Thermal sensitivities of respiration and protein synthesis differ among larval families of the Pacific oyster, Crassostrea gigas

Animals differ in their range of temperature tolerances throughout their life-history, and across latitudes and ocean depths (Schmidt-Nielsen, 1997; Pörtner, 2002; Somero et al., 2017). Early developmental stages, including larval forms, are often more vulnerable to temperature and other environmental perturbations than are adult stages (Pörtner and Knust, 2007; Dupont et al., 2010; Ross et al., 2011; Dahlke et al., 2020). Mechanisms for varied thermal tolerance across life stages and life spans are often related to bioenergetics of stress tolerance (Hand and Hardewig, 1996; Sokolova et al., 2012; Sokolova, 2021). While many studies focus on increases in metabolic rate in response to temperature increase, a major question remains as to which physiological systems are responsible for setting thermal limits (Somero, 2010). Models of oxygen supply and demand for marine animals currently predict an imminent mass extinction due to the acceleration of ocean warming (Penn and Deutsch, 2022). Understanding the mechanisms of physiological plasticity is a basis for better predictions of resilience to environmental change (Melzner et al., 2009; Seebacher et al., 2015; Tresguerres, 2016). Phenotypic variation resulting from genetic variation will ultimately dictate long-term evolutionary responses (a focus of the present study).

Regulation of respiration rate (energy supply) and protein synthesis rate (energy demand) are two critical processes for maintaining energy balance in an organism. Protein synthesis is one of the most energetically costly processes in cellular biology (Siems et al., 1992; Wieser and Krumschnabel, 2001; Cherkasov et al., 2006). For developmental stages of marine invertebrates, over 50% of the total pool of adenosine triphosphate (ATP) can be allocated to the single process of protein synthesis (e.g. white sea urchin Lytechinus pictus: Pace and Manahan, 2006; purple sea urchin Strongylocentrotus purpuratus: Pan et al., 2015; Pacific oyster Crassostrea gigas: Frieder et al., 2018). The thermodynamic cost of synthesizing a unit-mass of protein for larvae of C. gigas has previously been determined and, notably, this cost is fixed and independent of environment, genotype and phenotype (Lee et al., 2016).

Proteins are constantly degraded and recycled in organisms, and the functioning of protein synthesis is essential for maintaining homeostasis for growth and development under changing environmental conditions (Hawkins, 1991; Tsukada and Ohsumi, 1993; Vavra and Manahan, 1999; Frieder et al., 2018). Environmental stress results in changes in the allocation of energy in marine invertebrate larvae. In larvae of the sea urchin S. purpuratus, for instance, significant increases were measured in ATP allocation to support the cost of increased protein synthesis and turnover under conditions of ocean acidification (Pan et al., 2015). Larvae of the oyster C. gigas also had increased rates of protein synthesis and turnover under conditions of ocean acidification (Frieder et al., 2018). In that study, Frieder et al. (2018) reported different strategies used by larvae to support the increased demand for protein synthesis – reallocating a larger portion of the available ATP pool, or increasing metabolic rate to increase available ATP. Changes in the bioenergetic allocation of ATP were further analyzed for C. gigas larvae, where thermal sensitivity (measured as Q₁₀ value) for respiration was reported to have a lower Q₁₀ of ∼2 than the Q₁₀ for protein synthesis of ∼3 (Pan et al., 2021). The differential thermal sensitivities of supply and demand of the ATP pool have implications for larvae experiencing different scenarios of rising temperature, as the energy demand of biosynthesis would increase at a greater rate (higher Q₁₀ value) than the ability to supply energy. This imbalance would result in an increased – and eventually unsustainable – allocation of available ATP to protein synthesis. Of note is that the disproportionate increase in the energy demand of protein synthesis is a result of the higher thermal sensitivity of the rate of protein synthesis per se, not a change in the energy cost per unit-mass of protein synthesized (the latter cost is temperature independent: Lee et al., 2016). As with many evolutionary scenarios there may, however, be sufficient standing genetic variation to provide physiological variance within a population of a given species that could confer resilience to environmental change. This is the focus of the present study.

Studies of variation in respiration have mostly focused on environmental influences. Dissecting the complex contributions of ‘nature versus nurture’ is challenging in animal developmental physiology (Applebaum et al., 2014). For species with well-developed genetic resources, there is genetic variance in respiration and metabolic responses to temperature (Einum et al., 2019; Fossen et al., 2019). Among different lines of the fruit fly Drosophila melanogaster, adaptation to temperature change alters the relationship between metabolism and growth rate (Williams et al., 2016), with different lines having complex interactions between the thermal sensitivity of metabolic rate and heat tolerance (Berrigan and Partridge, 1997; Folk et al., 2007). In the present study, a series of crosses were undertaken with (single) male and female, pairwise matings of the Pacific oyster, Crassostrea gigas, to study the temperature physiology of 21 distinct larval families. The genealogical record of each parent used in crosses was confirmed by genotyping, based on analyses of single-nucleotide polymorphisms (SNPs). We report that several larval families have the physiological capacity to increase respiration rate with a Q₁₀ of 3, to match the same high Q₁₀ value for thermal sensitivity of protein synthesis (previously reported by Pan et al., 2021). In addition, we present a novel finding of an alternative mechanism to reduce the high cost of living under thermal stress – a lowering of the thermal sensitivity (Q₁₀ less than 2) of protein synthesis. We summarize these findings in bioenergetic models of ATP trade-offs that have implications for resilience to rising temperature.

Parental samples for all individuals used in crosses were genotyped by high-resolution melting of SNP-containing amplicons, following the methods in Sun et al. (2015). These mapped genetic markers in the genome are referred to as composite SNP markers (or simply ‘markers’), because most amplicons contain multiple SNP sites and are multi-allelic in contrast to single, bi-allelic SNP markers. Individual adults in the present study were genotyped at 45 different markers (Sun et al., 2015). For genotyping in the present study, to confirm parentage, over 1700 PCRs were conducted from tissue samples taken from adult broodstock, each tested at 45 SNP markers for determination of genotype. Gravid adult broodstock of the Pacific oyster, Crassostrea gigas (Thunberg 1793), were obtained from our transgenerational (‘egg to egg’) breeding program to establish different lines from wild populations (see similar approaches in Hedgecock et al., 1995; Pace et al., 2006; Hedgecock and Davis, 2007). To confirm parentage, each animal was briefly treated with a solution of MgSO₄ (Epsom salt, added to seawater at 73 g l⁻¹) to relax the adductor muscle and allow a sample of mantle tissue to be taken. Tissue was preserved in 70% ethanol and DNA was extracted using standard protocols (DNeasy Blood & Tissue, Qiagen, Germantown, MD, USA).

