Understanding how tropical corals respond to temperatures is important to evaluating their capacity to persist in a warmer future. We studied the common Pacific coral Pocillopora over 44° of latitude, and used populations at three islands with different thermal regimes to compare their responses to temperature using thermal performance curves (TPCs) for respiration and gross photosynthesis. Corals were sampled in the local autumn from Moorea, Guam, and Okinawa where mean (± s.d.) annual seawater temperature is 28.0 ± 0.9°C, 28.9 ± 0.7°C, and 25.1 ± 3.4°C, respectively. TPCs for respiration were similar among latitudes, the thermal optimum (T_opt) was above the local maximum temperature at all three islands, and maximum respiration was lowest at Okinawa. TPCs for gross photosynthesis were wider, implying greater thermal eurytopy, with a higher T_opt in Moorea versus Guam and Okinawa. T_opt was above the maximum temperature in Moorea, but was similar to daily temperatures over 13% of the year in Okinawa, and 53% of the year in Guam. There was greater annual variation in daily temperatures in Okinawa than Guam or Moorea, which translated to large variation in the supply of metabolic energy and photosynthetically fixed carbon at higher latitudes. Despite these trends, the differences in TPCs for Pocillopora were not profoundly different across latitudes, reducing the likelihood that populations of these corals could better match their phenotypes to future more extreme temperatures through migration. Any such response would place a premium on high metabolic plasticity and tolerance of large seasonal variations in energy budgets.

Overview

The study contrasted the thermal biology of corals from Moorea (-17.5°, -149.8°), Guam (13.5°, 144.7°), and Okinawa (26.6°, 127.9°), crossing 44° of latitude and different thermal environments. Pocillopora is a common genus of scleractinian corals found throughout the Indo-Pacific, and it was used to evaluate plasticity in the phenotypic response to differing temperatures that can provide insight into the capacity of corals to tolerate warming seas. Most corals in this genus (excluding the brooders P. acuta and P. damicornis) are considered “competitive” because they are efficient at resource use, they reproduce by broadcast spawning, and can dominate fore reef habitats.

Thermal performance curves (TPCs) were used to quantify the effects of temperature using dark respiration and gross photosynthesis as response variables. The thermal context of each location was characterized through the mean daily seawater temperatures over the 5 years preceding the present analyses, as calculated with daily resolution through ongoing monitoring programs in each location. Temperature was recorded with finer temporal resolution in Moorea and Guam (described below), but these values were averaged by day for the present analysis. These programs were maintained in different ways in each location, and while they broadly indicate the extent to which the environments vary among locations, the different methodologies indicate that absolute contrasts among locations should be completed with caution. In most cases, the mean daily seawater temperature was calculated using 5 years of daily records, but sample sizes were < 5 for the days of the year in which 5 replicate days were not sampled every year due to logistical constraints.

Common approaches used in all three locations

In Moorea, seawater temperature was recorded at 10-m depth on the fore reef of the north shore (using a Sea-Bird SBE39, accuracy ± 0.002°C, resolution 0.0001°C, Sea-Bird Electronics, Bellevue, WA) with the sensor < 0.1 km from the site of coral collection; values were augmented with data from a Hobo logger (U22-001, accuracy ± 0.2°C, resolution 0.02°C, Onset Computer Corp., Bourne, MA) for 136 days from 18 August 2019 (accessed 12 December 2019). The Hobo logger was utilized with the manufacturer’s calibration. Temperature was recorded at 0.0083 Hz and averaged by day before calculating the daily mean ± s.e.m. over five years (2015–2019). In Guam, seawater temperature was recorded at 0.5-m depth by the Ritidian wave rider buoy Station 52202 (https://www.ndbc.noaa.gov/, accessed 2 June 2023), which is ~ 14 km from the site of coral collection. Temperature was recorded at 0.0006 Hz and averaged by day before calculating the daily mean ± s.e.m. over five years (2018-2022). In Sesoko, daily seawater temperature was recorded manually using a 5 L bucket deployed from the lab jetty, which was ~ 1 km from the site of coral collection. Temperature records (recorded with a thermometer ± 0.1°C, 1-NM-11, Ando Keiki Co., Ltd., Japan) were used to calculate the daily mean ± SE over five years (2016–2020).

