Marine heatwaves (MHWs) profoundly disturb tropical coral reefs, imperilling species fitness and survival. Ectothermic piscivorous reef fishes are particularly vulnerable to MHWs since all aspects of their survival are dictated by ambient temperature. Severe +4°C MHWs are projected to escalate daily energy demands by ~32-55%, compelling piscivores to pursue larger or more frequent prey to survive. However, the feasibility of these responses have been questioned, as evolved predation and digestive strategies are constrained to specific prey types and sizes to safeguard residual aerobic scope (AS) during digestion for other vital processes. Instead, prevailing theory proposes appetite reductions at temperatures above optimal, preserving AS at the expense of growth and/or fitness. We investigated this dichotomy in the thermal foraging responses of Arc-eye hawkfish (Paracirrhites arcatus) and blacktail snapper (Lutjanus fulvus), evaluating energetic demand (standard metabolic rate, SMR), AS, appetite (meal mass intake), and capacity for digestion (specific dynamic action, SDA). Spanning a thermal gradient encompassing present-day winter (24.0±0.1°C), summer (27.5±0.1°C), and MHW (31.0±0.1°C), we show that SMR increased by ~65% from winter to MHW for both species, while AS increased by ~31-67%. Contrary to predictions of reduced appetite, both species consumed ~106% larger meals, yielding a ~35-105% greater SDA magnitude. Surprisingly, increased appetite did not encroach on residual AS as both species maintained the physiological flexibility to process larger meals while retaining ~45-60% of AS at the post-prandial peak. Although larger meals take longer to digest, both species exhibited faster digestion with rising temperatures resulting in a maintained or shortened SDA duration during MHWs, simultaneously enabling increased feeding rates while preserving aerobic reserves to support heightened predation. Our findings underscore the physiological feasibility of increasing appetite for some piscivores, while highlighting the ecological challenge of increasing prey numbers and sizes amid declining prey densities and prey size-reductions caused by ocean warming.

Study species, collection and holding conditions

Arc-eye hawkfish (Paracirrhites arcatus) and blacktail snapper (Lutjanus fulvus) were selected for this study. Both species adopt a ‘feast and famine’ approach to foraging (i.e., intermittently consume single large meals consisting of fish, crustaceans or other invertebrates), are from evolutionary distinct lineages (Cirrhitidae and Lutjanidae, respectively), and are found in high abundance on shallow water reefs throughout the equatorial Pacific (Armstrong & Schindler, 2011; Randall, 1998; Sheppard et al., 2017). Fish were collected with hand nets and barrier nets by scuba divers during the winter (January-March) and summer (July-September) of 2021 and 2022 from reefs around Kāneʻohe Bay on the eastern coast of Oʻahu, Hawaiʻi (main collection site for L. fulvus: 21.414234, -157.784047 and P. arcatus: 21.477181, -157.791538).

Following collection, fish were transported to the Johansen Fish Resilience Laboratory at Moku o Loʻe in Kāneʻohe Bay, and housed in groups of four individuals in large conical holding tanks (height: 86.0 cm; diameter: 82.5 cm; volume: 260 L) containing sections of PVC pipe for shelter. Tanks were supplied with flow-through, filtered, aerated and UV-sterilized seawater and subjected to a 12hr:12hr light:dark photoperiod. Water within the holding tanks were maintained at 24.0±0.1°C and 27.5±0.1°C during the winter and summer, respectively, reflecting mean seasonal sea surface temperatures (SST) of Kāneʻohe Bay (data retrieved online from tidesandcurrents.noaa.gov; Moku o Loʻe weather station; ID: 1612480; depth = 1 m; 21.433° N, 157.786° W). Water temperature within each tank was regulated using temperature control relays (WH1436, Willhi, China) that activated to either heat or cool the tank when water temperatures were not within the desired range. Heating within each tank was achieved with an 800 W aquarium heater (TH-08005, Finnex Inc., USA), whereas cooling was achieved with a submersible pump (D08V045CD, Kedsum, China) that circulated water from within the tank through a stainless-steel coil that was submerged in a 95 L external reservoir held at 5°C by a water chiller (ECO-1 1/2HP, Ecoplus, USA).