High-resolution melting assays were based on amplification of SNP-containing sequences by PCR, followed by one cycle of cooling and heating to determine the melting temperature of sequences containing different SNPs (Sun et al., 2015). Each reaction had a final volume of 10 μl, consisting of 1 µl DNA, 0.35 µl forward and reverse primers, 1.2 µl MgCl₂, 2.45 µl molecular grade H₂O and 5 µl of the high-resolution melting Master Mix (Roche Diagnostics, Indianapolis, IN, USA). Target amplicons were amplified by PCR in 384-well plates, using a LightCycler 480 thermocycling instrument (Roche Diagnostics) using reaction conditions given by Sun et al. (2015). Following PCR amplification, products were denatured at 95°C for 1 min, cooled to 40°C, and then heated gradually, at a rate of 0.05°C s⁻¹, to a target temperature of 95°C. Fluorescence (SYBR Green) was monitored in each well during heating. Genotypes were determined by examining the normalized melting curves, using LightCycler 480 software (v1.5) Gene Scanning 384-II module.

Parental genotypes were compared with the genotypes of their putative parents (i.e. grandparents of the larval families) using CERVUS 3.0 (Kalinowski et al., 2007). The wild-type broodstock used as the foundation population for the breeding program was used as the base population for simulations to determine critical thresholds for parentage analyses (Kalinowski et al., 2007). The proportion of loci typed was set to 1.00 and the genotyping error rate set at the default value of 1%. Determination of parentage was based on 95% confidence levels for assignment to the most likely candidate parent pair.

Eggs and sperm were removed directly from the gonads of genotyped, gravid C. gigas. To avoid cross-contamination, clean Pasteur pipettes were used to obtain gametes from each adult. Eggs were placed in filtered seawater for enumeration. Aliquots of sperm from each male were added to separate 1.5 ml micro-centrifuge tubes and kept ‘dry’ (no seawater added) at 4°C. Once the number of eggs was known, seawater was added to the sperm and approximately 1 ml of this sperm suspension was added to 1 liter of seawater containing eggs. Fertilization success was confirmed by microscopic observation of first polar body formation (∼20 min).

Table 1 lists the genotyped males and females selected for crosses to generate 21 larval families. Of the 21 larval families produced, 10 represented first-generation (G₁) families from sibling parents and the remaining 11 families represented G₀ families from unrelated parents (Table 1). For each sire×dam cross, the combination of each letter and number designates an individual male or female. Depending upon the fecundity (and sperm and egg quality) of each individual male and female used for crosses, some larval families were produced by using the same genotyped parent or sibling adult from the same parental line. This is indicated with repetition of the same letter and number, when the same individual adult was used to generate two families, producing half-sibling offspring larvae. For example, larval family 1 was produced by crossing male A and female B in a sire×dam cross. Male A was also used to produce half-sibling offspring for larval family 4, by crossing male A with a different female, female E. Hence, larval families 1 and 4 have a relatedness (co-ancestry) of 0.25. Larval families 2 and 3 have a relatedness of 0.375 because females D.1 and D.2 were siblings and were crossed with the same parent, male C. Larval families 1–4 were all generation G₀. Larval families 6 and 15 are examples of generation G₁ larval families, produced from four separate parents that were bred from the same pedigreed line (larval family 6 from G.1×G.2; larval family 15 from G.3×G.4). See Lynch and Walsh (1998) for calculations of genetic coefficients for relatedness.

Fertilized eggs from each of the 21 sire×dam crosses were stocked at an initial number of 10 eggs ml⁻¹ in a series of 200 l culture vessels. All larval families were reared at 25°C in a custom-designed, large-scale culturing facility at the Wrigley Marine Science Center (University of Southern California) located on Santa Catalina Island. Seawater used to culture all larval families was obtained directly from the Pacific Ocean surrounding Catalina Island. Ambient seawater was filtered (0.2 µm pore -size) and heated with inert (titanium) heat exchangers to the desired temperatures. During larval rearing, seawater in each culture vessel was replaced every 2 days by removing larvae from culture vessels and gently washing them on size-appropriate mesh sieves with filtered seawater.

During experiments, temperature was measured every 30 min (HOBO data logger U12, Onset Computer Corp., Bourne, MA, USA) in selected 200 l culture vessels. A total of over 2400 such temperature measurements were made, yielding a mean (±s.e.m.) temperature of 24.9±0.01°C. At 25°C – a routine temperature used to grow larvae of C. gigas – the first larval feeding stage (D-hinge veliger larva) developed by 2 days. The alga Tisochrysis lutea (T-ISO, CCMP1324, National Center for Marine Algae and Microbiota, Bigelow, ME, USA) (Helm and Bourne, 2004) was added to each larval culture vessel at 30,000 cells ml⁻¹. Algal ration was later increased to 50,000 cells ml⁻¹ as larvae grew (per standard culturing protocols for larvae of C. gigas, as described by Breese and Malouf, 1975; Helm and Bourne, 2004).

Growth and survivorship of individuals from each larval family were evaluated throughout experiments. Survivorship was based on enumerations of known aliquots of larvae, re-suspended in a volume of 200–500 ml (dependent on the number of larvae collected from each culture vessel). Replicate (at least 3) enumerations of larvae were completed for each larval family to yield a coefficient of variation of less than 10%. Known numbers of larvae were then placed in a range of experimental containers used for the measurement of respiration, amino acid transport and protein synthesis rates. Growth rate was measured from the shell lengths of larvae over time. Larvae were photographed at 40× magnification in Sedgewick Rafter counting chambers, with 1–2 drops of ethanol added for immobilization. At least 50 larvae at specific time points selected within each culture vessel were measured for length, using digital assay software for shell length analyses (ImageJ, National Institutes of Health, Bethesda, MD, USA). A total of approximately 5000 larval sizes were measured during the course of the larval culturing required for these experiments. Protein content of selected larval families was also assayed (Bradford, 1976; as modified by Jaeckle and Manahan, 1989).