The study was completed during October (Guam) and November (Okinawa) 2022, and May (Moorea) 2023. These periods sampled during the local autumn in the northern and southern hemispheres, so that the corals experienced similar relative changes in their recent thermal histories when they were collected. The experiments were completed in near-identical ways in each location, with the same equipment, and by the same investigator (Edmunds). Some aspects of the experiments differed among locations through the constraints of local logistic, and these differences are described below.

Coral collections sampled Pocillopora spp. from a fore reef habitat (~ 4–5 m depth), and targeted colonies that morphologically resembled ‘Pocillopora verrucosa’. Recent work has highlighted the extent to which in situ gross morphology is unreliable for resolving species in this genus, so all samples were genetically identified to species. The study corals are referred to as Pocillopora spp. Five colonies were collected in each location to each supply > 9 branches (i.e., clone mates) that were prepared as nubbins. Following collection, colonies were fragmented into pieces < 60-mm tall to fit in the respiration chamber, and were attached upright onto PVC bases using Coral Glue (Ecotech, Bethlehem, PA). Nine replicates were prepared from each colony for the measurement of metabolism using independent replicates at eight temperatures, with one replicate of each host genotype tested at each temperature and one replicate preserved for genetic identification of the coral host. This design supported a test of temperature effects without being confounded by donor colony-level genetic variation. Prepared corals were maintained in a shallow tank supplied with flowing seawater at a temperature and photon flux density (PFD, from sunlight) close to that expected at the collection depth, and metabolism was measured over 4–9 days beginning ~ 4 days after collection.

The experiment exposed corals to each one of the treatment temperatures for 60–120 minutes before measuring respiration and photosynthesis. Exposure durations were standardized to prevent confounding effects of varying incubation times at each temperature. The eight temperature treatments employed were chosen to span the annual range of temperatures at the three locations, and they were tested in a random sequence, usually with one temperature each day, and independent replicates of each host genotype for each temperature. The five corals tested in any one group were staged into the treatment tank (~ 20 L) to keep exposure durations close to 60–120 minutes. The treatment tank was darkened and regulated through heating (200W Jager Heaters, Eheim AEH3617090) and cooling (with water pumped through a closed loop submerged in an ice bath) that was controlled using a digital controller (Neptune A3 Apex, Morgan Hill, CA). The water jacket surrounding the respiration chamber was connected in series with the seawater in the temperature tank, thus ensuring the metabolism was recorded at the same temperature as the treatment. Treatment temperatures were periodically monitored by hand using a certified digital thermometer (± 0.001°C, model 15-077, Fisher Scientific, Pittsburgh, PA, USA).

Preceding each measurement day, the corals were kept in darkness overnight, with darkness maintained the following morning until the corals were transferred to the treatment tank. Each day, one coral from each of the five genotypes was transferred to a darkened tank that was maintained at the temperature selected for that day. On some days, two treatments were completed sequentially in the same manner. Respiration and photosynthesis were measured in a respirometer made from a cylindrical acrylic chamber (238 mL) that was jacketed with a water bath to maintain temperature. Corals were supported on a perforated disk within the chamber, beneath which a 30-mm spin bar circulated seawater. The lid was sealed to the chamber, and oxygen (O₂) saturation in the seawater was measured with an optode (FOSPOR-R with silicon overcoat, Ocean Insight, Orlando, FL). The manufacturers specifications indicate 0.05% accuracy and 0.1% resolution at 90% O₂ saturation, with 0.0003%/hr drift. The effects of drift and accuracy were mitigated through daily two point calibrations under controlled conditions (described below).

