Ocean warming shapes embryonic developmental prospects of the next g ener ation in Atlantic cod

Although early embryonic protein production relies exclusively on maternal molecules such as messenger RNAs (mRNAs) incorporated into ovarian follicles, knowledge about any thermally induced, intergenerational effects is scarce in ectotherms. Here, we in vestig ated how elevated temperatures (9 ◦ –12 ◦ C vs. 6 ◦ C) during oogenesis influenced the next generation by targeting maternal mRNAs in Atlantic cod ( Gadus morhua ) embryos, in view of up-and down-regulated genes in ovarian follicles of pre-spawning adults. Overall, the spawning female liver showed significantly higher levels of free amino acids and N-metabolites at 9 ◦ C than at 6 ◦ C. Higher-than-optimal temperatures induced adjustments in embryo transcriptome proportional to the temperature increase relative to the control group. The adjustments included alterations in maternal-effect genes, which are developmentally conserved among vertebrates. The transcriptomic differences for a selection of genes in embryos were reflected in ovarian follicles (containing multiple cell types) several months ahead of the spawning season, implying that environmental conditions of the adults are key for adjusting the genetic instructions for offspring development. This programming of fundamental traits from mother to offspring appears part of a sophisticated process to adapt the offspring to a changing ocean, though within life stage-specific, physiological thermal tolerance windows.


Introduction
Under on-going climate change, the world's ocean temperature and the occurrence of marine heatwaves are increasing at unprecedented rates (Boeke andTaylor 2018 , IPCC 2021 ).Variations in ambient temperature directly affect ectothermic physiology (Little et al. 2020 ), such as reproduction (Alix et al. 2020 ), sex determination (Geffroy and Wedekind 2020 ), and stress (Alfonso et al. 2021 ).Ultimately, these modifications at the individual level may scale up to population dynamics responses and altered surplus production (Kjesbu et al. 2014 ).Certain life stages, foremost adults during reproduction (spawners) and embryos, appear particularly susceptible to temperature change (Dahlke et al. 2020 ).Despite that, many studies on teleosts have clarified that warmer-thanoptimal temperatures during oogenesis may disturb the natural ovulation cycle and affect egg quality and offspring development (Bobe andLabbé 2010 , Alix et al. 2020 ), the fundamental mechanisms in operation behind these observational patterns are little understood.
This knowledge gap is to some extent reduced in recent studies describing parental and intergenerational effects at warmer temperatures (Fellous et al. 2021, Bernal et al. 2022 ).More specifically, these studies suggest a potential preconditioning of teleost offspring to ocean warming, resulting in notable changes in gene expression levels.Tropical damselfish ( Acanthochromis pol y acanthus ) exposed to warmer temperatures for two generations produced a third-generation off-spring with improved aerobic capacity, which coincided with differentially expressed genes activating immunity and cellular stress responses (Bernal et al. 2022 ).In three-spined stickleback ( Gasterosteus aculeatus ), where parents experienced elevated temperatures during gametogenesis, alterations in reprogramming and expression of genes involved in epigenetic modifications throughout embryogenesis were evidently in place (Fellous et al. 2021 ).Furthermore, elevated maternal temperature in rainbow trout ( Oncorhynchus mykiss ) impaired offspring fear-related behaviour, in line with differential expression of maternally deposited transcripts involved in neurodevelopment (Colson et al. 2019 ).
A key point is that the embryonic development depends on maternally inherited molecules, since cleavage-stage cells are transcriptionally inactive during the initial cell cycles (Schier 2007 ).While the exact functions of most maternally provided transcripts are unknown, some are highly conserved among vertebrates and have demonstrated distinct essential roles, defined as maternal-effect genes, in early embryonic processes like cell division, chromatin regulation, and embryonic patterning (Abrams andMullins 2009 , Dosch 2015 ).Therefore, changes in maternal transcripts could have phenotypic effects later in embryogenesis.
Other maternally inherited molecules, such as free amino acids (FAAs) originating from yolk proteins, synthesized in the liver and incorporated in developing oocytes, are also abundant in pelagic marine fish eggs.FAAs are important for oocyte swelling prior to ovulation (Lubzens et al. 2017 ), energy fuel, and protein synthesis in developing embryos (Rønnestad and Fyhn 1993 ).Since the FAA content in fish eggs directly depends on vitellogenin synthesis in the liver of the mother, a suboptimal environment at warmer-than-normal temperatures during vitellogenesis may alter the nutritional status of the liver, and subsequently the nutrient transfer to eggs and future embryos.As an example, in Atlantic salmon ( Salmo salar ), changes in environmental conditions, including temperature and light, may lead to altered nutritional status in maternal muscle and liver tissues, which is reflected in the nutritional status in the embryos, suggesting an intergenerational effect of changing environment (Skjaerven et al. 2022 ).
In a climate change scenario, high-latitude oceans are projected to warm faster than elsewhere in the Earth's marine basins (IPCC 2021 ).Although Atlantic cod inhabits North-Atlantic waters and coastal zones with temperatures ranging from −1.5 • to 19 • C, the local thermal range is typically narrower due to the existence of metapopulations (ecotypes) (Righton et al. 2010 ).Temperatures above ∼9.6 • C during spawning cause ovulation failure (Kjesbu et al. 2023 ), leading to significant reductions in fertilization rates and abnormal embryonic cell cleavage (van der Meeren and Ivannikov 2006 ).In the latter respect, mild thermal stress may alter the level of maternally loaded mRNAs regulating proliferation and DNA methylation already at the pre-gastrula stages (Skjaerven et al. 2011(Skjaerven et al. , 2014 ) ).Thus, temperature is considered the main abiotic factor accelerating enzymatic processes and cell proliferation in ectotherms.Environmental exposure also potentially influences chromatin regulation (epigenetic pathways) during these sensitive early stages of embryonic patterning (Abrams andMullins 2009 , Dosch 2015 ).
In this study, we first investigated if elevated temperatures during Atlantic cod gametogenesis affected the developmental potential of the resulting embryos, including testing if the experimentally set extreme 12 • C would be critical and lead to 100% non-viable eggs.Furthermore, we aimed at describing any associated changes in the transcriptome of embryos from maternal origin to identify potentially affected molecular mechanisms.We studied if the observed changes in mRNA levels in the embryos are reflected in the mother prior to spawning during the development of oocyte (ovarian) follicles.Lastly, we explored if other essential maternally inherited molecules were affected by elevated temperature in the form of altered amino acid composition in the maternal liver.To achieve this, wild-caught cod broodstock were subjected to different temperatures throughout gametogenesis and spawning in state-of-the-art settings ( Fig. 1 ), mimicking different ocean warming scenarios.Embryo survival, hatching rate, and mRNA composition were assessed in the offspring, while follicular expression of selected candidate genes and FAA composition were assessed in the female broodstock.

