Gene expression patterns of red sea urchins (Mesocentrotus franciscanus) exposed to different combinations of temperature and pCO2 during early development

Background The red sea urchin Mesocentrotus franciscanus is an ecologically important kelp forest herbivore and an economically valuable wild fishery species. To examine how M. franciscanus responds to its environment on a molecular level, differences in gene expression patterns were observed in embryos raised under combinations of two temperatures (13 °C or 17 °C) and two pCO2 levels (475 μatm or 1050 μatm). These combinations mimic various present-day conditions measured during and between upwelling events in the highly dynamic California Current System with the exception of the 17 °C and 1050 μatm combination, which does not currently occur. However, as ocean warming and acidification continues, warmer temperatures and higher pCO2 conditions are expected to increase in frequency and to occur simultaneously. The transcriptomic responses of the embryos were assessed at two developmental stages (gastrula and prism) in light of previously described plasticity in body size and thermotolerance under these temperature and pCO2 treatments. Results Although transcriptomic patterns primarily varied by developmental stage, there were pronounced differences in gene expression as a result of the treatment conditions. Temperature and pCO2 treatments led to the differential expression of genes related to the cellular stress response, transmembrane transport, metabolic processes, and the regulation of gene expression. At each developmental stage, temperature contributed significantly to the observed variance in gene expression, which was also correlated to the phenotypic attributes of the embryos. On the other hand, the transcriptomic response to pCO2 was relatively muted, particularly at the prism stage. Conclusions M. franciscanus exhibited transcriptomic plasticity under different temperatures, indicating their capacity for a molecular-level response that may facilitate red sea urchins facing ocean warming as climate change continues. In contrast, the lack of a robust transcriptomic response, in combination with observations of decreased body size, under elevated pCO2 levels suggest that this species may be negatively affected by ocean acidification. High present-day pCO2 conditions that occur due to coastal upwelling may already be influencing populations of M. franciscanus. Supplementary Information The online version contains supplementary material available at 10.1186/s12864-020-07327-x.

acidi cation continues, warmer temperatures and higher pCO 2 conditions are expected to increase in frequency and to occur simultaneously. The transcriptomic responses of the embryos were assessed at two developmental stages (gastrula and prism) in light of previously described plasticity in body size and thermotolerance under these temperature and pCO 2 treatments.
Results: Although transcriptomic patterns primarily varied by developmental stage, there were pronounced differences in gene expression as a result of the treatment conditions. Temperature and pCO 2 treatments led to the differential expression of genes related to the cellular stress response, transmembrane transport, metabolic processes, and the regulation of gene expression. At each developmental stage, temperature contributed signi cantly to the observed variance in gene expression, which was also correlated to the phenotypic attributes of the embryos. On the other hand, the transcriptomic response to pCO 2 was relatively muted, particularly at the prism stage.
Conclusions: M. franciscanus exhibited transcriptomic plasticity under different temperatures, indicating their capacity for a molecular-level response that may facilitate red sea urchins facing ocean warming as climate change continues. In contrast, the lack of a robust transcriptomic response, in combination with observations of decreased body size, under elevated pCO 2 levels suggest that this species may be negatively affected by ocean acidi cation. High present-day pCO 2 conditions that occur due to coastal upwelling may already be in uencing populations of M. franciscanus.

Background
The red sea urchin Mesocentrotus franciscanus (A. Agassiz, 1863) is an ecologically and economically valuable species found along the Paci c Coast of western North America [1]. In subtidal areas, especially within kelp forests, these echinoderms are herbivorous ecosystem engineers that can shape the ow of resources within marine habitats [2]. Overgrazing by M. franciscanus, often in combination with overgrazing by the purple sea urchin Strongylocentrotus purpuratus, can lead to the formation of urchin barrens in which macroalgal communities are severely reduced or depleted [3,4]. Red sea urchins also function as prey to animals at higher trophic levels, including spiny lobsters and sea otters [5][6][7]. In addition to its removal by natural predators, M. franciscanus is widely collected as a lucrative wild shery species. Fisheries in Mexico, the United States, and Canada harvest M. franciscanus for their gonads (i.e., roe) that supply domestic markets as well as international exports, principally to Japan [8,9]. Over recent In this study, M. franciscanus embryos were raised under a combination of two temperatures (13 °C or 17°C ) and two pCO 2 levels (475 matm or 1050 matm) that re ect current and future ocean conditions in their natural habitat [15,[43][44][45]. This generated four different treatment combinations: 1) 17 °C and 1050 matm pCO 2 , 2) 17 °C and 475 matm pCO 2 , 3) 13 °C and 1050 matm pCO 2 , and 4) 13 °C and 475 matm pCO 2. In the highly dynamic CCS, treatment combinations #2-4 are currently measured during and between upwelling events [15,[43][44][45]. Future ocean conditions are represented by treatment combination #1 (i.e., the simultaneous occurrence of 17 °C and 1050 matm pCO 2 ) given continued ocean warming and acidi cation. Additional detail regarding the selection of these temperature and pCO 2 treatment levels is located in the Methods. Gene expression patterns were assessed at both the gastrula and prism embryo stages. We also discuss the gene expression results within the context of previously reported physiological assessments from this experiment, including body size and thermotolerance [46]. Here, we describe the effects of the temperature and pCO 2 treatments at the molecular level and whether they relate to observations made at the level of the organism.
Temperature elicited a robust transcriptomic response at both developmental stages. Gene expression analyses indicated that the warmer temperature (i.e., 17 °C) induced a cellular stress response, amongst other processes. Additionally, the variation in gene expression that was signi cantly correlated to the temperature treatment was also signi cantly correlated to embryo body size and thermotolerance, characteristics that were neutrally or positively in uenced by the warmer temperature treatment [46]. In contrast, the transcriptomic response to the pCO 2 treatment was comparatively muted. This minor molecular-level response may explain the reduction in embryo body size that is observed under elevated pCO 2 levels (i.e., 1050 matm) [46]. Overall, we examined a valuable shery species that is capable of dramatically shaping coastal ecosystems, and determined that during early development M. franciscanus exhibits different magnitudes of transcriptomic plasticity in response to two climate change-related stressors. This study provides much needed insight into a species that is important for many local sheries on the Paci c coast of North America, facilitating our understanding of how M. franciscanus development is affected by current ocean conditions, as well as our predictive capacity of how this species will respond to future ocean change.

