Energetic costs increase with faster heating in an aquatic ectotherm

We show that rate of temperature change has a systematic effect on the oxygen consumption of ectotherms and that as temperature increases more rapidly, the rate of oxygen consumption increases. These findings have implications for how we predict ectotherm responses to climate change, for example heat waves.


Introduction
A broad body of research exists examining the thermal sensitivity of metabolism, and the importance of this relationship reaches into many branches of biology and ecology (Brown et al., 2004).This relationship is particularly paramount for ectotherms as temperature directly affects many facets of their physiology such as metabolism and performance as their body temperatures reflect their environments (Angilletta Jr, 2009).This is even more important given that 99% of species on Earth are ectothermic (Ohlberger, 2013) and that climate change is increasingly altering global temperatures and patterns (Kefford et al., 2022).
An extensive suite of studies have explored the thermal sensitivity of metabolism and revealed many persistent patterns and trends.Firstly, when modelling the increase in metabolic rate with increasing temperature seen in ectotherms, physiologists and ecologists widely used an Arrhenius function (Robinson et al., 1983;Gillooly et al., 2001), an application promoted by Brown et al. (2004) in the metabolic theory of ecology (MTE).However, although this could be used to construct the thermal performance curve (Huey and Stevenson, 1979) for many biological rates, it began to receive criticism for its inability to fully account for the varying thermal sensitivity of metabolic rate, and several other models have been proposed to describe this relationship (DeLong et al., 2017;Padfield et al., 2021;Arroyo et al., 2022).A key component of focus in these models is the activation energy (E a ).Dell et al. (2011) showed overwhelming evidence that the mean activation energy for metabolic reactions is 0.65 eV, a value which holds true for the thermal responses of many biological traits.An E a value of ∼0.65 eV corresponds to a Q 10 value in the range of 2-3 (Dell et al., 2011), indicating a doubling or tripling in metabolic rate with every 10 • C change in temperature.This understanding of the thermal sensitivity of metabolic rate is widely used to model and predict high-level processes such as predicting ecosystem level responses (Dillon et al., 2010;Rubalcaba et al., 2020), understanding species climate resilience (Vajedsamiei et al., 2021) and predicting population growth under a changing climate (Stearns, 1992;DeLong et al., 2018).
Although the thermal sensitivity of metabolism has been widely studied, research has primarily focused on static temperature experiences (i.e.estimating metabolic rate at a range of constant temperatures) (Enders et al., 2006;Williams et al., 2012;Halsey et al., 2015).However, temperature in nature rarely remains static, and so examining metabolic responses under static thermal conditions may not be an accurate reflection of what is occurring in nature.Thermal variability in the environment can have varying effects on fitness components of organisms, potentially hindering the adaptive capacity (Kingsolver and Buckley, 2020).Therefore, we need to better understand how the thermal sensitivity of metabolism varies when temperature is undergoing change.Understanding this is essential to informing predictive models of ectotherm responses to climate change-induced temperature change (Dillon et al., 2010).Some previous studies have investigated the effect of dynamic and stable temperature regimes on metabolism of ectotherms (Via, 1985;García-Guerrero et al., 2011;Latournerié et al., 2011;Williams et al., 2012;Lake et al., 2013;Vasseur et al., 2014;Zhang et al., 2015;Killen et al., 2017;Semsar-kazerouni and Verberk, 2018;Guzzo et al., 2019).However, the dynamic temperature scenarios used typically involve stepwise temperature changes (i.e. as opposed to continuous rate of increase or decrease) or oscillating temperature regimes (i.e.multiple increases and decreases in temperature over a period of time).The average metabolic rate of the entire exposure period is then estimated and directly compared to a stable temperature regime.This approach makes it difficult to understand the direct influence of heating or cooling rate on the thermal sensitivity of metabolism.To the best of our knowledge, there exists only one study that measured metabolic rate of a terrestrial insect species Gryllus pennsylvanicus, continually under continuous temperature change (Lake et al., 2013), which found mixed results: no significant difference was found in estimates of thermal sensitivity of metabolism between dynamic and stable treatments but, absolute estimates of metabolic rate were higher under dynamic thermal conditions than stable.There could be several reasons to expect rates of temperature change to influence the thermal sensitivity of metabolism, which have been raised previously, for example: (i) cost of acclimation may differ with rate of temperature change (Rohr et al., 2018).Faster rates of change may incur additional energetic costs for ectotherms as they are forced to acclimate to a new temperature environment over a shorter period of time; (ii) the balance between supply and demand of resources may be altered as animals operating at higher temperatures, with higher metabolic rates, demand more resources, but also have a greater capacity to supply resources (Gillooly et al., 2006); and (iii) overall metabolic cost may be impacted by rate of temperature change, specifically increasing metabolic cost with slower rates of temperature change, everything else being equal (Rezende et al., 2011).
Given the inherent complexity of natural systems and the accelerating pace of global warming, there is a pressing need for experimental approaches capable of accurately measuring the thermal response of organisms.In this study, we examined whether the rate of temperature change affects the thermal sensitivity of aerobic metabolism.We do this by conducting closed-chamber respirometry on an aquatic ectotherm, the Atlantic ditch shrimp (Palaemonetes varians) (Leach, 1814), to estimate oxygen consumption under varying thermal conditions.Closedchamber respirometry (which measures gas exchange between an animal and its environment) is widely recognised as a valid proxy for measuring aerobic metabolic rate of aquatic ectotherms (Nelson, 2016;Killen et al., 2017;Guzzo et al., 2019;Vajedsamiei et al., 2021) as the rate of oxygen uptake from the environment is expected to relate stoichiometrically to ATP production rate via mitochondrial oxidative phosphorylation (Clarke and Fraser, 2004;Nelson, 2016;Killen et al., 2021).Previous respirometry experiments on shrimps demonstrated that oxygen consumption generally increases with an increase in temperature: Litopenaeus stylirostris (Spanopoulos-Hernández et al., 2005) and Penaeus chinensis (Chen and Nan, 1993).Furthermore, P. varians has been used in temperature studies in the past (Nielsen and Hagerman, 1998;Palma et al., 2009;Cottin et al., 2010;Oliphant et al., 2011;Ravaux et al., 2012;New et al., 2014), making it a suitable study species.In addition to their value as a model organism, shrimp have considerable ecological value in aquatic food webs, serving both as predator and prey.They serve as a primary food source for many large aquatic The findings of this study may have significant implications for understanding how aquatic ectotherms, particularly economically important species such as shrimp, respond to global warming, with findings potentially transferable to other aquatic ectothermic species of similar body sizes and biology to the study species.We aim to contribute insights into how temperature fluctuations affect aerobic metabolism in ectotherms, further strengthening predictions of organismal and population response to environmental stressors in aquatic environments, thereby informing sustainability and conservation strategies.