Materials and methods
Field sampling and rationale behind the adopted thermal experimental design Wild Norwegian coastal cod ( N = 150: 58% of female, 28% of male, and 14% unknown sex; Supplementary Table S1 ) were captured in specially designed fish pots by local fishermen on the southwest Norwegian coast (59 • N, 05 • E, Bømlo, Norway; Fig. 1 ) in the spring of 2018.After transfer to IMR Matre Research Station (60 • N, 05 • E, Matredal, Norway), the individuals-on average 3.5 ± 1.8 kg (whole body weight) and 66.2 ± 9.8 cm (total length)-were equally and randomly distributed in 9 tanks (each 8 m 3 and supplied with seawater originating from 90 m depth) and maintained at 8 • C until the experiment began.In September 2018, each fish was PIT tagged (Glass Tag 2.12 × 12 mm, Smartrac) to track whole body weight and total length as well as reproductive descriptors (see below) throughout the experiment.
As the onset of secondary growth of ovarian follicles in Atlantic cod generally commences around the autumn equinox (Kjesbu et al. 2010 ), experimental temperatures were changed gradually from 1 October 2018 by a maximum of 1 • C per day (acclimatization) to establish the three experimental regimes: 6 • C as control, 9 • C as moderate, and 12 • C as 'stress temperature'.The rationale for choosing these experimental temperatures is present in Supplementary Materials and Methods and Supplementary Fig. 1 .
A triplicate design, 3 temperatures × 3 tanks, was adopted, which resulted in 12-18 individuals per tank ( Supplementary Table S1 ) and thereby 47-51 individuals for each of the three experimental regimes.The sex ratio (male: female) varied from 1:1 to 1:3 depending on the tank ( Supplementary Table S1 ).Sex was difficult to identify in some individuals and these were classified as 'Unknown' ( Supplementary Table S1 ).
Adults were held at these temperatures under as far as possible natural environmental conditions: a simulated natural photoperiod at 60 • N (following natural sunrise and sunset), salinity (35 ppt), oxygen (over 80%), and pH (7.8) (Kjesbu et al. 2022 ) until the end of the spawning season (end of April 2019).From September until spawning started (on 3 January at 9 • C and 8 January at 6 • C and 12 • C), the fish were fed to satiation 3 times a week with 9 mm commercial pellets (Amber Neptun, Skretting ARC).Feeding was arrested during the spawning season.
In December 2018 and January 2019, ahead of the spawning season (Brander 2005 ) ( Fig. 1 ), specimens were anaesthetized (Finquel, 0.6 g L −1 ) and checked for sex and reproductive status (in January, N = 127: 64% females, 35% males, and 1% unknown; Supplementary Table S1 ).Gonad catheterizations ('biopsies', ∼1 cm 3 ) were performed using Pipelle ® (Laboratoire C.C.D., France).Half of the collected ovarian follicles were immediately fixed in 3.6% buffered formaldehyde for histological and image analyses (undertaken at least 2 weeks later for complete fixation), and the rest of the material was frozen in liquid nitrogen and stored at −80 • C until gene expression analysis by qPCR.
At the termination of the experiment (end of April 2019), past the spawning season (Brander 2005 ) (22-26 April 2019 depending on temperatures), all individuals were euthanized by an overdose of MS222 (500 mg L −1 ).Sets of subsamples of the ovary and liver underwent either fixation or were stored in liquid nitrogen (as described above) for later analyses.

Automated image and histological analyses of ovarian follicle samples
Each fixed ovarian sample ( Supplementary Table S2 ) was ultrasonicated for 10 s (Vibra-Cell™, VCX130, Sonics & Materials Inc., USA) (Anderson et al. 2020 ), stained with 2% toluidine blue and 1% sodium tetraborate and photographed under a stereomicroscope (Olympus SZX10, Olympus Corp., Japan) at × 12.5 magnification by a SPOT Insight CMOS 5MP colour camera operated by Spot™ 5.6 software (Diagnostic Instrument Inc. 5.3.3).Ovarian follicle stage (previtellogenic, early vitellogenic, late vitellogenic, and hydrated ovarian follicles) and the presence/absence of ovulated oocytes (eggs) were assessed from the micrographs (Kjesbu et al. 2011 ).Further insights in ovarian follicle developmental aspects were given from histology ( Fig. 1 ).The automated image and histological protocols are further detailed in Supplementary Materials and Methods .