The optode was connected to a NeoFox (Ocean Insight, Orlando, FL) spectrometer, that was temperature compensated using the manufacturer’s thermister inserted into the chamber, and operated using NeoFox Viewer 2.9 software (Ocean Insight, Orlando, FL) in a Windows environment. The probe was calibrated daily in a chemical zero (sodium sulphite and 0.01M sodium tetraborate) and air-saturated seawater (100% saturation) at the treatment temperature, and O₂ solubility in seawater was determined using gas tables (N. Ramsing and J. Gundersen, Unisense, Aarhus, Denmark). The chamber was surrounded by a darkened shroud during measurements of respiration, and photosynthesis was recorded beneath the same shroud augmented with an LED lamp (Aqua Illumination, Hydra 64) suspended above the chamber. The lamp was operated at 75% power across all wavelengths of light to provide a PFD of 890–1,058 µmol photons m^-2 s^-1 within the range of photosynthetically active radiation (PAR). PFD was measured with a cosine-corrected sensor (Li-Cor LI 192) attached to a meter (Li-Cor LI 250A).

The chamber was filled with unfiltered seawater from the seawater system used to maintain the corals in the lab, and after the coral was placed in the chamber, it was sealed and the optode inserted. Unfiltered seawater was used as the effects of coral metabolism on the O₂ saturation of the seawater in the chambers was large relative to fluxes in control trials that were attributed, in part, to the microbial flora of unfiltered seawater. Dark respiration was measured within ~ 5–10 minutes of sealing the chamber, or until a stable decline in O₂ was recorded, and was recorded at ≥ 80% O₂ saturation to avoid saturation-dependency of respiration. Following the measurement of respiration, the LED lamp was switched on at 75% power and net photosynthesis recorded for ~ 5–20 minutes or until a stable increase in O₂ was recorded to < 110% O₂ saturation. Gross photosynthesis was calculated by subtracting dark respiration from net photosynthesis with adherence to the sign convention that O₂ uptake is negative and evolution is positive. Each set of experiments included two dark controls and two light controls that were run in identical way to the trials with corals, except the chamber contained only seawater. The displacement volume of the corals was recorded to calculate the volume of seawater in the chamber, and the area of the coral tissue was measured using aluminum foil (Marsh 1970). Respiration and gross photosynthesis were normalized to coral area and time in units of nmol O₂ cm^-2 min^-1 and both were expressed on a positive scale for clarity of presentation.

A separate experiment was completed with two nubbins from each site to quantify the relationship between photosynthesis and PFD within the PAR range, and to ensure that photosynthesis in the TPC trials was measured under saturating light conditions. Trials were completed at the conclusion of the thermal experiments using the same respiration chamber and optode described above. O₂ flux was measured in darkness and eight sequentially increasing PFDs created by operating the LED lamp at power outputs ranging from 10–100%. The corals were maintained at each PFD for 2–10 minutes, or until a steady rate of change in O₂ saturation was achieved, with trials constrained to 80–110% O₂ saturation. O₂ fluxes were corrected for controls, and standardized to coral area (measured as described above) and time in units of nmol O₂ cm^-2 min^-1. Line plots of net photosynthesis versus PFD were used to evaluate the PFD at which photosynthesis reached an approximate plateau (i.e., it saturated with respect to light).

Thermal performance of Pocillopora spp. in Moorea

Measurement of coral thermal performance were completed in May 2023, and the corals (n = 5 colonies) were collected from the north shore on 26 April (-17.476°, -149.838°). Corals were prepared as nubbins and placed into a tank at 28.1 ± 0.1°C that was illuminated at 1,060 µmol photons m^-2 s^-1 (recorded with Li-Cor LI 193 sensor) using LED lamps, and supplied with a constant flow of seawater; the corals remained in this tank until trials commenced on 1 May. Two randomly selected temperatures were tested each day, and the analysis contrasted 22.1 ± < 0.1°C, 25.4 ± < 0.1°C, 27.1 ± < 0.1°C, 28.3 ± < 0.1°C, 30.1 ± 0.1°C, 31.9 ± 0.1°C, 32.6 ± < 0.1°C, and 33.9 ± 0.1°C (mean ± s.e.m., N = 4–7), and photosynthesis was measured at 995 µmol photons m^-2 s^-1 (recorded with a Li-Cor LI 192 sensor).

