOVA with Holm–Sidak post hoc comparisons and individual tanks treated as replicates (N=3 per temperature). The relationship between MS and temperature was also assessed with a second-order (non-linear) polynomial regression and linear regression was used to examine the relationship between SGR and MS across tank replicates. RMR during thermal ramping was compared between temperature acclimation treatments using two-way repeated-measures ANOVA with individual tanks set as subjects. Thermal ramping for each temperature acclimation treatment started at a different temperature so RMR was compared over a standardised temperature range of 24 to 28°C. This temperature range was chosen as it was common to all four acclimation temperatures and did not include temperatures where any fish had obviously declining Ṁ_O2 or had lost equilibrium. Finally, the relationship between RMR at 24°C and CT_max and RMR at 28°C and CT_max was examined using linear regression. All statistical tests were carried out using R (version 3.1.1) and Sigma Plot (Systat Software, CA, USA).

There were nine mortalities during the 12 week growth trial but there were no significant differences in mortality rate between the temperature acclimation treatments (χ²=2.33, d.f.=3, P=0.507). The surviving fish appeared in good health at all temperatures with no obvious signs of disease or abnormal behaviour indicative of a compromised health status.

SMR was significantly different between the temperature acclimation treatments (ANOVA, d.f.=3, F=20.2, P<0.001) and there was an overall trend of SMR to increase with higher temperature (Fig. 1A). Post hoc tests showed that SMR was significantly higher at 24°C than at 21°C (P=0.027), 18°C (P=0.002) and 15°C (P<0.001), and significantly higher at 21°C compared with 15°C (P=0.024). Ṁ_O2,max appeared lowest in fish acclimated to 24°C but no significant differences (ANOVA, d.f.=3, F=2.79, P=0.11) were found between temperature treatment groups (Fig. 1A). A parabolic relationship described the effect of acclimation temperature on MS, with MS peaking at 18.2°C (Fig. 1A). There was a significant main effect of temperature on MS (ANOVA, d.f.=3, F=10.58, P=0.004) and post hoc tests showed significantly lower MS at 24°C than at 15°C (P=0.019), 18°C (P=0.005) and 21°C (P=0.011). Fish held at all temperatures gained mass during the 12 week temperature acclimation period and there was a parabolic relationship between SGR and temperature, with the highest growth rate seen at 18°C (0.68±0.041%M_b day⁻¹) but lower growth rates at 21°C (0.59±0.047%M_b day⁻¹), 15°C (0.51±0.034%M_b day⁻¹) and 24°C (0.39±0.048%M_b day⁻¹; Fig. 1B). Mixed-model analysis confirmed a significant main effect of temperature (d.f.=3, F=5.28, P=0.027) on SGR and post hoc comparisons showed SGR was significantly higher at both 18°C (P<0.001) and 21°C (P=0.041) compared with 24°C. The similar pattern in SGR and MS across the temperature acclimation treatments resulted in a significant linear relationship between these variables (R²=0.48, P=0.01), with an overall trend for higher SGR to be associated with greater MS across the individual tank replicates (Fig. 1C).

Under a present-day annual temperature regime, optimum temperatures for growth in F. lapillum occur over the warmer summer and early autumn months between December and May but growth potential drops substantially over the cooler winter period between June and November (Fig. 2). Under a scenario where the temperature at all times of the year is 3°C warmer than present day, optimum temperatures for growth occur over late autumn, winter and spring (May to November), whereas growth potential declines substantially over the warm summer months (Fig. 2). Despite substantial changes in the projected annual pattern of growth under warming scenarios, only minor differences are projected in the mean annual growth rate of F. lapillum, where a small but possibly insignificant increase exists compared with present day under +1.5 and +3°C warming scenarios (Fig. 2).