Fish were allowed one week to acclimate to laboratory conditions and were then subjected to a ‘treatment’ for two weeks (see Treatment groups below). During this time, fish were fed to satiation on a daily basis with ~0.2-1.0 g pieces of Pacific squid (Loligo opalescens). Animal care and all of the experimental procedures described below complied with the ethical standards of the Institutional Animal Care and Use Committee at the University of Hawaiʻi at Mānoa, approved under the permit number 3200.

Treatment groups

Fish collected in winter were assigned to the ‘winter’ treatment, whereas fish collected in summer were randomly assigned to either the ‘summer’ or ‘MHW’ treatment. Following acclimation to laboratory conditions all fish were exposed to a two-week trial. Specifically, fish in the winter and summer treatments were held at mean winter and summer SST for two weeks (i.e., 24.0±0.1°C and 27.5±0.1°C, respectively), while fish in the MHW treatment were held at mean summer SST for one week before being subjected to a simulated MHW for one week. A simulated MHW in this study was defined as a discrete, prolonged warm water anomaly above the climatological mean with a minimum duration of five days (following Hobday et al., 2016). Based on the characterization of MHWs in Kāneʻohe Bay between 1994-2020 (Tran & Johansen, 2023), MHWs were simulated by increasing water temperature ~0.9ºC per day from mean summer SST to the peak of the MHW (31.0±0.1°C), whereafter water temperatures were maintained at the peak of the MHW for at least three more days. These conditions closely matched the mean duration (days), maximum heating rate (°C day-1), and maximum intensity (°C) for MHWs detected on the reefs within Kāneʻohe Bay, ensuring that MHW simulations followed real-world conditions (Tran & Johansen, 2023). Based on pilot trials, all fish were fasted during the final three days of the treatment period to avoid the confounding effects of specific dynamic action on pre-prandial whole-animal aerobic metabolic rates (Secor, 2009).

Intermittent-flow respirometry

Oxygen uptake rates (ṀO2) of P. arcatus and L. fulvus were measured using best practices for intermittent-flow respirometry (Clark et al., 2013; Roche et al., 2013; Svendsen et al., 2016). Specifically, respirometry trials were conducted in a temperature-controlled room containing two experimental tanks (length: 97 cm, width: 53 cm, height: 37 cm) supplied with flow-through, filtered, aerated and UV-sterilized seawater under a 12hr:12hr light:dark photoperiod. Water temperature within the experimental tanks were maintained within 0.1°C of winter, summer, and peak MHW treatment temperatures. Each experimental tank contained either two large cylindrical acrylic respirometers for L. fulvus (length: 22.0 cm, diameter: 10.0 cm, volume: 1.600 L) or four smaller respirometers for P. arcatus (length: 10.6 or 14.6 cm, diameter: 3.8 or 3.8 cm, volume: 0.166 or 0.214 L for small and large individuals, respectively). To feed fish within the respirometer, a 1 cm hole was drilled into the top of the anterior end of the respirometer, which was sealed with an oxygen impermeable rubber stopper when obtaining ṀO2 measurements. Water was continuously circulated through each respirometer using an in-line submersible pump (AD20P-1230E, DollaTek, USA) within a recirculation loop to ensure a homogenous concentration of oxygen throughout the respirometer. Automated flush pumps intermittently refreshed the water within each respirometer according to specific flush cycles (see Experimental protocol) set in the AquaResp software (v3.04, https://github.com/bigb8/AquaResp. This ensured that oxygen levels in the respirometers always remained above 80% air saturation. The partial pressure of oxygen in the water within each respirometer was measured continuously at 1 Hz using a fiber optic oxygen sensor mounted in the recirculation loop where the flow is sufficient to ensure a rapid response time of the sensor. The optode was connected to a 4-channel Firesting Optical Oxygen Meters (Pyro Science, Germany), which in turn were connected to a PC that logged the data. Oxygen measurements were automatically compensated for temperature (via a Pt100 temperature probe connected to the temperature port of the oxygen meter) and salinity (via manual input of water salinity into the Firesting logger software). Mass-specific ṀO2 were then automatically calculated by the AquaResp software from the linear decline in the partial pressure of oxygen during the measurement period when the flush pumps were off. Only ṀO2 calculations with an R2 of greater than 0.95 were kept for further analysis.