For analysis of the temperature sensitivity (Q₁₀ value) of respiration, a total of 41 Q₁₀ values were measured using a total of 1600 individual micro-respiration chambers. To obtain the respiration rate, the amount of oxygen in each chamber was measured at least 4 times during a time course experiment (Fig. 1A–C). Respiration rates of larvae were measured as oxygen consumption over time, using optode technology (Witrox-1 Oxygen Meter, Loligo Systems, Viborg, Denmark). This technology was previously calibrated and determined to be consistent with respirometry measurements made with polarographic oxygen sensors for C. gigas larvae (Pan et al., 2021). A known number of larvae (between 250 and 400, depending on size) was placed in a series of replicate glass, sealed micro-biological oxygen demand respiration chambers. Each respiration chamber (up to 40 different respiration chambers were used for a given assay) was individually calibrated for volume (between 400 and 700 µl). Each respiration chamber contained an internal 2 mm diameter sensor spot (Presens, Regensburg, Germany), to which a fiber optic cable was non-invasively and manually held near the sensor spot for each measurement of oxygen. This permitted a series of time course assays for the rate of oxygen depletion by larvae in each individual respiration chamber. For each larval family, oxygen consumption was measured with at least seven replicate chambers over the range of selected temperatures (Fig. 1D–F: 14–28°C) used to determine Q₁₀ value. This temperature range was chosen as it is representative of the natural environments where larvae of C. gigas are known to be present in the water column (Pacific Northwest: Quayle, 1988).

ANCOVA linear regressions were used to test whether a measured Q₁₀ value for respiration within a larval family changed with growth, over a range of developmental sizes spanning 80 µm (newly formed veliger larva) to 280 µm (nearing metamorphic competence) in shell length (size range shown in Fig. 2). ANOVA with post hoc Tukey tests were used to test for differences in respiration Q₁₀ values among the 21 larval families. ANCOVA was used to test for differences between G₀ and G₁ larval families. For statistical tests, values of total corrected degrees of freedom are provided with the F-statistic.

For analysis of the temperature sensitivity of protein synthesis, two larval families were studied in detail (families 9 and 21). These larval families were chosen because family 9 had an average respiration Q₁₀ of 2.3, in contrast to family 21 which had a lower respiration Q₁₀ of 1.6 (Table 2).

Amino acid transport rates were determined by in vivo time course assays with [¹⁴C]glycine as the substrate added to seawater. Previously selected for studies of C. gigas larvae (Lee et al., 2016; Pan et al., 2018, 2021), [¹⁴C]glycine was chosen as the radioisotope tracer of protein synthesis based on the composition of the intracellular free amino acid pool of C. gigas. Each time course assay was conducted in duplicate 20 ml sealed glass vials, each containing 10,000 larvae in 10 ml of filtered seawater. An addition of 74 kBq of [¹⁴C]glycine (Perkin Elmer, Boston, MA, USA) was supplemented with non-radioactive glycine (Sigma-Aldrich, St Louis, MO, USA) to provide a total glycine concentration of 10 µmol l⁻¹ in seawater prior to the start of each transport assay. A 15 µl aliquot of seawater was taken at the start of each assay to quantify the specific activity of [¹⁴C]glycine in each experimental assay vial. The specific activity was used to correct the amount of [¹⁴C]glycine to total glycine transported from seawater by larvae. A series of six 1 ml samples (each containing 1000 larvae) was removed from each replicate vial every 8 min during a 48-min time course assay. Samples were transferred onto an 8 µm pore-size (Nuclepore, GE Healthcare, Pittsburgh, PA, USA) membrane filter and, under gentle vacuum, the excess radioactivity in seawater was removed with additional washes of filtered seawater. Filters holding larvae were then placed in 1.5 ml micro-centrifuge tubes and frozen at −80°C until further processing.

Frozen larvae were processed by adding 400 µl of deionized water (Nanopure, Barnstead^TM Nanopure Bioresearch Deionization System, Dubuque, IA, USA) to each sample tube kept on ice, followed by homogenization with a Vibra-cell ultrasonic processor fitted with a 3 mm micro-probe (Sonics & Materials, Inc., Newton, CT, USA) for two 15 s intervals. A 15 µl aliquot of the homogenate was placed directly into a scintillation vial containing liquid scintillation counting fluid (Ultima Gold^TM, Perkin Elmer). Total radioactivity was counted with appropriate quench correction (Beckman Coulter Liquid Scintillation Counter, Model 6500). The rate of glycine transport by larvae (Fig. 3A) was calculated from the slope of the accumulation of [¹⁴C]glycine over time, corrected for the starting specific activity of [¹⁴C]glycine in each assay vial and for the number of larvae per sample. Duplicate time course assays at each of four temperatures were used to calculate the Q₁₀ value for glycine transport. An ANCOVA was used to compare the Q₁₀ values for glycine transport within and between families. This analysis was used to compare the effect of temperature on the acquisition of substrate (amino acid transport) with the metabolism of substrate (amino acid incorporation into protein).

Protein synthesis rates were calculated from data obtained during the corresponding in vivo time course assays of the transport rate by larvae of [¹⁴C]glycine (Fig. 3A). From each of the 400 µl larval homogenate samples prepared from each time point of a time course assay as described above, a 300 µl aliquot was used to determine the amount of [¹⁴C]glycine that was incorporated into protein. An additional 42 µl of homogenate was used to determine the intracellular specific activity of [¹⁴C]glycine in the extracted free amino acid pool of larvae (Fig. 3B). The rate of protein synthesis was calculated by combining (1) the rate of incorporation of [¹⁴C]glycine into the trichloroacetic acid (TCA)-precipitable protein fraction (Fig. 3C) with (2) the rate of change in the intracellular specific activity of [¹⁴C]glycine in the free amino acid pool (Fig. 3B).

The intracellular specific activity of [¹⁴C]glycine in the free amino acid pool of larvae was determined by high-performance liquid chromatography. For extraction of the free amino acid pool from larval tissues, a 42 µl aliquot of the original (400 µl) larval homogenate (as prepared above) was combined with 100 µl chromatographic-grade 100% ethanol to yield a final concentration of 70%. Samples were incubated at 4°C for at least 24 h and then centrifuged for 5 min at 12,000 rpm prior to chromatographic analysis (for further details see Lee et al., 2016). In brief, the chromatographic separation and quantification of the amount (moles) of glycine in each sample (100 µl injection chromatographic sample loop) was combined with quantification of the amount of radioactivity in the glycine chromatographic peak (collected by fraction collector; Model FC 203B, Gilson Inc., Middleton, WI, USA). These data were used to calculate the rate of change of the intracellular specific activity of [¹⁴C]glycine in the free amino acid pool (Fig. 3B), required for calculations of absolute rates of protein synthesis. These methods were used to calculate protein synthesis rates for different larval families, with the full suite of chromatographic and radiochemical measurements repeated at each different temperature.