Following thermal acclimation, CT_max was highest in fish acclimated to 24°C and lowest in those acclimated to 15°C, and there was a significant main effect of acclimation temperature (ANOVA, d.f.=3, F=32.39, P<0.001; Fig. 3). Post hoc tests showed that fish acclimated to 24 and 21°C had significantly higher CT_max than fish acclimated to 18°C (P=0.004, P=0.007) and 15°C (P<0.001, P<0.001) and fish acclimated to 18°C had significantly higher CT_max than fish acclimated to 15°C (P=0.015). The difference between acclimation temperature and CT_max (thermal safety margin) was progressively reduced with increasing acclimation temperature and there were significant differences in thermal safety margin between all temperature acclimation treatments (ANOVA, d.f.=3, F=570.87, P<0.001; Fig. 3).

During thermal ramping, the RMR of fish acclimated to all temperatures increased to a peak in an exponential manner and then declined suddenly (Fig. 4A). Across a common temperature range (24–28°C) there was a significant effect of acclimation temperature (ANOVA, d.f.=3, F=17.13, P<0.001) and thermal ramping temperature (ANOVA, d.f.=4, F=195.6, P<0.001) on RMR and the interaction was significant (ANOVA, d.f.=12, F=2.59, P=0.016). Post hoc tests showed that RMR increased significantly (P<0.05) between 24 and 28°C, but there were also significant differences between acclimation temperatures that were dependent on thermal ramping temperature. Between 24 and 26°C, RMR was significantly higher in fish acclimated to 15 and 18°C than in fish acclimated to 21 and 24°C (Fig. 4A). However, at 27 and 28°C, fish acclimated to 24°C had significantly lower RMR than those acclimated to 15, 18 and 21°C, and those acclimated to 21°C had significantly lower RMR than 15°C acclimated fish (Fig. 4A). RMR attained a similar maximal Ṁ_O2 value in the 15°C (0.58 mg O₂ g⁻¹ h⁻¹±0.028), 18°C (0.59 mg O₂ g⁻¹ h⁻¹±0.006) and 21°C (0.56 mg O₂ g⁻¹ h⁻¹±0.021) groups but appeared to increase less in the 24°C group (0.49 mg O₂ g⁻¹ h⁻¹±0.015; Fig. 4A,B). As a result, peak values of RMR during thermal ramping were significantly different between the temperature acclimation groups (ANOVA, d.f.=3, F=4.98, P=0.031), and post hoc tests confirmed that fish acclimated to 24°C had significantly lower RMR_maxTR than those acclimated to 18°C (P=0.043; Fig. 4B). Interestingly, there was a significant inverse linear relationship between RMR and CT_max at 24°C (R²=0.67, P=0.001) and also between RMR and CT_max at 28°C (R²=0.48, P=0.013). As such, low RMR was consistently associated with a high CT_max and vice versa (Fig. 5).

To better understand how global warming may affect fish populations and their ecosystems, previous research has stimulated the need for more data on the response of respiratory variables, fitness-related variables and upper thermal tolerance limits to warming in species of interest (Nilsson and Lefevre, 2016). In this respect, the current research has adopted a novel approach in addressing growth and metabolic responses of the common triplefin fish to chronic and acute thermal changes. This provides insights into possible impacts of climate change for a highly abundant coastal species that resides in shallow subtidal and intertidal habitats. This work also allows an opportunity to assess the relevance of the OCLTT hypothesis in chronic and acute thermal tolerance.

There are few studies that allow for the potential effects of thermal acclimation of metabolism on MS under chronic time scales (weeks to years) (Norin et al., 2014; Sandblom et al., 2014, 2016) and even fewer that include measurements of growth (Donelson, 2015; Grans et al., 2014; Khan et al., 2014b). In the warm northern geographical range of F. lapillum, mean SST peaks in February (20.6°C Leigh Laboratory SST record) and remains within 0.5°C of this value for a period of 11 weeks over the austral summer. Thus tolerance assessments at higher mean temperatures approximating future temperature predictions should be tested over a similar time frame. This justifies our 12 week period of continuous temperature acclimation.