Materials & Methods
Closed-chamber respirometry was used to estimate the aerobic metabolic rate of P. varians by recording the rate of decline in dissolved oxygen (DO) concentration (mg l −1 ) over time, under varying thermal conditions.Measurements were taken individually on 161 shrimp, with 38 experiencing stable temperature, 67 experiencing heating and 56 experiencing cooling.
P. varians is a shallow water shrimp native to Western Europe (Ravaux et al., 2012).P. varians is known to inhabit areas where the environmental temperature can seasonally fluctuate from 0 to 33 • C (Lofts, 1956;Jefferies, 1964;Healy, 1997;Oliphant et al., 2011).They are known to typically inhabit brackish waters but have been found in very low salinities, including fresh water (Hagerman and Uglow, 1984).These shrimps are often used for fishing bait or as live feed for aquarium fish (Palma et al., 2008a;Palma et al., 2008b;Palma et al., 2009).This species was chosen because it has previously served as a model species for temperature studies (Ravaux et al., 2012) due to their adaptation strategies to temperature (Cottin et al., 2010) and their capacity to tolerate periods of severe hypoxia (Hagerman and Uglow, 1984), tolerating low-oxygen tensions before entering anaerobic respiration (Nielsen and Hagerman, 1998;Peruzza et al., 2018).