The amount of [¹⁴C]glycine incorporated into protein was measured by taking 300 µl of the larval homogenate with an equal amount of 10% trichloroacetic acid added to yield a final concentration of 5%. After a 30 min incubation at 4°C, homogenates were vortexed and placed on a GF/C glass microfiber filter (Whatman, Tisch Scientific, North Bend, OH, USA) to retain TCA-precipitable protein. Under gentle vacuum, the [¹⁴C]glycine not incorporated into the TCA-precipitable protein fraction was removed by additional washes with 5% TCA and methanol. Each filter containing the protein precipitate was transferred into a liquid scintillation vial for subsequent determination of the amount of radioactivity. The slope of the increase with time of the rate of incorporation of [¹⁴C]glycine into protein was measured in duplicate assays for a given larval family at each temperature tested (Fig. 3C).

The rates of respiration (moles of oxygen) and protein synthesis (grams) were converted to energy equivalents (joules: Gnaiger, 1983; Lee et al., 2016). This conversion allowed the calculation of energy allocation strategies used by different larval families to support increased protein synthesis rates in response to increased temperature.

Protein and lipid are the major biochemical components in C. gigas larvae (Moran and Manahan, 2004; Pan et al., 2018). A value of 484 kJ mol⁻¹ O₂ was used to convert moles of oxygen to joules (mean value of 441 kJ mol⁻¹ O₂ for lipid; 527 kJ mol⁻¹ O₂ for protein). For physiological energetics, these oxyenthalpic equivalents for combustion of catabolized biomass are routinely used (Gnaiger, 1983). An additional justification for the use of 484 kJ mol⁻¹ O₂ is that the measured calorimetric–respirometric ratio for C. gigas larvae under normoxic conditions is 478 kJ mol⁻¹ O₂ (Hand, 1999; see also Haws et al., 1993), a value that is within 1% of the value (484 kJ mol⁻¹ O₂) used here to calculate energy allocation. For protein synthesis in C. gigas larvae, the energy to synthesize a unit mass of protein is 2.1±0.2 J mg⁻¹ protein synthesized (mean±s.e.m.) (Lee et al., 2016). Of importance for calculating energy allocation models in the current study is that this cost of protein synthesis is independent of temperature, growth rate (size) and genotype.

The project goal was attained by generating multiple larval families of Crassostrea gigas for physiological analysis. To ensure that each larval family was distinct (single sire×dam crosses), genetic relatedness was confirmed by genotyping each set of parents (Table 1). The larvae in all 21 families grew under the culturing conditions utilized for food ration and 25°C rearing temperature. Of the fertilized eggs that developed at 25°C to the first larval feeding stage at day 2 (57 million D-hinge veliger larvae across all 21 larval families tested), a mean of 45% survived to day 11. This value compares well to the typical survival rate during the larval stage of oysters of approximately 50% (Helm and Bourne, 2004). Additionally, larvae from 17 of the 21 families underwent successful metamorphosis to the early juvenile stages, indicating the general viability and health of the larvae studied.

Fig. 1A–C illustrates an example of the level of replication used to measure rates of oxygen consumption of larvae from three different families (families 2, 9 and 21). The primary respiration data obtained for these three larval families span a broad range of Q₁₀ values and illustrate the size independence of respiration Q₁₀ values (Fig. 2). In addition, data from larval families 9 and 21 were used to calculate the model illustrated in Fig. 4, highlighting the differential allocation of energy to protein synthesis with rising temperature. Respiration rates of larvae were measured in a series of seven independent micro-biological oxygen demand respiration chambers. Also illustrated in Fig. 1A–C are blank control measurements, showing the insignificant depletion of oxygen in respiration chambers when no larvae were present (filtered seawater only) (e.g. Fig. 1A: combined ANOVA of the two control regressions, P=0.9). All of the independent rate measurements of respiration (from each respiration chamber) shown in Fig. 1A–C were obtained at a single temperature. Larvae used in these measurements from each of these three different families were of similar size (mean±s.e.m., family 2: 153±4.5 µm, family 9: 148±1.3 µm, family 21: 146±1.4 µm). By ANOVA, these three larval sizes were not significantly different (F_2,153=2.34, P=0.1). Those replicate measurements of respiration within a family were pooled and graphed as a single point (circled in Fig. 1D–F) for a specific temperature (with corresponding measurement error). These analyses of respiration rates were repeated at temperatures of 14, 18 and 28°C for each larval family (Fig. 1D–F).

Rates of oxygen consumption were then used to calculate Q₁₀ values for the respiration rates for each larval family. Values of Q₁₀ were calculated from the log value of the respiration rate at each of the four temperatures assayed. For instance, larval family 2 (Fig. 1D) had a high Q₁₀ value (±s.e.) of 2.8±0.90 (N=28, R²=0.90), larval family 9 (Fig. 1E) had an intermediate Q₁₀ value of 2.4±0.22 (N=28, R²=0.99) and larval family 21 (Fig. 1F) had the lowest Q₁₀ value of 1.7±0.20 (N=28, R²=0.99). ANCOVA for these three Q₁₀ values showed significant differences (F_2,78=15.32, P<0.0001). Hence, there was a significant difference (Tukey's comparison) between these three larval families with respect to temperature, with family 2 having a greater sensitivity to temperature increase than family 9, and both families having a greater sensitivity than family 21.

Combined for all 21 families studied, a total of 41 Q₁₀ values were measured for larvae reared to different sizes (shell length range from 80 to 280 µm). This experimental analysis was designed to address the question of whether larval growth and size alter the physiological response to temperature within specific larval families. As detailed below, statistical analysis revealed that larval size did not change the Q₁₀ value for respiration within a family. Importantly, there were significant differences (statistical values below) among larval families with respect to the response to temperature.

The 21 larval families that were assayed during the course of this study were ranked from the highest Q₁₀ value of 3.5 for larval family 1 to the lowest Q₁₀ of 1.6 for larval family 21 (Table 2). For the latter family, for instance, four different larval sizes were assayed for Q₁₀ respiration values (mean±s.e.m. larval shell length columns i–iv, respectively: 188±3.4, 146±1.6, 146±1.4 and 202±2.2 µm). The two mean values of 146 µm for larval family 21 represent larvae of similar size, but differing in age by 2 days (size-fractioned, see Materials and Methods, ‘Respiration rate and Q₁₀ values’). Complete details of the range of larval sizes tested within each larval family are given in Fig. 2. ANOVA of all data for the 21 larval families and all larval sizes assayed revealed that size did not change the Q₁₀ value within a family (F_1,39=0.56, P=0.46). There was, however, a significant difference for Q₁₀ values among families (F_20,40=7.03, P<0.001), with larval family accounting for 87% of the variance. This level of difference explained by family was calculated based on the ratio of the sum of squared deviation from the mean for the family effect (7.1) to the total sum of squared variance (8.2).