Chronic exposure to 24°C resulted in decreased growth rates and MS compared with lower temperatures (15, 18 and 21°C). Our data therefore indicate that an average sea temperature rise to 24°C owing to climate change will impair the whole-animal and metabolic performance of this species during the hottest months of the year. Depressed MS at 24°C was driven by higher SMR, and a marginally decreased Ṁ_O2,max in comparison to fish acclimated to lower temperatures. The metabolic response of F. lapillum to chronic warming therefore follows the response pattern outlined in the OCLTT hypothesis, where maximum aerobic capacity is limited, and warming beyond an optimal temperature increases basal costs and reduces MS (Pörtner and Knust, 2007; Pörtner and Farrell, 2008; Pörtner, 2010). However, this pattern of restricted MS at higher temperature is not universal across species, as the MS of some fish increases continuously with temperature (i.e. it does not peak) (Lefevre, 2016), because of the expansion of Ṁ_O2,max at higher temperatures. If MS theoretically sets a limit to the extent to which aerobically demanding processes can be performed beyond maintenance requirements (Clark et al., 2013), it is perhaps not surprising that the growth of F. lapillum fell at higher temperatures. MS was decreased and competing oxygen-demanding processes were potentially partitioned or prioritised across a shrinking pool of aerobically derived energy. While the correlation between MS and growth found in this study is not unique (Claireaux and Lefrancois, 2007; Jobling, 1981; Khan et al., 2014b), there is limited evidence linking key performance measures to MS under chronic temperature exposures (Clark et al., 2013). Other studies have found no association between temperature optimums for growth and MS in fish (Grans et al., 2014; Healy and Schulte, 2012; Norin et al., 2014) and it is has become apparent that, although the OCLTT hypothesis may be applicable in some species, it is not universal (Lefevre, 2016; Nilsson and Lefevre, 2016).

Although it was not quantified, egg deposition was evident during the growth trial at 15, 18 and 21°C but not at 24°C. This reproductive activity was unavoidable as wild F. lapillum spawn year round (Wellenreuther and Clements, 2007). Despite investment of energy in reproductive activity at 15–21°C, these fish still grew faster than fish at 24°C, where no reproductive activity was observed. The absence of reproductive activity at 24°C in this study could mean that spawning activity is constrained to cooler months if predicted warmer summertime conditions eventuate. Studies that specifically aim to address the influence of predicted future temperature regimes on the seasonality and performance of reproductive activity in this species would be of interest.

The pattern of SMR, Ṁ_O2,max and MS in the present study contrasts with a previous work for F. lapillum where, despite increasing SMR with higher temperature, a peak in Ṁ_O2,max at 24°C resulted in higher MS at 24°C than at lower temperatures (Khan et al., 2014a). However, Khan et al. (2014a) acclimated fish for only 4 weeks (versus 12 weeks in the present study) and fish may therefore have been in a different physiological state. The discrepancy in these findings highlights the importance of acclimation length in studies of this nature, and future studies investigating the time course of physiological adjustments in this species would be valuable. The different pattern of MS may also result from methodological differences, as Ṁ_O2,max was estimated over a substantially longer period by Khan et al. (2014a), and this may have underestimated the Ṁ_O2,max at lower temperatures.