Data Collection Acclimation phase
Animals for this experiment were sourced from mainland Europe by a specialist aquarium stockist (Seahorse Aquariums Ltd).They were transported in transparent plastic bags filled with brackish water at ∼18 • C, with transport from source location to laboratory taking ∼6 days.On delivery to the laboratory, animals were removed from the bags and placed in five large buckets.Before experimental treatments, animals were acclimated stepwise to the five desired accli-mation temperatures (10, 12, 15, 18 and 21 • C) during a 24-h period, using temperature-controlled buckets with a photoperiod of 12 h:12 h light:dark.Shrimp were not fed during this period.Shrimp were then transferred to their corresponding acclimation tanks and fed.The laboratory setup consisted of five aerated tanks (37 × 23.5 × 26 cm) filled with unfiltered, treated tap water (∼0.5 ), in a room kept with a 12 h:12 h light:dark phase.Tank water was treated with 'Organic Aqua Start Up' water treatment and water quality was maintained with weekly 'Organic Aqua Fish Care'.Temperature in the tanks was maintained at stable temperatures, controlled by programmable units (Inkbird ITC-310 T-B; constant desired temp ±1.0 • C), and shrimp were kept at these stable acclimation temperatures for 2-30 days before experiments commenced.A minimum acclimation phase duration of 2 days was chosen in accordance with previous studies of a similar nature on similar (or the same) species (Berglund and Bengtsson, 1981;Nielsen and Hagerman, 1998;Miller et al., 2002;González-Ortegón et al., 2013).The shrimp were provided with flake fish food approximately every second day.Before respirometry, shrimp were fasted for a minimum of 12 h (average fasting time for stable trials was 26 ± 11 h and for heating/cooling trials was 23 ± 12 h).

Temperature trials
For the stable measurements, DO concentration was recorded for 1-2 h at five stable treatments corresponding to the acclimation treatments (i.e. 10, 12, 15, 18 and 21 • C).For the dynamic temperature treatments, DO concentration was measured under four thermal regimes: (i) heating from 15 to 21 • C, (ii) 10 to 21 • C, and then cooling from (iii) 15 to 10 • C and (iv) 21 to 10 • C, with starting and ending temperatures corresponding to acclimation temperatures.These regimes were repeated at four ramping rates (0.0083 , where temperature was increased or decreased continuously.Individual shrimps were removed from the acclimation tanks and placed in 200ml plastic chambers, filled with water taken from the acclimation tanks.The chambers were then partially submerged in temperature-controlled water baths.The temperature of the water baths was controlled by a programmable unit (Inkbird ITC-310 T-B; desired temp ±0.5 • C).Shrimp were allowed to settle in the chambers for 1-2 h before commencing measurements to allow them to recover from handling during transfer.Once settled, a lid was placed on the chamber, a DO probe (Go Direct Optical Dissolved Oxygen Probe) was inserted through an opening in the top of the lid and the chamber was hermetically sealed.Probes were 100% calibrated before the beginning of each experimental phase following the manufacturer calibration procedure.Probes began recording as soon as the chamber was sealed, measuring DO concentration (mg l −1 ), DO saturation (%) and temperature ( • C) continuously.Sampling frequency of the probe was varied between temperature treatments due to probe memory limitations; when conducting experiments at λ = 0.0083, 0.0167, 0. , probes recorded at a frequency of 0.2, 0.5, 1 and 1 sample min −1 , respectively.All experiments were conducted during the photophase on shrimp that had not been fed for ≥12 h (for more details on acclimation and feeding times, see Supplementary Material, Table S1).Individual shrimps were measured alongside 'control' chambers (chambers containing tank water but no shrimp) to record bacterial respiration and later correct the DO concentration data.Chambers were visually shielded from external disturbance during respirometry by the high, opaque walls on the water baths.After data collection, shrimp were euthanised by submersion in 90% ethanol, blotted dry and weighed (recording wet weight ± 0.001 g).Chambers were cleaned with 90% ethanol between all experiments to reduce bacterial growth within.
All work was conducted following the principles of the '3Rs': reduction, refinement and replacement.Animal ethics approval for all works was sought from and approved by Trinity College Dublin School of Natural Sciences (SNS) Research Ethics Committee (project number '2021-06 (revised)').