Spawning season and collection of eggs
Egg collectors were installed on each tank (one collector for alive/floating eggs and another one for dead/sinking eggs), allowing precise, daily collection and monitoring of natural spawning.The collected and naturally fertilized eggs (embryos) were observed under a stereomicroscope Olympus SZX10 (magnification × 8 or × 6.3) to identify developmental stage (Fri ð geirsson 1978 ) and record egg diameter ( μm).As the studied egg batches from each tank at a given day could originate from several individuals ( Supplementary Table S1 ), special care was made during sorting (by size and external appearance) and staging of the developmental stage (2-to > 128cell stage) (Fri ð geirsson 1978 ) to separate any simultaneously occurring egg batches (Kjesbu 1989 ). Totally, 50 eggs from each batch were frozen at −80 • C for later molecular analyses.The risk that more than one female may have spawned eggs of exactly the same size and appearance at a given hour was overlooked.

Embryonic developmental success
To track embryonic development, 18 egg batches per temperature were randomly collected and incubated in a NUNC™ 24-well plate (Thermo Fisher Scientific Inc.) until hatching, with 1 embryo per well and 4 plates per spawn.Incubation temperature was set to reflect the above-mentioned tem-perature regimes experienced by the parents: 6.1 ± 0.3 • C, 9.0 ± 0.2 • C, and 12.0 ± 0.2 • C, respectively.Thus, we assumed that eggs in oceanic waters would experience the same temperatures as the local temperature at or near the spawning ground for the first 1-2 days of incubation (Sundby 1991 ).The total number of incubated embryos (96 embryos) formed the denominator to calculate survival and hatching rates.For consistency, the reported survival rates always referred to the same developmental stage, here blastula embryos (Fri ð geirsson 1978 ), as the developmental time would be expected to vary with temperature (Geffen et al. 2006 ).

RNA extraction from eggs
To investigate how maternal temperature influences the offspring ( Fig. 1 ), we targeted maternal mRNAs in early cleavage stage (2-16 cells) embryos, which is prior to the maternalzygotic transition (Schier 2007 ).Among the egg batches incubated, 6 samples of ∼50 eggs from each of the 3 spawning tanks per temperature regime were selected for RNA extraction ( Supplementary Table S3 ), with methods described in Supplementary Materials and Methods .

RNA sequencing of eggs, and bioinformatic analysis of RNA-seq reads
Eighteen libraries for RNA-sequencing (RNA-seq) were constructed using 1 μg embryonic RNA per sample in the NEB-Next Ultra II Directional RNA library Prep Kit (New England Biolabs, USA) with the NEBNext ® rRNA Depletion module (New England Biolabs, USA), according to the manufacturer's protocol.The final libraries were quantified, normalized, pooled at equimolar ratio, and sequenced on the Illumina NextSeq 500 sequencer (Illumina, USA) with 75 base pairs of paired-end reads.Methods for the assessment of sequence quality, tests for outliers, and differential gene expression appear in Supplementary Materials and Methods .

RNA extraction of ovarian follicle samples, cDNA synthesis, and RT-qPCR
To verify mRNA expression levels in pre-spawning females, ovarian follicles collected in December 2018 and January 2019 were analysed by RT-qPCR ( Supplementary Table S4 ).The ovarian follicles were sampled from two females to six females from each of the three tanks per experimental temperature-and the RNA extracted as described for eggs, except that the starting material was 50 mg of tissue.Selected females were of the same maturity stage, according to the results obtained in histology.The RNA quantity and quality were checked with the same protocol as for the embryos.The mean quantity was 1055.5 ng μL −1 , the mean RIN value was 9.5 (9.2-9.9), and absorbance 260/280 ratios of 1.9-2.1 (average 2.08).The quality of the RNA was thereby considered suitable for further analysis.A two-step protocol was used to quantify the RNA by RT-qPCR ( Supplementary Materials and Methods ).
The primer sequences used for detection of the reference genes β-actin ( actb ) and elongation factor 1 alpha ( ef1a ) were from Skjaerven et al. (2011) , while primers for the target genes ( , pancreatic progenitor cell differentiation and proliferation factor A-like [ ppdpfal ], monocarboxylate transporter 7-like [ slc16a6l ], and solute carrier f amil y 41 member 1 [ slc41a1 ]) were designed using Geneious 2020.2.4 ( https://www.geneious.com).The primer for dync1i1 (LOC115535959), also matched the gene pdk4 (LOC115535965) as both genes have overlapping regions on chromosome 22.Primer sequences and efficiencies are shown in Supplementary Table S4 .All primers were approved by a one-step PCR test, including a 50-bp GelPilot R ladder (Qiagen, Germany), according to the manufacturer protocols.

Amino acid composition in female broodstock liver
Free amino acid composition and N-metabolites (urea, ammonium, etc.) were determined from samples of broodstock livers from the 6 • to 9 • C regime on a Biochrom 20 plus amino acid bioanalyser (Amersham Pharmacia Biotech, Sweden) using post-column derivatization with ninhydrin (Espe et al. 2014 ).We analysed 24 and 18 liver samples from each of the 3 replicate tanks at 6 • -9 • C, respectively, collected in May 2019 at the end of the experiment from the rest of the live fish.Samples from the 12 • C regime were not included in this chemical analysis as none of the embryos from these spawners survived.