Experimental protocol

Following the treatment period, the experimental protocol consisted of initially determining pre-prandial whole-animal aerobic metabolic rates such as maximum metabolic rate (MMR, day 1) and SMR (days 2-3), and then subsequently determining post-prandial whole-animal aerobic metabolic rates to evaluate SDA (days 4-6) for each fish.

To determine MMR, fish were individually transferred from their treatment tank to an identical tank with a reduced water level (~15 cm) and subjected to a ‘chase protocol’ (Clark et al., 2013; Norin & Clark, 2016; Roche et al., 2013). This protocol consisted of chasing the fish by hand for three minutes followed by a period of air exposure for one minute during which they were weighed and measured. Fish were then placed into respirometers and ṀO2 were obtained overnight with the AquaResp software set to ‘MMR mode’ (i.e., ‘Wait’ = 0.5 min, ‘Measure’ = 1 min, and ‘Flush’ = 2.5 min). Then to determine SMR, ṀO2 measurements were obtained for 48 h starting the following day at 07:00 HST with the AquaResp software set to ‘SMR mode’ (i.e., ‘Flush’ = 5 min, ‘Wait’ = 1 min, and ‘Measure’ = 1 min).

Following the evaluation of pre-prandial whole-animal aerobic metabolic rates, fish were fed inside the respirometers by gently removing the rubber stopper from the hole in the respirometer and inserting ~0.2-1.0 g pieces of Pacific squid until the individual was satiated. Fish were deemed to be satiated when a piece of squid within the respirometer was not consumed within 15 min, whereafter unconsumed pieces of squid were gently removed and the respirometer was re-sealed with the rubber stopper. The feeding process typically took ~30 min. To determine SDA, post-prandial ṀO2 measurements were then obtained for another 72 h with the AquaResp software set to ‘SMR” mode. At the end of the experiment, fish were removed from the respirometers for subsequent release back into the wild at the approximate location of capture.

To account for background respiration, linear regression over time using measures of bacterial ṀO2 obtained from empty respirometers before and after each respirometry trial were subtracted from all ṀO2 measurements. To limit background respiration rates, all equipment was disinfected with a 1% bleach solution, thoroughly rinsed with freshwater, and allowed to dry before commencing further trials.

Determining pre-prandial responses

A total of 44 individual P. arcatus (n: winter=14, summer=17, MHW=19) and 35 L. fulvus (n: winter=11, summer=12, MHW=12) were used to determine pre-prandial metabolic rates. MMR was defined as the highest pre-prandial Ṁ*O2 measurement (Norin & Clark, 2016, 2017), while the 20% quantile method on *Ṁ*O2 measurements obtained during the 48 h pre-prandial period yielded the most consistent SMR estimates based on visual inspection of *Ṁ*O2 and the cumulative variance of the mean lowest normal distribution (Chabot *et al., 2016). In addition, Q10 was calculated (i.e., Q10 = (R2/R1)[10/(T2-T1)]) using mean SMR values for each species between winter and summer (i.e., R1: SMRwinter; R2: SMRsummer; T1: 24.0°C; T2: 27.5°C), as well as between summer and at the peak of a simulated MHW (i.e., R1: SMRsummer; R2: SMRpeak of MHW; T1: 27.5°C; T2: 31.0°C). AS was calculated as MMR minus SMR (Clark et al., 2013).

In order to accurately quantify SDA (see Determining post-prandial responses below), it is recommended to account for the 'metabolic circadian rhythm' of the study organism to account for brief periods of spontaneous activity (Roe et al., 2004). This was achieved for each individual by calculating an hourly moving median of the ṀO2 measurements obtained during the 48 h pre-prandial period. From the smoothed data, parameters of the ‘metabolic circadian rhythm’ such as the mesor (i.e., mean of the circadian rhythm) and amplitude (i.e., difference between the mesor and peak of the ‘metabolic circadian rhythm’) were calculated for each individual. Likewise, routine metabolic rate (RMR) was calculated for each individual by subtracting SMR from the mesor of the 'metabolic circadian rhythm’ to evaluate rates of surplus energy expenditure (i.e., energetic expenditure in excess of that required for maintaining homeostasis).