Data Analysis
Several animals were removed from the analysis phase due to varying reasons: (i) 105T1021R5 was removed because it was found to have lost the ability to right itself at the end of the experiment, (ii) 28T2110R1 and 13T1021R0.5died during the experiment, (iii) 37T2110R0.5 and 38T2110R0.5 had too short of a fasting period (5.5 h) and (iv) 122T1521R1 timer was set incorrectly.This resulted in a total of 117 ramping treatments and 38 stable treatments for use in the data analysis phase.
All animals were recorded concurrently with control 'blank' chambers, containing water from the same source but no shrimp, to remove the effect of background (bacterial) respiration.A linear regression was fitted to the DO concentration of each blank chamber.The DO concentration for each animal was then corrected by subtracting the negative slope of the blank chamber from the DO concentration raw data, such that.DO corrected = DO animal −(− slope background respiration ).

Modelling
Average metabolic rate (MO 2 ) was estimated for each animal held at one of the five stable temperature treatments.MO 2 was estimated following the methods of Svendsen et al. (2016), by fitting a linear model to all individual DO concentration data, extracting the slope (m) of this line and then correcting for 'effective volume' (volume of the chamber − volume of the animal).To avoid difficulties in measuring the volume of the animal, it is common practice to instead subtract the mass of the animal (in kg) from the volume of the chamber (in l) (Svendsen et al., 2016) because shrimp were assumed to be neutrally buoyant in water, so densities are assumed equivalent and equal to 1. Metabolic estimates were also mass-controlled and allometrically scaled (coefficient = 0.8; Christensen et al. (2020)): where MO 2 is the metabolic rate in milligrams O 2 per minute per kilogram, V eff is the 'effective volume', mass is in kilograms and m is the slope of the line of DO concentration versus time.These MO 2 values were then plotted against average temperature, and a thermal performance curve (TPC) was fitted using the 'rTPC' R package (Padfield and O'Sullivan, 2021; R Core Team, 2022).Several models were chosen for fitting based on their ecological applicability to data of this kind, and the AICc score (used for small sample size, n = 37) was calculated (Supplementary Material, Table S1) and used for model selection based on the lowest AICc score.The "Rezende" model (Rezende and Bozinovic, 2019) gave the lowest AICc score and thus was chosen for use.Total performance pf was calculated using the full model: where Q 10 defines the fold change in performance as a result of increasing the temperature by 10 • C, d is a constant controlling the rate of decay from threshold temperature T th upwards and Ce is the rate of DO decline in this scenario.
This TPC curve was used to predict the total oxygen consumption (mg O 2 ) for animals that experienced heating or cooling with bounds set as the starting (T 1 ) and ending The relationship between predicted and observed total O 2 consumption was investigated to see if the stable temperature trials could be used to predict how much oxygen animals would consume in comparable thermal ranges but under differing rates and directions of temperature change.To do this, the area under the curve was calculated by numerical integration using the 'integrate' function in R (Piessens et al., 2011), with upper and lower limits set to the start and end temperature of each individual ramping trial.This integral was then converted to total predicted oxygen consumption (mg) by incorporating duration of the trial and mass of the organism using the following equation: where λ is the rate of temperature change and mass in kilograms.These predicted values were then plotted against the recorded (or observed) total oxygen consumption.Observed total oxygen consumption was calculated, using the recorded DO concentration data, by subtracting the average of the last five DO concentration values from the average of the first five DO concentration values and dividing by the effective volume.These data were log transformed (to overcome the effect of larger O 2 measurements being prone to wider uncertainties) and plotted against each other.
To further investigate any potential effect of direction of temperature change, the residuals of the previous plot were calculated (i.e.log of observed O 2 consumption minus log of predicted O 2 consumption) and the relationship between these residuals and λ was investigated.A breakpoint regression was fit, using Bayesian inference, using the 'JAGS' R package (Plummer et al., 2022).