Statistical analysis
For leading cohort oocyte (LC) diameter, egg diameter, and gene expression values, a nonparametric Kruskal-Wallis test, complemented with the post hoc Dunn's multiple comparisons test, was applied to examine any significant differences among the temperature regimes.We used Spearman's rank correlation coefficient to test for significant correlations between gene expression and LC diameter.Significant differences in survival rate (at blastula stage) and hatching rate between temperature regimes were tested by a Chi-square test with a pairwise comparison for proportions.Differences in free amino acid composition between regimes in female broodstock liver were assessed on the mean values from each tank using unpaired t -test.Statistical analyses were performed by GraphPad Prism and R 3.4.4/RStudio1.1.447,with significance set at P < .05.

Temperature affects egg quality
Egg developmental success, characterized by survival and hatching rates, were influenced by temperature.In egg batches undergoing transcriptomic analysis, the egg diameter was significantly larger at 9 • C compared to control temperature (6 • C) and stress temperature (12 • C) ( Fig. 2 a).This diameter varied less in the two warmer temperature regimes compared to the control ( Fig. 2 a).Warmer temperatures significantly increased mortalities of embryos at the blastula stage ( Fig. 2 b, Supplementary Fig. S2 ).Thus, hatching rates significantly decreased with increasing temperatures ( Fig. 2 c).Increasing incubation temperature from 6 • to 9 • C resulted in an increase in developmental rates: embryos incubated at 6 • C hatched on average at 17.7 ± 2.2 days post fertilization (dpf) (range: 13-22), while those incubated at 9 • C hatched at 10.2 days ± 0.9 dpf (range: 9-12) ( Fig. 2 d).None of the embryos incubated at 12 • C survived until hatching.

Embryonic mRNA expression clustered by temperature
The RNA sequences were clearly separated in the PCA plot between the three temperature-specific mRNAs present in the early cleavage stage embryos ( Fig. 3 a).These RNA expression profiles were uninfluenced by any variation in the associated cell number in the cleavage-stage embryos sampled ( Supplementary Fig. S3 ).No outliers were statistically identified.
An increase from 6 • to 9 • C altered the expression of 165 genes, of which 83 exhibited lower transcript levels and 82 higher transcript levels at 9 • C compared to at 6 • C, respectively ( Fig. 3 b, Supplementary Fig. S4a , and Supplementary Table S5a ).The following comparison between 6 • and 12 • C led to 408 differential expressed genes (DEGs), of which 207 exhibited lower mRNA transcript levels and 201 higher transcript levels at 12 • C compared to 6 • C ( Fig. 3 b, Supplementary Fig. S4b , and Supplementary Table S5b ).A total of 210 mR-NAs, with 125 showing lower transcript levels and 95 higher transcript levels, were identified at 12 • C compared to 9 • C ( Fig. 3 b, Supplementary Fig. S4c , and Supplementary Table S5c ).The DEGs showed clear hierarchical temperature-related clustering of sequenced embryos, indicating that the identified DEGs had an expression profile that was influenced by ambient temperature ( Fig. 3 c and Supplementary Fig. S5 [9 • C vs. 12 • C]).This finding was further strengthened by focusing on the 40 most significant DEGs.The heatmaps revealed a dose-dependent temperature response as the samples clustered according to ambient temperature ( Supplementary Fig. S6 ).
Among the genes presenting significantly higher transcript levels ( Supplementary Tables S5a -c ), the most enriched pathway was ribosome (annotated to 26.9% of the genes with higher levels of transcripts in the 12 • C regime compared to the 6 • C regime), followed by cAMP signalling (14.5% of those genes; Supplementary Fig. S7a ).The latter pathway was also annotated to a large proportion of genes with lower transcript levels both in the 6 • C vs. 12 • C (10.1%; Supplementary Fig. S7b ) and in the 9 • C vs .12 • C comparison (6.7%; Supplementary Fig. S7c ).Other pathways, such as PI3K-Akt signalling, focal adhesion, and axon guidance, were among the most enriched pathways in both categories of genes [higher or lower transcript levels ( Supplementary Fig. S7a -c )].In the 6 • C vs. 12 • C comparison, endocytosis, inositol phosphate metabolism, and phosphatidylinositol signalling system pathways were each annotated to 12.1% of the genes exhibiting lower transcript levels ( Supplementary Fig. S7a ).All genes within each pathway in Supplementary Fig. S7 can be viewed in Supplementary Table S6 .
Climate scenarios of 3 • -6 • C increase in temperature modulate mRNAs in ovarian follicles for embryo development several months ahead of spawning The altered composition of maternally inherited embryonic mRNAs in response to the two elevated temperature regimes during oogenesis suggested that the changes may have arisen through adjusting the incorporation of maternal mRNAs in the ovarian tissue several months previously, before final maturation of the ovarian follicles and ovulation.This was confirmed by investigating follicular mRNA levels of eight selected embryonic DEGs.These DEGs had a consistent dosedependent response of either greater or lesser expression in the elevated temperature regimes ( Figs 3 e-f and 4 m-t).The samples analysed were ovarian follicles originating from females  of the same maturity stage at two times prior to spawning (December-January) ( Fig. 5 ).To be sure that the differences in gene expression were due to temperature and not follicle maturation stage, we checked for correlation between gene expression and LC diameter.In most cases, expressions of the selected genes were statistically uncorrelated to this oocyte size ( Supplementary Fig. S8 ), and therefore to ovarian follicle developmental stages (early vitellogenic, late vitellogenic, and final maturation of ovarian follicles; Fig. 1 ).Four genes ( dync1i1 , pnpla8l , sgpp2 , and znf592 ) presented higher transcript levels in embryos ( Fig. 4 m-p).As early as December, pn-pla8l and sgpp2 exhibited significantly higher transcript levels in ovarian samples of the 12 • C regime compared to the 6 • C regime.The same trend was observed for pnpla8l and sgpp2 in January, but the differences were then statistically insignificant ( Fig. 5 b-c).Expression of dync1i1 and znf592 did not differ significantly between any temperature regime; however, there was a trend towards greater expression with increasing temperature ( Fig. 5 a-d).Among the selected genes exhibiting lower transcript levels in embryos ( arhgap20l , ppdpfal , slc16a6l , and slc41a1 ; Fig. 4 q-t), slc41a1 displayed a significantly lower expression in the 12 • C than in the 6 • C regime in December ( Fig. 5 h).Expression of arhgap20l was lower in the 9 • C than in the 6 • C regime in December ( Fig. 5 e).A tendency towards lesser slc41a1 and arhgap20l expression with increasing temperature was also observed in January but this result was non-significant ( Fig. 5 h-e).Likewise, no significant differences between regimes could be detected for ppdpfal and slc16a6l ( Fig. 5 f-g).