Few studies have assessed climate change with respect to seasonal growth. Despite the forecast of impaired growth at projected higher summer temperatures, when considered on an annual cycle and taking into account seasonal temperature fluctuations, the annual growth of F. lapillum might be slightly improved with a universal 1.5 and 3.0°C increase in temperature (Fig. 2). Warmer summer temperatures owing to climate change are expected to result in chronic exposure to temperatures beyond present-day conditions and compromised growth for F. lapillum. However, the influence of warming on growth during late autumn, winter and early spring may actually improve overall growth performance as temperature shifts closer to an optimal range (90% of maximum) (16.5–20.5°C; Fig. 1). The geographical range of this species extends to the southern end of New Zealand, where mean sea temperatures are cooler at all times of the year. Therefore, if southern populations have optimum temperature ranges for growth similar to the northern populations considered in this study, warming may serve to improve growth performance as temperatures are shifted closer to the optimum range for growth at all times of the year. The predicted growth rates, however, need to be interpreted with caution as they are based on average growth rates measured over an extended period (12 weeks) of continuous exposure to a single temperature. Wild populations face gradually rising and falling temperatures in spring and autumn and thus may be in a continual state of acclimation during these months, which could alter the temperature–growth relationship. It is also not known how gradual seasonal temperature changes between summer highs and winter lows, as occurs over 3 to 4 months in natural environments, impacts the temperature–growth relationship during the relatively thermally stable winter and summer periods that follow. While seasonal temperature fluctuations undoubtedly play a role in determining when a species can grow fast, other factors such as food availability and competition also likely play an important role (Jones, 1986; Metcalfe, 1986). Consideration of how factors such as these are impacted by warming is required to develop a more complete picture of the growth response to climate change. Furthermore, the relationship between temperature and growth rate under laboratory conditions may not be applicable to wild populations.

Chronically exposing F. lapillum to higher temperatures over 12 weeks increased its upper thermal tolerance limit (CT_max), and this has been reported for other fish species (Akhtar et al., 2013; Beitinger et al., 2000; Donelson, 2015; Drost et al., 2016; Healy and Schulte, 2012; McDonnell and Chapman, 2015; Sandblom et al., 2016; Zhang and Kieffer, 2014). However, thermal safety margins were reduced progressively because the increase in CT_max under acclimation to higher temperature was proportionately less than the increase in acclimation temperature (i.e. acclimation response ratios are less than 1). Low CT_max acclimation response ratios appear to be common among ectotherms and this suggests that plasticity in upper thermal tolerance has limited potential to reduce overheating risk (Gunderson and Stillman, 2015). Indeed, CT_max only increased 0.12°C between the acclimation temperatures of 21 and 24°C in the present study (Fig. 3), suggesting that F. lapillum exposed to temperatures beyond the present-day range do not have the acclimatory capacity to further increase CT_max. As a consequence, the thermal safety margin was reduced to its lowest value upon acclimation to 24°C.

While subtidal populations of F. lapillum reside in more thermally stable environments, this species also occupies intertidal rock pools where large temperature fluctuations occur. The significant reduction in thermal safety margin upon acclimation to a predicted future summertime temperature could therefore threaten survival or restrict available habitat for these intertidal populations. The one-off exposure to an acute temperature rise in the present study does not fully represent rock pool conditions and it is acknowledged that repeated acute temperature rises (as occurs over a tidal cycle) may further condition upper thermal tolerance. However, there was no evidence of repeated heat shock improving CT_max of the common killifish (Fundulus heteroclitus) (Healy and Schulte, 2012). Acute temperature rises in rock pools may also coincide with algal-mediated photosynthetic hyperoxia, but hyperoxia also appears to have no benefit to the CT_max of fish (Brijs et al., 2015; Healy and Schulte, 2012; Rutledge and Beitinger, 1989). Intergenerational or adaptive genetic responses in the upper thermal tolerance of F. lapillum could therefore be essential for mitigating overheating risk in a warmer future, although these too appear limited for the upper thermal tolerance of fish (Sandblom et al., 2016).

The OCLTT hypothesis, with the differential availability of MS as the primary mechanism, seeks to unify explanations of thermal tolerance across the optimum to critical temperature range (Pörtner and Farrell, 2008). There is a difference, however, between how upper thermal tolerance is defined under the OCLTT hypothesis and the majority of studies investigating upper thermal tolerance in fish, which use CT_max (i.e. the temperature at which equilibrium is lost). Under the OCLTT hypothesis, the upper critical temperature (T_crit) is the point at which metabolic scope is nil (Farrell, 2016; Pörtner and Farrell, 2008), and how the CT_max fits within the OCLTT framework is unclear (Farrell, 2016; Healy and Schulte, 2012). Regardless, warm acclimation increases upper thermal tolerance in fish, and aerobic metabolism in fish with higher CT_max should reflect either an extension or retention of MS at high temperatures.