Tukey's pairwise comparison (α=0.05) following ANOVA identified the larval families with the highest and lowest response of respiration rate (Q₁₀ values) to temperature. This clustering of families is shown in Fig. 2. Larval families 1–4 (Fig. 2, upper ellipse) had the greatest sensitivity to temperature, compared with all other families. Each of the respective Q₁₀ values for these four larval families – representing a wide range of different larval sizes – was by 95% confidence interval not different from a Q₁₀ of 3.0 (calculated from the s.e. of the slope regression analysis for each Q₁₀ value, Table 2). In contrast, larval family 21 had the lowest respiration Q₁₀ of 1.6 (Table 2), illustrated as the lower ellipse in Fig. 2. The other 16 larval families (families 5–20; see Fig. 2) had Q₁₀ values for respiration that were less than the Q₁₀ value of 3.0 for larval families 1–4 and were greater than the Q₁₀ value of 1.6 for larval family 21. The detailed oxygen consumption data for larvae of the same size in families 2, 9 and 21 (Fig. 1), with different Q₁₀ values, are indicated by arrows on Fig. 2. An ANCOVA regression of 41 respiration Q₁₀ values, for all 21 families, and 11 G₀ and 10 G₁ generations revealed a significant effect of generation (F_1,40=8.05, P=0.007). There was a marginally significant interaction between family and generation (F_1,40=4.5, P=0.04) (least square mean±s.e.m. Q₁₀ value: G₀=2.4±0.04, G₁=2.2±0.06; P=0.04).

In summary, Q₁₀ values for respiration within a family did not change during larval growth, as at any given size the Q₁₀ values were similar. Significant differences in Q₁₀ values among families and between generations were evident, which has important implications for biological variation in response to temperature (Figs 1 and 2, Table 2).

To address a possible relationship between respiration rates for larval families that were (or were not) successful in reaching metamorphosis as late-stage larvae, respiration rates were compared for larval families that completed metamorphosis with those larval families that did not complete metamorphosis. To correct for size-dependent effects in respiration, larvae of a similar size were compared (see size range in Fig. 2 for larvae of approximately 150 µm in shell length: 137–163 µm). This dataset contains a total of 11 larval families for which respiration rates were measured for larvae of ∼150 µm. Of these 11 larval families, eight were successful in metamorphosing to the early juvenile stage (larval families 2, 3, 10, 11, 17, 18, 19, 21) and three larval families were not (larval families 4, 9, 13). A comparison of respiration rates measured at the rearing temperature of 25°C shows that larvae from families that completed metamorphosis had similar (t-test, P=0.121) respiration rates (mean±s.e.m. 68.7±5.9 pmol O₂ h⁻¹) to larvae that did not complete metamorphosis (51.0±6.8 pmol O₂ h⁻¹), i.e. there was no relationship between larval respiration and successful metamorphosis.

The change in the rate of protein synthesis with increasing temperature was measured for larval families 9 and 21 (Fig. 3). Family 9 was chosen because it had an average Q₁₀ value of 2.3, the same as the overall mean Q₁₀ of 2.3±0.1 for all 21 families (Table 2). Family 21 was chosen because it had the lowest Q₁₀ value for respiration (Q₁₀ of 1.6, from Table 2). Notably, larvae from each of these two families had similar rates of glycine transport from seawater, averaging 1.7±0.06 pmol glycine larva⁻¹ h⁻¹ (mean±s.e.m.) (Fig. 3A, at 28°C). By ANCOVA, this transport rate of [¹⁴C]glycine from seawater by larvae was similar for the duplicate assays within a larval family and between the two families (ANCOVA F_1,16=0.24, P=0.6). The rates of protein synthesis, however, differed significantly (statistical values below) between these two larval families. Fig. 3B,C illustrates the measurement steps used to determine the rate of protein synthesis for larval families 9 and 21. Notably, the intracellular specific activity differed for the transported [¹⁴C]glycine in the free amino acid pools of larvae for these two families (Fig. 3B). Replicate assays within a family did not differ statistically (family 9, ANCOVA slope: F_1,6=1.49, P=0.3; family 21, ANCOVA slope: F_1,6=1.63, P=0.3). Within each larval family, a single regression line is shown calculated from the pooled data from the two time course assays. There was no difference in slope between the two regression lines (ANCOVA F_1,16=1.25, P=0.3), but there was a significant difference in the intercepts of the regressions between families 9 and 21 (ANCOVA F_1,17=18.51, P=0.0005). The difference in intercept, and not in slope, indicated a difference in the amount (moles) of glycine in the free amino acid pool of larvae from families 9 and 21. For example at 28°C, family 9 had an average of 4.4±0.2 pmol glycine larva⁻¹, which was significantly lower than that of family 21, which averaged 6.3±0.2 pmol glycine larva⁻¹ (ANOVA F_1,18=31.1, P<0.0001). Hence, the higher specific activity of glycine in the free amino acid pool measured for family 9 (Fig. 3B) is attributable to a smaller amount of total glycine in the pool. As the same amount of [¹⁴C]glycine was transported by larvae from each family (Fig. 3A), that radioactivity was differentially diluted by the amount of glycine in the free amino acid pool of larval families 9 and 21. This is reflected in the higher intracellular specific activity of [¹⁴C]glycine in family 9 (the ratio of [¹⁴C]glycine to total glycine).

Additionally, larval families 9 and 21 showed differences in the rate of incorporation of [¹⁴C]glycine into protein (Fig. 3C). Replicate assays within a family did not differ statistically (family 9, ANCOVA slope: F_1,6=0.15, P=0.7; family 21 ANCOVA slope F_1,6=2.57, P=0.2); those data were pooled into a single regression for each larval family. Larvae from family 21 had a significantly faster rate of incorporation of [¹⁴C]glycine into protein than larvae from family 9 (ANCOVA F_1,16=9.50, P=0.007). Similar sets of assays were conducted over the temperature range shown in Fig. 3D. The log value of the average protein synthesis rate at each temperature was used to calculate the Q₁₀ value. An ANOVA regression gives the Q₁₀ value as 10^slope and was 1.7±0.04 (Q₁₀ ±s.e.) for the rate of protein synthesis for larvae from family 9, and 3.1±0.7 for the rate of protein synthesis for larvae from family 21. Larvae from family 21 had a significantly higher temperature sensitivity (Q₁₀ value) for protein synthesis rate than larvae from family 9 (ANCOVA slope F_1,12=42.95, P<0.0001).