Liver free amino acid content was generally higher in moderately warmer water
The results on the potential role of temperature-induced nutritional programming in broodstock-with reference to liver free amino acid and N-metabolite composition-indicated that the females held at moderately warmer water, 9 • C, contained significantly higher, overall levels compared to those at 6 • C ( Fig. 6 a).No significant differences were found in tests of each free amino acid for the influence of spawning temperature; however, alanine, valine, proline, tyrosine, and aspartic acid were examples of free amino acids, which showed a trend towards larger mean value at 9 • C compared to 6 • C ( Fig. 6 bf).The N-metabolite taurine was significantly greater at 9 • C than at 6 • C ( Fig. 6 g), whereas the β-alanine (beta amino acid, not incorporated into proteins, which function as a metabolic buffer) showed a trend toward a smaller mean value at 9 • C against 6 • C ( Fig. 6 h).In addition, the ammonium content in liver indicated a trend towards a change in N-metabolite turnover in broodstock due to temperature ( Fig. 6 i).

Discussion
The world's dynamic climatic factors, such as temperature, constantly interact and thereby play a key role in forming local biodiversity (Bindoff et al. 2019 ).As ectotherms, most fish are directly impacted by ambient temperatures, though active displacement to other water masses, as commonly seen during ocean warming (Poloczanska et al. 2013 ), may be an effective mechanism to fine-tune the in-situ body temperature to more optimal thermal conditions.Nevertheless, when the temperature rises, so does the metabolic rate and the need for oxygen as all enzymatic processes accelerate (Pörtner and Peck 2010 ).
In this study, we first explored the impact of elevated temperature (from 6 • C to 9-12 • C), preceding the spring spawning season, on the expression of maternal mRNAs in early cleavage stage cod embryos.Prior to the onset of zygotic production of mRNAs, which in fish takes place after the cleavage stage during the maternal-to-zygotic transition (MZT) (Despic et al. 2017 ), maternally deposited mRNAs control development from the first cell division through early embryonic stages (Tadros and Lipshitz 2009 ).Variations in maternal mRNA may therefore have cascading consequences on offspring phenotypes (Mitchell 2022 ).Our study, which specifically focuses on Gadus morhua , a species of high ecological and commercial significance, demonstrates the clear temperature dose-dependent deposition of maternal mRNAs into ovarian follicles and fertilized eggs ( Fig. 7 ).The greater the deviation from the seemingly, corresponding optimum temperature (6 • C) (van der Meeren and Ivannikov 2006 ), the greater was the number of differentially expressed transcripts in embryos.Among the most consistently affected genes, all wellknown to be conserved among vertebrate species, we detected members categorized as maternal-effect genes, which play key roles in defining the offspring phenotype later in development, as shown for mammals (Mitchell 2022 ).

The fate of early life stages
To place our molecular biology and nutritional chemistry findings into a life-history perspective, the warmest temperature of 12 • C resulted in a smaller embryonic survival rate and thus offspring hatching rate, reflecting the existence of a physiological temperature sensitivity curve at the embryonic stages (Laurence and Rogers 1976 ).In general, the fish embryonic period (from fertilization to hatching) becomes exponentially shorter as temperature rises [Atlantic cod: −1 • C: 44 days post fertilization (dpf), 2 • C: 22 dpf, 4 • C: 15 dpf, 6 • C: 12 dpf, 8 • C: 10 dpf, 10 • C: 9 dpf, and 12 • C: 8 dpf].As the natural thermal window for cod embryos is between 0 • and 9 • C (Dahlke et al. 2018 ), an incubation temperature outside this range is expected to lead to decreased rate of somite development (Hall and Johnston 2003 ) and to greater mortality (Dahlke et al. 2018 ).So, incubation temperature conclusively affects the fate of the cod embryonic stages.