Theoretically, upper thermal tolerance could be increased through MS in two ways: (1) maximum aerobic capacity (i.e. Ṁ_O2,max) could be expanded at temperatures approaching critical limits, and/or (2) basal metabolic costs (i.e. SMR) could be reduced at temperatures approaching critical limits. In consideration of the first scenario, this study found no evidence of an expansion of maximum aerobic capacity with acclimation to warmer temperatures. In fact, despite having the highest CT_max, acclimation to the warmest temperature of 24°C restricted maximum aerobic capacity when measured at acclimation temperature (Fig. 1) and also limited maximum RMR at temperatures approaching critical limits (Fig. 4). Although Ṁ_O2,max was not directly measured at high temperatures, the RMR approaching critical limits should reflect maximum capacity, because under the OCLTT framework SMR and Ṁ_O2,max are the same at the upper critical temperature. This study finds more support for decreases in basal metabolic costs supporting improved upper thermal tolerance. Acclimation to increasingly warmer temperatures resulted in a lower RMR at equivalent temperatures during thermal ramping (Fig. 4A) and a lower RMR at 24 and 28°C was associated with higher CT_max (Fig. 5). These findings therefore fit the theory that physiological adjustments occur under warm acclimation to prevent the occurrence of high basal energetic costs (Norin et al., 2014; Pörtner, 2001; Sandblom et al., 2014; Seebacher et al., 2010). Lower metabolic rates in warm-acclimated fish may result from downregulation of mitochondrial function, which prevents excessively high metabolic rate during warming (Fangue et al., 2009; Pörtner, 2001) and could preserve MS at higher temperatures. However, despite the observed correlations in Fig. 5, it remains unclear whether lower RMR during thermal ramping actually plays a functional role in setting upper thermal tolerance because a clear difference in the metabolic rate of fish from the 21 and 24°C acclimation groups approaching their lethal thermal limit did not appear to dictate any difference in CT_max. Furthermore, experimental manipulations of oxygen transport capacity through anaemia (Brijs et al., 2015; Wang et al., 2014), hypoxia (Ern et al., 2016) and hyperoxia (Brijs et al., 2015) have little effect on upper thermal tolerance in fish.

In regard to chronic thermal tolerance, F. lapillum acclimated to 15, 18, 21 and 24°C (over 12 weeks) experience a decline in MS and growth fitness at 24°C. This finding is in support of the OCLTT hypothesis because of the correlation between metabolic and growth performance measures (Fig. 1). Unless genetic or transgenerational phenotypic acclimation can compensate, the main cause for concern would be that the metabolism and growth of F. lapillum would be sub-optimal with a 3°C temperature increase above present summertime SSTs. Few studies, however, assess climate change with respect to seasonal growth, and our modelling approach suggests that projected climate change will shift optimal growth to alternative times of the year and would thus result in an annual growth rate that is not impacted negatively (Fig. 2). Chronic warm acclimation only modestly increased upper thermal tolerance and, consequently, thermal safety margins declined at high acclimation temperatures (Fig. 3). In line with expectations under the OCLTT hypothesis, the observed increase in upper thermal tolerance under warm acclimation was associated with reduced RMR on acute thermal ramping (Fig. 4). In the context of global warming, the acute thermal tolerance of F. lapillum is a cause for concern because the ability of this species to withstand acute temperature change will decrease with every unit increase in peak chronic temperature. On that basis, the resilience of F. lapillum to future climate change will be reduced progressively as temperatures increase and the distribution of this common coastal species may become limited to thermally stable sub-tidal habitats or higher latitudes to cope.

Technical staff Peter Browne and Errol Murray are acknowledged for their assistance with design and fabricating the respirometry setups, and John Atkins is acknowledged for design of the respirometry software.