Thermal performance of Pocillopora spp. in Guam

Measurements of coral thermal performance were completed in October 2022, and the corals (n = 5 colonies) were collected in front of the closed Tanguisson Power Plant on 20 October (13.543°, 144.807°). Corals were prepared as nubbins and placed in a tank at 30.1 ± 0.2°C that was illuminated by natural sunlight screened to a maximum of 825 µmol photons m^-2 s^-1 (recorded with a Li-Cor LI 192 sensor), and supplied with a constant flow of seawater; the corals remained in this tank until trials commenced on 24 October. One randomly selected temperature was tested daily, and the analysis contrasted 22.9 ± 0.1°C, 26.1 ± 0.1°C, 28.1 ± < 0.1°C, 29.0 ± < 0.1°C, 30.9 ± 0.1°C, 32.7 ± < 0.1°C, 33.9 ± 0.1°C, and 35.5 ± 0.1°C (mean ± s.e.m., N = 6–15), and photosynthesis was measured at 890 µmol photons m^-2 s^-1 (recorded with a Li-Cor LI 192 sensor).

Thermal performance of Pocillopora spp. in Okinawa

Measurements of coral thermal performance were completed in November 2022, and the corals (n = 5 colonies) were collected off Sesoko Island on 24 November (26.629°, 127.858°). Corals were prepared as nubbins and placed in a tank at 23.8 ± 1.6°C that was illuminated by natural sunlight screened to a maximum of 326 µmol photons m^-2 s^-1 (recorded with a Li-Cor LI 192 sensor), and supplied with a constant flow of seawater; the corals remained in this tank until trials commenced on 27 November. One randomly selected temperature was tested daily, and the analysis contrasted 21.0 ± 0.1°C, 26.8 ± < 0.1°C, 28.3 ± 0.1°C, 29.9 ± 0.1°C, 31.9 ± 0.1°C, 32.4 ± < 0.1°C, 33.2 ± < 0.1°C, and 35.0 ± < 0.1°C (mean ± s.e.m., N = 7–11), and photosynthesis was measured at 1,058 µmol photons m^-2 s^-1 (recorded with a Li-Cor LI 192 sensor).

Thermal Performance Curves (TPC)

TPCs were estimated for respiration and photosynthesis with best-fit relationships selected from: (i) a symmetrical Gaussian, and (ii) a Gaussian-Gompertz. These relationships were selected because of their common use in preparing TPCs, the principles of parsimony and the utility of simpler models, and our objective of comparing the temperature at which the dependent variable was maximized (T_opt) and curve widths among locations using eight temperature treatments. Relationships were fit using least squares non-linear regressions, with model selection using AIC_c. The constants mathematically defining the TPCs (see equations 1 and 2 below) were obtained from the parameter values selected by the curve fitting routines for the best-fit relationships. Since the objective was to compare TPCs among corals from different latitudes, a single functional relationship was selected as the best compromise for corals from the three locations. Curves were independently fit to the five coral genotypes from each site, and to the data pooled among coral genotypes by site. The two approaches contrasted the value of less robust fits for each individual coral (with eight data points) versus more robust fits by pooling results among corals within each location (with ~ 40 data points).

(i) Symmetrical Gaussian

                                                                        equation 1

where Rate = dependent variable (dark respiration or gross photosynthesis, nmol O₂ cm^-2 min^-1), Max = maximum fitted value of respiration or photosynthesis, Temp = independent variable (temperature, °C), T_opt = optimum temperature on fitted relationship for dark respiration or gross photosynthesis, and alpha (α) = measures half the width of the thermal performance curve (°C). In the best-fit relationships, T_opt represents the temperature at which the dependent variable is highest, and the terminology reflects historic precedence rather than the inference that the maximum rate is optimal with respect to fitness.

(ii) Guassian-Gompertz

                                        equation 2

where Rate = dependent variable (respiration or photosynthesis, nmol O₂ cm^-2 min^-1), Max = maximum fitted value of respiration or photosynthesis, Temp = independent variable (temperature, °C), T_opt  = optimum temperature on fitted relationship for dark respiration or gross photosynthesis, alpha (α) = slope of ascending portion of the relationship (nmol O₂ cm^-2 min^-1 °C^-1), and beta (β) = slope of descending portion of the relationship (nmol O₂ cm^-2 min^-1 °C^-1).