Maternal mRNAs are temperature dependent and act over generations
Besides showing that the egg incubation temperature directly impacts developmental success, we propose that maternal temperature history indirectly contributes to egg quality and embryogenesis via maternal mRNAs.We analysed cleavagestage embryos from three temperature regimes and found that the levels of hundreds of maternal mRNAs were altered.Several of those mRNAs encode proteins with functions in embryonic development, such as cytoskeleton assembly ( sdk2 , eml6 , and elmo2 ), cell fate ( hipk2 , yap1 , and yrk ), folding of proteins ( dnajb1b) , transcriptional regulator and mitosis ( mapk4 ), protein methylation ( eef2kmt ), translation ( zar1 ), chromatin configuration ( smarce1 and kdm2b ), all which may change developmental gene regulatory networks, causing a 'domino effect' through subsequent embryonic tissue architecture, growth, and development (Davidson et al. 2022 ).In our study, the sequenced mRNAs from embryo samples clustered into three groups depending on maternal temperature experienced during the   reproductive cycle.This statistical outcome implies that this climate stressor undoubtedly has an effect on embryonic transcripts during early cleavage stages, suggesting a temperaturedependent phenotypic egg plasticity function.However, the fates of the spawned eggs in the field are also determined by other effectors, e.g. the degree of predation (Allan et al. 2021 ).
Maternal-effect genes are believed to be essential in early embryonic development but the impact on the phenotypic expression of the offspring might be manifested relatively late in development (Mitchell 2022 ).We found that both the transcripts encoding the Yes-associated protein 1 ( yap1 ) and Zygote arrest 1 ( zar1 ) were influenced by temperature.Both of these genes are classified as maternal-effect genes.The gene yap1 is critically important for the development of the noto-chord and body elongation (Kimelman et al. 2017 ), whereas zar1 regulates translational repression for eggshell (zona pellucida) proteins in zebrafish ( Danio rerio ) (Miao et al. 2017 ) and is embryonic lethal in knockout mice models (Wu et al. 2003 ).For other consistently affected DEGs, similar genes are maternal-effect genes in mammals, for instance, mitogenactivated protein kinases 1 and 3 (similar to mapk4 ), lysine demethylase 4a (similar to kdm2b ), and lysine methyltransferase 2D (similar to eef2kmt ), all of which have important functions for post-translational modifications of proteins or regulation of chromatin structure (Mitchell 2022 ).Another gene, smarce1 , is influenced by temperature in the cod early stage embryos .This gene encodes an evolutionarily conserved protein that interacts with a large chromatin remodelling complex called the SWI/SNF.This complex regulates the transcription of pluripotency genes, which play a crucial role in balancing proliferation and differentiation during embryo development (Mitchell 2022 ).These observations indicate that temperatures during reproduction are of special importance for patterning and cell fate.Moreover, these changes in mRNA levels may adjust the starting point for protein production and signalling during development when the embryo is especially vulnerable as it does not yet have its own transcriptional machinery to cope with environmental stressors.
The maternal-to-zygotic gene activation occurs gradually prior to the late blastula stage in fish (Tadros andLipshitz 2009 , Lubzens et al. 2017 ).Maternal mRNAs are transferred to the ovarian follicle during the reproductive cycle ( Fig. 7 ) and are progressively activated to support essential oogenesis and pre-MZT embryogenesis functions (Winata and Korzh 2018 ).Our results reveal an intriguing trend where maternal mRNAs with lower expression in embryos exposed to warmer temperatures encode proteins that are important for such functions as microtubule binding ( eml6 ), cell migration ( elmo2 and yrk ), growth ( hipk2 ), and mitosis ( mapk4 ).In contrast, transcripts related to cell adhesion ( sdk2 ), negative effects on translation ( zar1 ), and inflammatory response ( sgpp2 ) exhibited higher levels of expression under increased environmental temperatures.Notably, the zar1 transcript, which is highly expressed at the warmest temperature, has been named the 'zygotic arrest gene 1' (Mitchell 2022 ), suggesting that it may partly contribute to the decreased survival rate of embryos at this temperature.Cross-disciplinary literature indicates that changes in maternal mRNA levels do not act by themselves, but rather in concert with both metabolites and nutrients, as well as epigenetic mechanisms like microRNAs, DNA methylation, and histone tail modifications (Mitchell 2022 ).The smarce1 and kdm2b genes are involved with chromatin regulation and both displayed a reduced transcript level when temperature increased.
The composition of the egg and the interaction with environmental signals collectively lay the foundation for the next generation being adapted to climate fluctuations, and the maternal mRNAs thereby act across generations (also called intergenerational effects transferred from mother to offspring) (Heard andMartienssen 2014 , Bautista andCrespel 2021 ).These early developmental changes can convey a cascading effect during cell cleavage possibly in all the thousand cells of the embryo at the blastula stage, as maternal mRNA influences the translation of mRNA into proteins (Tadros andLipshitz 2009 , Mitchell 2022 ).Our study of the changes in maternal mRNAs due to reproductive temperature suggests that these changes could be a crucial mechanism affecting future cell proliferation and differentiation.This finding sheds light on one possible underlying mechanism of long-term plasticity and phenotypic traits in fish as recently described (Colson et al. 2019, Fellous et al. 2021 ).As oogenesis and embryonic development are both sensitive life stages (Dahlke et al. 2020 ), more 'consequences', or adaptations, are expected the earlier this stressor comes into play.Further studies are needed to evaluate the longterm implications for offspring phenotypic traits depending on maternal mRNA deposition.The alteration in maternal mRNA between temperature regimes seems to be a key mechanism.
Genes consistently regulated by elevated temperature, and their expression in ovarian follicles Among eight selected genes that displayed a change in embryonic expression in response to elevated maternal temperature, four showed the same significantly altered expression within ovarian follicles prior to spawning, suggesting that differential embryonic expression may be initiated in the mother long before final maturation of the ovarian follicles.Elevated temperatures from 17 • to 18.5-21 • C during the reproductive season of the nest-building stickleback affect the transcriptomic profile of the ovary as well as the mature egg, which is further linked to changes in epigenetic reprogramming as well as phenotypic effects in offspring (Fellous et al. 2021 ).Nevertheless, it is largely unknown which of the regulated ovarian transcripts actually affect embryonic developmental potential, when and in which tissue they become regulated.Cheung et al. ( 2019) , using transcriptomic analysis and genome editing, demonstrated that two genes showing differential expression in eggs of poor quality in zebrafish were indeed essential for proper fertilization.Similar studies may reveal additional genes that, upon dysregulation, lead to significant effects on reproductive and offspring success.In cold-temperate cod, ovary transcripts from several hundred genes accumulate or degrade throughout oogenesis, with large differences between each stage of maturation (Kleppe et al. 2014 ).This reflects complex transcriptome dynamics with many follicular transcripts being used prior to final maturation, and thus may not be deposited into the mature egg for embryonic development.
Different maternal-effect genes likely have unique expression dynamics during oogenesis and early embryogenesis, so that dysregulation leading to effects on embryo developmental potential takes place at different times, depending on the gene.Here, pnpla8l and sgpp2 both showed temperature-induced higher levels of transcripts in embryos as well as in ovarian follicles in December before spawning as the LC diameter corresponded to early or late vitellogenic follicles ( Fig. 1 ).Pnpla8 codes for a calcium-independent phospholipase, involved in maintenance of lipid homeostasis, lipid mediator synthesis, and cytoprotection.In mammals, dysregulation of pnpla8 may lead to several conditions, including metabolic diseases (Hara et al. 2019 ).Sgpp2 encodes a sphingosine-1phosphate (S1P) phosphatase that has a role in controlling the metabolism of S1P.Induction of Sgpp2 has been associated with pro-inflammatory signalling (Mechtcheriakova et al. 2007 ).There is no information available on the functions of pnpla8 and sgpp2 in fish.However, they may have unknown functions for oogenesis and/or embryonic development, since their expressions are regulated in both ovarian follicles and embryos of cod, depending on temperature.Increased mRNA levels of pnpla8 and sgpp2 in response to elevated temperatures during oogenesis might reflect compensatory mechanisms in the form of increased cleavage of fatty acids and proinflammatory signalling.
We further identified two genes, arhgap20l and slc41a1 , with a temperature-induced lower transcript levels in embryos and follicles in December.Developmental expression data exist (White et al. 2017 ), but apparently not on the function of arhgap20l in fish.However, mammalian arhgap20 is involved in neurite outgrowth (Yamada et al. 2005 ).A recent study also found a link between arhgap20 and composite traits at birth in sheep (Esmaeili-Fard et al. 2021 ).Slc41a1 has been suggested to encode a magnesium transporter important for regulation in multiple tissues in fish, including kidney, intestine, and gill (Islam et al. 2013, Kodzhahinchev et al. 2017 ).Loss of function of slc41a1 in zebrafish demonstrated that this protein is essential for Mg 2 + homeostasis (Arjona et al. 2019 ), and Mg 2 + levels in fish may be regulated in response to water salinity and diet via Slc41a1 (Takvam et al. 2021 ).Mg 2 + is essential for gastrulation and neural fold change in vertebrates (Komiya et al. 2014 ).Thus, low transcript levels of slc41a1 in cod embryos in elevated temperature regimes may have contributed to the observed reduced survival.
Taken together, we show that some genes ( pnpla8l , sgpp2 , arhgap20l , and slc41a1 ) involved in various processes, such as lipid metabolism, pro-inflammatory signalling, neurite outgrowth, and Mg 2 + homeostasis, are differentially expressed in early cod embryos in response to elevated temperatures during oogenesis, and that their dysregulation is reflected in vitellogenic follicles.