The analysis of differences in Q₁₀ values for respiration (Table 2) highlighted overall significant differences in temperature response between larval family 9 (Q₁₀ of 2.3) and 21 (Q₁₀ of 1.6). These two larval families with different responses of respiration to temperature also had different temperature responses for rates of protein synthesis. This comparison for protein synthesis of the two families was based on larvae of similar size, at 147 µm (mean±s.e.m. family 9: 148±1.3 µm; family 21: 146±1.4 µm). By ANOVA, larval sizes between these two families were not different: F_1,107=1.02, P=0.3. Also, these same-sized larvae from family 9 and 21 had a similar protein content (mean±s.e.m. family 9: 64.9±4.2 ng protein larva⁻¹; family 21: 67.0±2.1 ng protein larva⁻¹). Larval family 9 at this size fraction had a greater respiration Q₁₀ of 2.4±0.22, compared with a Q₁₀ of 1.7±0.20 for family 21 (by ANCOVA, these Q₁₀ values were different: F_1,50=5.56, P=0.02). Strikingly, the reverse pattern was measured for the temperature sensitivity of protein synthesis (Fig. 3D), where larval family 21 had a greater Q₁₀ value of 3.1±0.7 and family 9 the lower Q₁₀ of 1.7±0.04 (by ANCOVA these Q₁₀ values were different: F_1,12=42.79, P<0.0001).

In summary, a physiological difference was identified between larval families 9 and 21 with respect to the alternation of thermal sensitivity for the supply of energy (respiration) and demand for energy (protein synthesis). This difference in protein synthesis was not driven by acquisition of substrate, because the rate of amino acid transport from seawater was equivalent between larval families at any given temperature tested (e.g. at 28°C, Fig. 3A: ANCOVA, F_1,16=0.24, P=0.6). Larval families 9 and 21 had a similar Q₁₀ value for glycine transport of 1.8±0.2.

The different Q₁₀ values for respiration and protein synthesis for families 9 and 21 have substantial consequences for the supply and demand of energy allocation and availability (illustrated in Fig. 4). In this analysis, the rates of respiration (Fig. 1E,F) and protein synthesis (Fig. 3D) were converted to energy equivalents to compare energy supply (respiration) with energy demand (protein synthesis) (see Materials and Methods, ‘Allocation of energy to support the cost of protein synthesis with rising temperate’). The total area of each pie chart represents the energy from respiration (Fig. 4A–D: 329, 594, 286 and 476 µJ, respectively). The percentage allocation to protein synthesis is represented by the area in black within each pie chart. For example, Fig. 4A shows the energy allocation at a temperature of 18°C is 42% to protein synthesis for a 147 µm larva of family 9. This allocation is calculated as follows (from data in Figs 1E and 3D): (a) respiration rate of 679 pmol O₂ larva⁻¹ day⁻¹ (based on 28.3 pmol O₂ larva⁻¹ h⁻¹ at 18°C for family 9) converts to 329 µJ day⁻¹ (679 pmol O₂×484 kJ mol⁻¹ O₂÷1000); (b) protein synthesis rate of 65 ng protein larva⁻¹ day⁻¹ (based on 2.7 ng protein larva⁻¹ h⁻¹ at 18°C) converts to 137 µJ day⁻¹ (65 ng protein×2.1 J mg⁻¹ protein synthesized). Thus, the ratio of the energy required to support protein synthesis (137 µJ day⁻¹) to the total energy supply (329 µJ day⁻¹) is 42% (as presented in Fig. 4A). For the same-sized larvae from family 9 exposed to a higher temperature of 28°C (Fig. 4B), the high Q₁₀ value of 2.4 for respiration would supply sufficient energy to meet the increased rate of protein synthesis at this higher temperature, as protein synthesis has a lower Q₁₀ of 1.7 relative to respiration at 2.4.

The physiological state labeled ‘optimal’ (Fig. 4B) is the scenario where larvae from family 9 have sufficient energy to meet (and exceed) the increased metabolic cost from an increased rate of protein synthesis, which requires 28% of total energy at the higher temperature. Herein, ‘optimal’ is considered to be a scenario where 50% or less of ATP is allocated to a single process – protein synthesis in this case – allowing for sufficient energy in the ATP pool to support other essential processes (e.g. ion regulation: Pan et al., 2018, 2021). In contrast, same-sized (147 µm) larvae from family 21 are considered to be in a ‘stressful’ physiological state at the same temperature (28°C; Fig. 4D). In this case, the cost of supporting the increased rate of protein synthesis requires an increased energy allocation from 34% at 18°C (Fig. 4C) to 67% at 28°C (Fig. 4D). It is notable in these scenarios that larvae from these two families allocate between 42% (family 9) and 34% (family 21) of total energy to protein synthesis. This sustainable physiological scenario at 18°C would permit allocation of cellular energy to other essential physiological processes in C. gigas larvae (e.g. the cost of ion transport and calcification; see Pan et al., 2018). It is only under conditions of rising temperature that the unsustainable physiological allocation to protein synthesis (67% at 28°C) becomes evident for family 21. Pan et al. (2021) provides a comparison of larval survival of C. gigas at 21°C and 28°C.

Following on from previous work (Pace et al., 2006; Hedgecock and Davis, 2007; Sun et al., 2015; Pan et al., 2021), this study aimed to develop phenotypic contrasts among different pedigreed larval families of the Pacific oyster (Crassostrea gigas). This offered a unique advantage for insights into biological variance and mechanisms of physiological resilience of marine organisms to environmental change. Of particular interest in the present study is the close relatedness of larval families 2 and 3 and their corresponding similarity in high Q₁₀ values of respiration averaging ∼3.0 (Table 2, Fig. 2). These two larval families have a genetic relatedness of 0.375, because they share the same male parent (male C, Table 1), and sibling female parents (females D.1 and D.2, indicating different individuals of the same family line, Table 1) (see Materials and Methods). Thus, reproducibility of phenotypic measurements is found among replicate observations of the same family and among different larval families sharing parents. Overall, these data suggest that there is standing genetic variation for temperature sensitivity in current populations. The availability of these pedigreed lines will allow genetic mapping of physiological phenotypes in subsequent generations (e.g. the ability to map loci for viability in F₂ families of C. gigas: Yin and Hedgecock, 2021).