Genetic analysis and identification of genetic lineages

Laboratory protocols for genetically identifying the Pocillopora samples slightly differed among islands as outlined in Supplementary Information. Briefly, genomic DNA was extracted from preserved samples and the mtORF barcoding region was amplified using PCR. Moorea samples were amplified using the FATP6.1 and RORF primers and sequenced in the forward direction with the FATP6.1 primer at Florida State University on an Applied Biosystems 3730 Genetic Analyzer with Capillary Electrophoresis. Guam samples were amplified using newly designed primers (Poc_MT_ORFf1: 5'-TGCAAAATTTAAGTAATGTGGGTTT-3' and Poc_MT_ORFr1 5'-CACCTGGAGGTGTTTCTACCTT-3') and sequenced at Epoch Life Sciences (Missouri City, TX) in both directions with diluted PCR primers on an ABI3730XL sequencer. Okinawan samples were amplified with the primer pair FATP6.1/RORF and by Macrogen Japan (http://www.macrogen-japan.co.jp/) using the same primer pairs as for the PCR. Forward and reverse sequences from Guam and Okinawa were aligned using MAFFT v7.49 in GENEIOUS v.9.1.8 (Biomatters). Forward sequences from Moorea were aligned using Clustal Omega in GENEIOUS PRIME 2023.2.1. Individual consensus sequences were generated and aligned with forward sequences from Moorea and previously published mtORF haplotypes. Alignments were manually trimmed to 927bp and then quality-trimmed using G-Blocks online (http://phylogeny.lirmm.fr) with default settings but allowing gap positions in final blocks, reducing the alignment length to 817 bp. Alignments were re-imported into GENEIOUS and phylogenetic analyses were conducted using Bayesian and Maximum Likelihood algorithms.

Maximum likelihood analyses were conducted using RAxML 8.2.8 as implemented in Geneious. A unique GTR model of sequence evolution was used with corrections for a discrete gamma distribution (G) for site-rate heterogeneity (GTRGAMMA) with a proportion of invariable sites and 1000 bootstraps.

In addition, Bayesian inference analyses were carried out with MrBayes 3.2.6. Analyses were conducted with the default GTR model of sequence evolution and a gamma-shaped rate variation with a proportion of invariable sites. MrBayes analyses started with random trees, default priors and two runs, each with 4 Markov chains. Convergence diagnostics were assessed in Geneious, and the runs were allowed to proceed until the average deviation of split frequencies reached < 0.01 (~ 2 M generations).

Naming conventions were used to associate haplotypes with nominal species names. Since both P. meandrina and P. grandis share the mtORF haplotype 1 sequence, we differentiated samples identified as haplotype 1 using a PCR-based restriction fragment length polymorphism (RFLP) gel assay of the histone 3 marker using PocHistone primers, Xho1 restriction enzyme, and protocols. The mtORF and Histone markers have previously been validated as suitable species-level markers to delineate Pocillopora species, based on whole mitochondrial genomes and genome-wide sequencing. The nuclear genomes of mtORF haplotypes 1 (P. meandrina) and 8 are indistinguishable, despite being distinct mitochondrial lineages. Therefore, colonies identified as haplotype 8 were included with colonies identified as haplotype 1 in P. meandrina, which was also justified based on their similar within-reef distribution patterns and responses to heat waves.

Statistical approaches. TPCs were fitted using least squares linear regression conducted with XLSTAT statistical software (version 2023.1.3, Lumivera, Denver, CO) that operates within an Excel environment (version 16.72, Microsoft, Redmond, WA). Since non-linear least squares regressions are sensitive to the choice of starting values for parameters, we thoroughly checked the sensitivity of the final parameter estimates to different starting values. As an additional check, we also compared our parameter estimates to those generated by the nls.multstart package in R, and got similar results. For models that converged, the best-fit options were selected by AIC_c. Temperature was the independent variable, and respiration and gross photosynthesis were dependent variables. Curve parameters (described above) were compared among latitudes using one way ANOVA using coral host genotypes as replicates.