Temperature-mediated metabolic shifts in broodstock liver
It is well known that fish during gametogenesis allocate energy via the liver to gonads for reproduction.Here, we considered whether the liver's content of amino acids and other nitrogenous metabolites was affected by temperature.Although we found large individual variations in the liver, fish exposed to 9 • C showed an overall significantly higher level of free amino acids after spawning compared to control fish at 6 • C. In addition, β-alanine, which is an intracellular metabolic buffer (Dolan et al. 2019 ), was reduced in the 9 • C regime compared to 6 • C. We detected a significantly higher level of taurine at 9 • C, which may be related to an increased metabolic rate (Pörtner andPeck 2010 , Little et al. 2020 ).The alterations in overall N -metabolites within the liver indicate a shift in energy (Skjaerven et al. 2022 ).Further experiments are essential to understand how elevated temperatures influence the next generation via regulations in the metabolism.The analysis of free amino acids represents an initial foray into investigating the concept of metabolic programming and shedding light on the vulnerability and potential adaptability of organisms in the face of climate change.

Conclusions
In this study, we combine reproduction biology and molecular techniques to possibly predict the adaptation process to a warmer climate for populations of Atlantic cod.In evolutionary ecology, a reaction norm aims at explaining the phenotypic variation based on the interaction between the genotype and the environment (Li et al. 2018 ).A refinement has been proposed as a 'genomic reaction norm', which represents the range of gene expression phenotypes by a given genotype along an environmental gradient (Oomen and Hutchings 2022 ).Different environmental temperatures adjust enzymatic reactions, transcripts as well as the physiological metabolism in general in ectotherms (Little et al. 2020 ).It is tempting to speculate that the change in maternal mRNAs into the eggs reported here might be a systematic mechanism at the most sensitive live-history stage whereby one genotype alters phenotypic expression by varying mRNAs as a heritability mechanism to cope with the environment to increase sur-vival possibilities of the offspring.However, more studies are clearly needed to scrutinize phenotypic variations caused by altered levels of mRNA in response to maternal temperature.Furthermore, enormous complexity exists in the interrelations between genetic and environmental factors for developmental traits, and, as such, experimental design protocols must carefully take into consideration in-built individual variation as different reaction norms can exist for every genotype, phenotypic trait, and environmental variable (Li et al. 2018 ).However, our results on changing maternal mRNA messages in the cod eggs, depending on incubation temperature, should be helpful to increase insights into the existence of a sophisticated, fundamental adaptation process.
Thermal tolerance is species-specific but also varies in relation to life-history stages.Here, we studied maturing and spawning females as well as embryonic stages, where the latter two are known to be the most sensitive stages (Dahlke et al. 2020 ).However, we show that elevated temperatures influence the next generation already during the reproductive cycle of Atlantic cod by targeting maternal mRNAs in the embryo, a process in line with up-and down-regulated genes in ovarian follicles of pre-spawning adults several months before spawning.Increased temperatures induced adjustments in the transcriptome of cleavage-stage embryos proportional to the temperature increase relative to the control.Most importantly, our results identified changes in maternal-effect genes.Under climate change, Atlantic cod may present stock (habitat)dependent responses, with more productive stocks in the north (e.g.northeast Arctic cod) and less productive stocks in the south (e.g.North Sea cod) (Kjesbu et al. 2022 ).Even so, the temperature response in maternal-effect genes at the embryo stage might be a key component of adaptation to climate change, though within a feasible range of response patterns to stressors.Further studies are needed to follow the implications of altering maternal gene products and their phenotypic consequences for the next generation.