In our previous study (Pan et al., 2021), we reported a Q₁₀ of ∼3 for the thermal sensitivity of protein synthesis rates in C. gigas larvae (from table 3 of that study, average Q₁₀ of 2.9±0.18). This finding was based on an analysis of protein synthesis across 13 different larval sizes, representing seven different larval cohorts. In the present study, two mechanisms of resilience were identified. First, specific larval families were found to have increased thermal sensitivity of respiration (Fig. 2). Notably, these differences remained constant during growth (size independent) (Fig. 2). Second, an alternative strategy was identified, of decreased thermal sensitivity (Q₁₀ value of 1.7) for protein synthesis (Fig. 3D). The observed crossover at a temperature of ∼20°C for larval families 9 and 21 (Fig. 3D) suggests a genotype-by-environment response for the change in protein synthesis rate with rising temperature. The outcome from these analyses is a unique approach to define a physiological basis for biological diversity within a species that permits optimal function for energy allocation with changing temperature (larval family 9, Fig. 4A,B). Larval family 9 exhibits an optimal phenotypic contrast to other larval families (e.g. larval family 21, Fig. 4C,D) regarding the physiological response to increased temperature.

While an increase in respiration rate with rising temperature will provide additional ATP to sustain essential physiological processes, the trade-off of having that capacity is the requirement for increased food intake to sustain elevated rates of respiration (see below). Considerations of genetic-based resilience, physiological trade-offs required to sustain the maintenance of metabolism and growth, and changes in food availability in the environment will help to improve models that predict survival under various scenarios of environmental change.

The current study identified two sets of related larval families of C. gigas (larval families 1 and 4, and 2 and 3) that have the physiological capacity to increase their respiration rate with a high thermal sensitivity (Fig. 2: Q₁₀ value of ∼3) to match the demand for ATP from thermally induced increases in protein synthesis (which is known to have a Q₁₀ of 3; Pan et al., 2021). Respiration Q₁₀ values of 3 or higher are not uncommon for a range of freshwater, marine and terrestrial ectotherms (Seebacher et al., 2015). In the present study, of the 21 families tested, four had the physiological capacity to increase the Q₁₀ of respiration to 3. Previously, Pan et al. (2021) reported that C. gigas larvae had high Q₁₀ values for protein synthesis. Here, we report that specific larval families of C. gigas have the physiological capacity to match the increased ATP demand from protein synthesis with rising temperature with a corresponding increase in respiration. This physiology would confer resilience to rising temperature.

Findings for resilience in larval families of C. gigas have previously been shown for other environmental stressors, (e.g. ocean acidification: Frieder et al., 2017, 2018). In these cases (see also Pan et al., 2015, 2018), a common response mechanism is the regulation of protein turnover and the increased metabolic cost of increasing synthesis to balance protein metabolic dynamics (synthesis, turnover and accretion). Under optimal conditions, the fraction of energy allocated to protein synthesis reported in the present study (Fig. 4: 34–42%) is consistent with previous studies (Lee et al., 2016; Frieder et al., 2017; Pan et al., 2018, 2021). Fig. 4 also shows the energy not accounted for by protein synthesis in this analysis (‘unaccounted’). For larval stages of C. gigas, this ‘unaccounted’ fraction has been well characterized biochemically, with the sodium pump (Na⁺,K⁺-ATPase activity), calcification and nucleic acid biosynthesis comprising the majority of the remaining energy not accounted for by protein synthesis (Pan et al., 2018). If one of these allocation values is altered in response to environmental change – such as protein synthesis increasing to 67% allocation under rising temperature (Fig. 4D: family 21) – support for other essential physiological processes will likely be compromised through energy limitation.

The physiological strategy highlighted in the present study of increasing respiration rate to meet the demands of protein biosynthesis (Fig. 4) requires additional energy to fuel metabolism. Specifically, the increase in respiration at 28°C for larval family 9, which has a Q₁₀ of 2.4, is 594 µJ day⁻¹. This increase in available energy is sufficient to meet the metabolic demand of protein synthesis at this higher temperature (Fig. 4B, ‘optimal’). Conversely larval family 21, which has a respiration Q₁₀ of 1.7, increased available energy to 476 µJ day⁻¹, an available energy that may be insufficient to support the physiological demand of other essential processes (Pan et al., 2018; 2021) beyond protein synthesis (Fig. 4D, ‘stressful’). The difference in available energy between the optimal and stressful physiological state is 118 µJ. What amount of additional algal foods would be required to meet that energy difference?

Three of the common algal foods used to grow larvae of C. gigas have a similar organic weight (from table 1 of Helm and Bourne, 2004: Isochrysis galbana, T-ISO, Pavlova lutheri). Droop et al. (1982) report a value of 0.54 µJ cell⁻¹ of P. lutheri. Using this estimate, the number of algal cells required to fuel a differential energy of 118 µJ is 219 algal cells per day (9 per hour, assuming 100% efficiency of energy conversion). Larvae of C. gigas grown on a similar algal ration to that used in the present study (30,000 cells ml⁻¹) had a feeding rate of 30 cells h⁻¹ (Pace et al., 2006: ∼150 µm larva). This would need to be increased to 39 cells h⁻¹ to provide the additional energy required to sustain the optimal energy amount of larval family 9 (assuming similar feeding and assimilation rates). This increase in feeding rate is physiologically feasible (see Pace et al., 2006, for studies of genetically determined variation in feeding and assimilation rates in C. gigas larvae). Information remains scant regarding the actual amounts of food available to support larval growth in the ocean (Crisp et al., 1985; Fenaux et al., 1994; see p. 639 of Bayne, 2017, for a review of more recent literature). In addition, major changes are anticipated in global phytoplankton species, production and biochemical composition in a warming ocean (Behrenfeld et al., 2016; Holm et al., 2022). The energy allocation model presented in Fig. 4 illustrates the physiological capacity to respond to ocean change. Integrating such physiological capacities into long-term monitoring and models of environmental change will assist in predicting ‘winners’ and ‘losers’ within populations of a single species.

Denaturation of proteins is time dependent: the longer an enzyme subsists in higher temperatures, the more likely it is that denaturation will occur (Schulte, 2015). Additionally, small changes in temperature have large impacts on protein conformation (Dong et al., 2022). At higher temperature, proteins are more likely to be changing between conformations, and are less likely to be in a proper conformation to perform designated functions. Thus, protein synthesis rates would need to be increased to maintain the same number of proteins capable of performing their function at a given moment in time. In C. gigas, all 88 Hsp70 heat shock genes in the genome show increased expression in response to thermal stress (Zhang et al., 2012). In the fruit fly Drosophila melanogaster, a higher number of the heat shock protein gene copies incur additional metabolic cost (Hoekstra and Montooth, 2013).