Results
In total, 155 datasets were analysed: 64 heating trials, 53 cooling trials and 38 stable trials, with shrimp wet weight ranging from 0.048 to 0.295 g (Supplementary Material, Table S2).We found that the relationship between MO 2 and average temperature was best described by a thermal performance curve (Figure 1), specifically that modelled by Rezende and Bozinovic (2019) because it gave the lowest AICc score of 76.83 compared with the second lowest Pawar model AIC of 77.295 (Supplementary Material, Table S2).This Rezende model yielded estimates of 3.03, 0.26, 17.93 and 0.05 for q 10 , Ce, T th and d, respectively.By integrating under this curve, total predicted oxygen consumption (mg O 2 ) was calculated for each individual that underwent heating or cooling (Figure 2).Three individuals (82T2110R10, 84T2110R10 and 36T2110R10) were excluded from this analysis as they showed near zero O 2 consumption (−0.005, −0.02 and −0.008 mg, respectively), likely due to the fast cooling rate (λ = 0.16 • C min −1 ) and short duration of exposure (∼60 min).
We see that generally, the points are clustering along the 1:1 line (Figure 2 grey dashed line), indicating that the stable temperature trials predict total oxygen consumption for the dynamic temperature trials.However, any potential effect of direction of temperature change is unclear.Therefore, the residuals of this were calculated (log of observed O 2 consumption minus log of predicted O 2 consumption), and the relationship between the residuals and λ was investigated (Figure 3).The breakpoint regression was fit with one breakpoint forced through λ = 0 because all points greater than zero related to heating trails and all points less than zero related to cooling trails.
Figure 3 reveals that faster rates of heating and cooling tend to increase total oxygen consumption.However, there appears to be a weaker influence of rate of cooling than rate of warming on total oxygen consumption: the slope of the heating data and cooling data are significantly greater than zero (slope heating = 7.71, P(slope > 0) = 1.00 for heating, and slope cooling = −2.73,P(slope < 0) = 0.96 for cooling data) with cooling rate having a weaker influence on total oxygen consumption than heating rate.Furthermore, as temperature increases more rapidly, there is an increasing rate of oxygen consumption (as the regression line crosses the grey dashed line, with a continual positive slope).In addition, animals consumed less oxygen when experiencing slow rates of temperature change than when they were experiencing no change in temperature (evident in the heating and cooling data points, from λ = −0.02 to 0.02).By examining this relationship (Figure 3) between the residuals and λ, it is evident that at slower rates, the stable temperature trials will tend to underestimate oxygen consumption of the dynamic temperature trials (evidenced by the larger majority of data points lying below the grey dashed line), and at faster rates, the predictions will be overestimated (evidenced by the larger majority of data points lying above the grey dashed line).