Determining post-prandial responses

Following exclusion of individuals due to high background respiration (i.e., background respiration >10% of SMR, likely due to the presence of faeces within some respirometers), noisy ṀO2 measurements (i.e., R2<0.95), or lack of feeding, a total of 36 individual P. arcatus (n: winter=12, summer=14, MHW=10) and 32 L. fulvus (n: winter=10, summer=11, MHW=11) were used to determine post-prandial metabolic rates.

To account for brief periods of spontaneous activity during the SDA period, an hourly moving median was applied to post-prandial ṀO2 measurements. The post-prandial response of an individual was then defined as the period of elevated ṀO2 occurring directly after feeding until post-prandial values had returned to pre-prandial levels – defined as >8 consecutive smoothed post-prandial ṀO2 values (>1 h of measurements) that were ≤ ‘metabolic circadian rhythm’ of that individual (i.e., median + median deviation of pre-prandial ṀO2 expected for that time of day). The accumulated energy expended from the ingestion, digestion, absorption and assimilation of a meal (SDAabsolute) was calculated from the amount of oxygen consumed above the ‘metabolic circadian rhythm’ during the entire post-prandial response, and converted to units of energy with a conversion factor of 20.083 kJ L O2-1 (Schmidt-Nielsen, 1997). The proportion of AS that was occupied by the SDA response (SDArelative) was calculated by dividing the total amount of oxygen consumed above SMR by AS during the entire post-prandial response. The absolute peak of the post-prandial response (SDAabsolute peak) was defined as the highest smoothed value of ṀO2 during the post-prandial response, and was also calculated as a percentage of AS (SDArelative peak). The duration of the post-prandial response (SDAduration) was calculated as the period of time between feeding and the completion of the post-prandial response. The efficiency of the SDA response (SDAcoefficient) was calculated by dividing SDAabsolute with the amount of digestible meal energy from Pacific squid, which was 2.510 kJ g-1 based on the nutritional information provided on the packaging (Emerald Calamari, Del Mar Seafoods, Inc., Watsonville, CA, USA) and assuming a conversion factor of 4.184 kJ kcal-1 (Schmidt-Nielsen, 1997).

Statistical analyses

All statistical analyses were performed using R. Detailed descriptions of the statistical analyses used for pre-prandial (i.e., SMR, MMR, AS, RMR, as well as the mesor and amplitude of ‘metabolic circadian rhythm’), prandial (i.e., meal size), and post-prandial data (i.e., SDAabsolute, SDArelative, SDAcoefficient, SDAabsolute peak, SDArelative peak, and SDAduration) are provided in the supplementary information (e.g., R packages used, data exploration, model fitting, model selection, model checking, model inference and model predictions; see Supp. Info. 1-2).

Briefly, pre-prandial and prandial linear regression models were fit using various combinations of body mass, treatment temperature, and species, as well as the interactions between them (Table 1). Post-prandial linear regression models were fit using various combinations of body mass, meal size, treatment temperature, and species, as well as the interactions between them (Table 1). The best-fitting models were selected based on Akaike’s Information Criterion (AIC) and model assumptions were checked (Burnham & Anderson, 2002). All variables with the exception of treatment temperature, species and SDA duration were log transformed to meet model assumptions.

The main inferences from the most parsimonious models are reported throughout the text, while back-transformed model predictions are displayed in the figures. For detailed model summary outputs and extensive model inferences refer to Table 1 and Supp. Info. 1-2, respectively. Planned contrasts were conducted to investigate treatment temperature differences for an average sized individual (i.e., mean body masses of 9 or 45 g for P. arcatus or L. fulvus, respectively) and/or meal size (i.e., mean meal sizes of 0.53 or 3.63 g for P. arcatus or *L. fulvu*s, respectively) for each species. The p-values resulting from the planned contrasts were subjected to False Discovery Rate (FDR) adjustment using the Benjamini-Hochberg procedure to account for multiple testing (Benjamini & Hochberg, 1995).