RNA structures can impact the regulation of protein synthesis rate (Mortimer et al., 2014). In response to heat shock, RNA secondary structure is involved in regulating gene expression (Su et al., 2018). Across a range of species of marine mollusks, Liao et al. (2021) have shown that stability of mRNA secondary structure (based on nucleotide sequence) differs adaptively among species across a wide temperature range. In our study of larval families of C. gigas, we identified a family that had a lower Q₁₀ value for protein synthesis (larval family 9, Fig. 4). We suggest that such larval families may have mRNA structures that result in a lower response of protein synthesis to higher temperature. Further studies are required to test the possibility that these structural changes would result in a rapid increase in translation with rising temperature, and that there is genetic variation associated with this process. In the current study, it has been established that high Q₁₀ values for protein synthesis require a substantial amount of energy to support increased rates with rising temperature (Fig. 4D). Conversely, a lower temperature sensitivity for protein synthesis would result in lower energy costs, and a more optimal allocation of ATP (Fig. 4B). This latter (‘optimal’) physiology would reduce the requirement to increase synthesis rates with rising temperature.

An analysis of fractional rates of protein synthesis supports this suggestion. Specifically, larval families 9 and 21 are of similar size and protein content (147 µm and 66 ng protein per larva). The rate of protein synthesis at 28°C was 3.3 and 6.3 ng protein larva⁻¹ h⁻¹ for larval families 9 and 21, respectively. This equates to a 2-fold difference in fractional rates of protein synthesis [family 9: 5% (3.3÷66); family 21: 10% (6.3÷66)]. These two larval families had similar growth rates (same size and protein content for a given age), hence the 2-fold higher fractional rate of protein synthesis for larval family 21 would represent a higher turnover rate of protein (i.e. protein degradation rate). The cost of this apparent protein instability is reflected in the increased allocation required to support protein synthesis (Fig. 4: family 9, 28%; family 21, 67%). As discussed above, these differences in temperature-dependent physiological and biochemical rates among larval families of C. gigas may reflect standing genetic variation, where the phenotype is a change in biochemical strategies of protein synthetic dynamics (synthesis, degradation, accretion). Increased protein turnover, not resulting in increased protein accretion, appears to be a common mechanism for the stress response in developmental stages of marine invertebrates. For instance, in larval stages of C. gigas and the sea urchin Stronglyocentrotus purpuratus, increased rates of protein turnover occurred under ocean acidification treatments (Pan et al., 2015; Frieder et al., 2018). Given the large number of 88 Hsp70 heat shock genes in the genome of C. gigas, it would certainly be of interest in future studies to link the physiology of the thermal sensitivity of protein synthesis to the biochemistry of changes in specific proteins of larval stages.

The ability of a species to cope with environmental change sets the evolutionary potential of ‘winners and losers’ (Somero, 2010). The genetic and physiological basis of adaptation to environmental change is highly complex, making predictions across generations particularly challenging. As highlighted by Somero (2010), few studies directly address this question by rearing animals across generations and measuring their physiology. The phenotypic similarity of larval families 1–4 in terms of high respiration Q₁₀ values (Table 2) and the close relatedness among these families (Table 1) suggest an additive genetic component to Q₁₀ variance. This further implies that animals with the trait of high respiration Q₁₀ values would likely contribute to the greater physiological resilience of an evolving natural population. In contrast to this classic scenario of directional selection on additive genetic variation, however, our finding of lower Q₁₀ values in G₁ compared with G₀ larval families suggests that there is also a substantial non-additive genetic component to variation in respiration Q₁₀. Non-additive genetic variation for a complex physiological trait – growth in larvae of C. gigas – was previously evidenced by hybrid vigor in crosses among inbred parent lines (Pace et al., 2006). Transgenerational evidence for non-additive and, therefore, non-heritable variation in respiration Q₁₀ is seen by comparing the findings of Pan et al. (2021) with those of the present study. Pan et al. (2021) report that a cohort of larvae (cohort 5, a full-sibling G₀ family) had a Q₁₀ value of 3.0. That cohort of larvae was reared to adulthood and the sibling broodstock spawned for the present study, yielding family 13. Notably, the high Q₁₀ value of 3.0 for respiration in the previous generation (Pan et al., 2021: cohort 5) was not evident in the next generation used in the present study (Table 2: family 13, Q₁₀=2.2 and 2.1 for larvae of different sizes; Fig. 2). These examples of non-additive genetic effects have important implications for transgenerational inheritance of complex traits and warrant caution regarding predicting inheritance of physiological resilience solely based on parental phenotypes (presumed ‘winners’). A focus on transgenerational (egg-to-egg) inheritance of physiological traits that provide thermal tolerance would significantly advance our understanding of and ability to predict resilience under various scenarios of future climate change.

For studies of biological responses to global ocean change, the study of bivalves (such as C. gigas as a model organism) provides the availability for reproducible analyses of existing genetic lines (cf. wild-type), and has well-developed genomic and physiological resources. Additionally, C. gigas has a global ecological distribution across six continents and tolerates a wide temperature range (2–35°C: Helm, 2009; Miossec et al., 2009). Of note is that the aquaculture of bivalves offers great potential as a source of future ‘blue food’ (Nature editorial, 2021; Naylor et al., 2021), given their high nutritional quality as a human dietary source. The future success of producing blue food will depend upon understanding the basic biology and resilience potential of farmed species. Although the present study was not specifically designed to allow quantitative dissection and analysis of genetic-based variation, the finding that several extant larval families have biochemical strategies of differential energy allocation to respond to rising temperature is notable, regarding the identification of resilient phenotypes. For future analyses, the experimental approaches presented here offer the potential to address questions of physiological resilience and genetic-based adaptation to global ocean change.

We are especially grateful to our colleague Dr Dennis Hedgecock for his intellectual input on this project and for insightful comments on earlier drafts of the manuscript. Special thanks to our industry partner, Thomas Grimm, and his dedicated staff at Carlsbad Aquafarm, without whose facilities, support and collaboration the long-term maintenance of pedigreed lines used in this study would not have been possible. Our thanks to the staff of the Wrigley Marine Science Center at the University of Southern California for their assistance in supporting the larval culturing work reported here. D.T.M. would like to dedicate this publication to his recently deceased mother, Mrs. Pamela Manahan.