Discussion
We show that the rate of temperature change has a systematic effect on the oxygen consumption of a model aquatic ectothermic species.Perhaps most compellingly, we find that as temperature increases more rapidly, the rate of oxygen consumption increases, and although we find a similar relationship as temperature decreases more rapidly, the rate of cooling had less of an influence on the rate of oxygen consumption than the rate of warming.We also see that at slower rates   of temperature change, the animals consumed less oxygen than when they were experiencing no change in temperature (i.e.stable temperature).These findings are somewhat unexpected as previous literature indicates that dynamic or variable temperature environments are more energetically costly to ectotherms than stable environments (Williams et al., 2012;Morón Lugo et al., 2020), following Jensen's inequality (Ruel and Ayres, 1999), and that metabolic costs increase with decreasing heating rate, likely due to extended exposure duration (Chown et al., 2009;Rezende et al., 2011).However, these differing findings may be related to Q 10 values.Chown et al. (2009)   Several tentative explanations arise to explain why faster rates of temperature change imply higher metabolic costs.Previous studies have shown that accumulated metabolic costs of individuals under different ramping regimes are not equal, and animals undergoing slower rates of temperature change may be undergoing acclimation due to the associated longer exposure durations (Rezende et al., 2011).Acclimation capacity may be allowing the animals to cope with the increased temperature (Vasseur et al., 2014;Rohr et al., 2018) and to downregulate the increase in metabolism.Furthermore, the longer experiments (i.e.those with the slowest rates of temperature change) may be providing enough time for hardening to occur, a form of phenotypic plasticity (Hoffmann et al., 2003) that protects cells from subsequent injury (Overgaard et al., 2006).Another potential explanation could be that there may be a mismatch between metabolic supply and demand as the temperature increases and the rate of temperature change increases.This could result in internal entropy, causing stress and/or damage, resulting in negative implications for performance over time (Pörtner, 2012;Ritchie, 2018;Vajedsamiei et al., 2021).Future studies to identify the mechanisms underlying the observed increase in metabolic costs at faster rates of temperature change could include: (i) investigating the genetic expression of heat shock proteins and their role in the hardening processes; (ii) energy budget modelling of organisms undergoing temperature change to assess potential mismatches in metabolic supply and demand; and (iii) exploring the potential generality of this identified relationship in other aquatic ectotherms with differing acclimation abilities and of varying body sizes, with the latter allowing for exploration of differing responses to thermal stress across body sizes.
As discussed in the Introduction, there exists only one study that measures metabolic rate continually under continuous temperature change.Lake et al. (2013) carried out closed-chamber respirometry on a species of terrestrial insect experiencing dynamic temperature change.Our study was conducted in a similar manner to that of Lake et al. (2013).They found that, although estimates of thermal sensitivity of metabolism did not differ significantly between dynamic and stable treatments, absolute estimates of metabolic rate were higher under dynamic thermal conditions than stable ones.Specifically, they reported significantly higher estimated oxygen consumption during cooling at a rate of 0.1 • C min −1 compared with stable temperature treatments.This finding, however, only occurred in the group of animals that had undergone repeated respirometry and did not occur in animals that were only tested once (as is the case with our study).This discrepancy led them to suggest that rate and direction of temperature change effect on metabolic rate is an aspect that needs to be considered when extrapolating from laboratory studies to the field.At least in a broad sense, our findings concur with those of Lake et al. (2013) and confirm that a rapid increase or decrease in ambient temperature results in a greater metabolic demand compared with situations of gradual and slow change.Furthermore, it is noteworthy that Lake et al. (2013) estimated oxygen consumption from the air, in contrast to this study, which examined an aquatic (marine) model organism.
A major consideration we acknowledge is the potential for a lag between the water temperature and the body temperature of the animals, and so to counteract this, we intentionally selected a small-bodied aquatic animal (with body wet weight ranging from 0.048 to 0.295 g), therefore increasing the potential for a large heat transfer coefficient and reducing the lag between body and water temperature.Notwithstanding this, should there have been a small lag present during our experiment, this would only increase the effect we show because it would mean that the higher metabolic rates are occurring at even slower rates of heating and cooling because the body would heat and cool slower than the water, thereby increasing the slope of the heating and cooling relationships seen in Figure 3, implying that the impact of λ could be even more influential.
If the dependence of metabolism on heating rate that we document proves to be a general phenomenon seen across species, this would have implications for how we predict ectotherm responses to climate change.The magnitude, duration and frequency of extreme thermal events (i.e.heat waves) is predicted to increase (Coumou and Rahmstorf, 2012;Kefford et al., 2022), and ectotherms are likely more vulnerable to changing temperatures as their body temperature follows that of their environment (Angilletta Jr, 2009;Åsheim et al., 2020).With heat waves increasing in frequency and intensity (Perkins et al., 2012;Minuti et al., 2021), our findings indicate that perhaps heat waves with faster rates of heating may be more costly to the animals than slower heat waves, even if they ultimately reach the same maximum temperature.This potential impact may be further exacerbated given the widespread deoxygenation of both oceanic and inland waters resulting from global warming (Oschlies et al., 2018;Jane et al., 2021).Furthermore, these findings highlight the potential impact of thermal pollution on aquatic ectotherms where discharges of heated effluents, such as sewage, or stagnant water from dams may pose a significant threat to aquatic ectotherms through acute temperature changes.This study is an important step in providing data on ectotherm capacity to adapt to change, through experimental manipulations, and feeding this into mechanistic models for predicting ectotherm responses to climate change and thermal pollution.Our results suggest a systematic influence of temperature change on energetics and encourage future work to determine the generality of this finding across species.

Figure 1 :
Figure 1: Average MO 2 , for the duration of the trial, against average temperature ( • C) for all stable temperature individuals, with thermal performance curve fitted, following the method of Rezende and Bozinovic (2019).

Figure 2 :
Figure 2: og observed versus predicted O 2 consumption (mg) for each ramping individual, with points coloured by rate of temperature change (λ • C min −1 ) and symbols indicating if the temperature was heating (triangle) or cooling (circle).The 1:1 (grey dashed) line indicates where we would expect the points to lie along if the stable treatments perfectly predicted the ramping trial O 2 consumption.