Figure 1 .
Figure 1.Experimental design: Atlantic cod stock from the southwestern coast of Norway were kept at 6 • , 9 • , and 12 • C from the onset of secondary growth of ovarian follicles until the end of spawning.Effect of temperature regime: tracing the early programming of the next generation (offspring) with maturing o v arian f ollicle transcripts (oocyte biopsy); clea v age stage embry os (RNA seq., embry onic phenot ypes); and the nutritional st atus in spawners (free amino acids in liver).N: nucleus.

Figure 2 .
Figure 2. De v elopmental success parameters according to temperature regime in the six cod egg batches analysed by RNA seq.(a) Egg diameter according to temperature regime.(b) Survival rate at blastula stage according to temperature regime.(c) Hatching rate according to temperature regime.(d) Timing of hatching embryos after fertilization: blue dots correspond to 6 • C, y ello w dots correspond to 9 • C, and size of the dots is related to the number of hatched embryos.Differences in egg diameter between temperature regimes tested with a Kruskal-Wallis test and Dunn's multiple comparisons test ( P < .05).Pairwise Chi-square tests in survival rate at blastula stage and hatching rate between temperature regimes ( P < .05).Panels (a)-(c) display 10th (whisker), 25th (box), 75th (box), and 90th percentile (whisker), mean (black symbol) and median (thick line) (box), and individuals (grey symbol).

Figure 3 .
Figure 3. Effects of environmental temperature during cod oogenesis on the next generation.(a) PCA plot of transformed normalized mean counts per sample quantified by RNA sequencing at three temperatures (blue: 6 • C, y ello w: 9 • C, and red: 12 • C).Shaded areas represent 85% confidence interval of variance per regime.(b) Violin plot indicating the proportion of genes exhibiting lower and higher transcript levels for each comparison.Log2 fold change scores of significantly differential expressed genes (DEGs) are represented by jittered dots [blue dots: significantly higher expression levels, and red: significantly lo w er e xpression le v els in 9 • C compared to 6 • C (lo w er panel), 12 • C compared to 6 • C (middle panel), and 12 • C compared to 9 • C (upper panel), respectively] and number of significantly altered transcripts are shown for each comparison in text boxes.(c and d) Heatmaps of mRNA e xpression le v els of significantly affected sequences of scaled and centred read counts for all significantly DEGs between (c) 6 • C vs. 9 • C, and (d) 6 • C vs. 12 • C. Genes are shown in rows, samples in columns.Expression levels indicated by colours, with blue to red indicating minimum and maximum e xpression le v els f or each gene, respectively.Hierarchical clustering of samples and sample references are indicated on the top of each heatmap.(e and f) Venn diagrams of transcripts showing greater (e) or lesser (f) expression in response to increasing temperature.

Figure 5 .
Figure 5. Expression of selected genes in de v eloping cod o v arian f ollicles sho wn b y RNA-sequencing to present higher (a-d) or lo w er (e-h) transcript le v els in embryos from mothers exposed to increased water temperatures during oogenesis.For each sampling point, gene expression values are relative to the average expression of the reference genes actin beta and elongation factor 1 alpha , calibrated to the average expression of the 6 • C regime, and shown as mean ± SEM, N = 5-6.Significances between regimes * P < .05 and * * P < .01.Blue, orange, and red bars represent 6 • , 9 • , and 12 • C, respectively.Rel., relative; expr., expression.

Figure 6 .
Figure 6.Free amino acid and N-metabolite composition in female cod liver at spawning.Heatmap scaled by tank mean (tank mean: max mean), including liver-free amino acids (categorized depending on the R-group) and N-metabolite composition.Colour intensity of the scalebar indicates the relative composition measured from low (white) to high (dark blue) (a).Temperature mean values of alanine, valine, proline, tyrosine, aspartic acid, taurine, β-alanine, and ammonium in liver from 6 • to 9 • C (b-i, respectively).Data are presented as mean ± SEM, N = 3. Significance between regimes * * P < .01.

Figure 7 .
Figure 7. Ov ervie w of influence of ele v ated temperatures (9 • C and 12 • C vs. 6 • C) during the reproductiv e cy cle on the ne xt generation (maternal mRNAs) of Atlantic cod embryos, including the expression of selected genes in ovarian follicles of pre-spawners and the liver content of the broodstock.Down and up arrows correspond to a decrease and an increase, respectively, in the measured variable.Developmental time is measured from fertilization to embryo hatching and survival rate is assessed at blastula stage.