Summary statistics and overview of RNA-seq
The samples used for RNA-seq were generated from triplicate cultures of embryos raised at each of the four combined temperature and pCO 2 treatments (i.e., 12 total cultures) (see Additional le 1). Each sample was collected as a pool of 5000 embryos from each of the 12 cultures at both the gastrula and prism stages during development to produce a total of 24 samples used for RNA extractions and library preparation. Sequencing of the 24 libraries yielded a total of 728,782,735 100-bp single reads. After quality trimming, an average of 30.3 ± 1.3 million reads per library remained. FASTQC reports [47] of trimmed sequences showed high sequence quality (>30) with limited adapter contamination or presence of overrepresented sequences. Per-library mapping e ciency to the developmental transcriptome [48] using RSEM [49] was at an average of 52.6%. The presence of mitochondrial rRNA appeared to have contributed to the percentage of unmapped reads, although mapping rate may have also been affected by the completeness of the reference transcriptome.
Developmental stage in uenced transcriptomic patterns A principal component analysis (PCA) of sample-to-sample distances showed that differences in gene expression pro les were primarily between the two developmental stages, gastrula and prism (Fig. 1a). Principal Component (PC) 1 captured the majority of the variance (67.5%) and revealed a clear separation between gastrula and prism stage embryos, while PC2 only captured 3.8% of the variance. Indeed, a permutational multivariate ANOVA across all 24 samples with developmental stage, temperature treatment, and pCO 2 treatment as xed factors, revealed that developmental stage explained 66.4% of the variance (p = 0.001) (Fig. 1b). In contrast, temperature treatment explained only 4.1% of the variance (p = 0.041) and pCO 2 treatment explained only 2.3% of the variance (p = 0.207). All factor interactions were not signi cant (p > 0.05). Results from gene expression analyses across all samples independently of stage (i.e., gastrula and prism stages not analyzed separately) are available in Additional le 2. Because we have previously explored the differences in gene expression patterns across M. franciscanus during early development [48] and it is not the main focus of the current study, from here onward we report separate gene expression analyses for the gastrula and prism stages.
Temperature and pCO 2 affected gastrula gene expression Separate PCA plots were generated for the gastrula and prism stages. At the gastrula stage, we generally observed that both temperature and pCO 2 treatments appeared to drive differences in gene expression patterns across samples. A PCA of only the gastrula stage showed that replicate samples grouped together (Fig. 1c). Here, PC1 captured 23.8% of the variance and was found to have a highly signi cant negative correlation to the temperature treatment (Fig. 2a). PC2 captured 12.0% of the variance (Fig. 1c) and was found to have a signi cant positive correlation with the pCO 2 treatment (Fig. 2a). Using average embryo length measurements from this experiment (previously reported in [46]), both PC1 and PC2 were also found to be negatively correlated with gastrula body size (Fig. 2a).
Upon examining the PC loadings for the gastrula stage using the PCAtools package [50], the genes most responsible for variation along PC1 included an elongation factor 1-alpha (loading = -0.0585) and a transcription factor SUM-1-like gene (loading = -0.0505) (Additional le 3). Genes contributing variation to PC2 included a poly(A)-speci c ribonuclease PARN (loading = 0.0458) and a putative DNA polymerase gene (loading = 0.0439). A rank-based gene ontology (GO) analysis was performed following the GO_MWU package [51] in R. Using complete sets of loading values calculated from the PCA (Additional le 3), this analysis identi ed GO categories enriched by genes contributing variance to the PCs. GO terms related to regulation of gene expression, ion binding, and DNA recombination were enriched by variable genes in PC1 (Additional le 4a). Genes contributing variation to PC2 enriched GO categories associated with heat shock protein binding, peptide metabolic process, and amide biosynthetic process (Additional le 4b).
A permutational multivariate ANOVA revealed that at the gastrula stage, 20.3% of the variance was explained by temperature treatment (p = 0.001) (Fig. 1d). Differential expression (DE) analyses conducted in limma [52] identi ed differentially expressed genes (relatively up-and down-regulated) between gastrula raised under different temperature treatments. A total of 2049 genes were signi cantly upregulated in embryos raised at 17 °C relative to embryos raised at 13 °C (adjusted p < 0.05) (Fig. 3a)  with up-regulated genes included macromolecule catabolic process, ion binding, and active transmembrane transporter (Fig. 4b, Additional le 5b). A total of 166 genes were down-regulated in embryos raised at 1050 matm relative to embryos raised at 475 matm (Fig. 3b), including a keratinassociated protein 4-4-like (log 2 FC = -1.62, adj. p = 0.008) and a carbonic anhydrase 14-like isoform X3 gene (log 2 FC = -0.89, adj. p = 0.047) (Additional le 3). Enriched GO categories included macromolecule biosynthetic process, macromolecule metabolic process, and nucleic acid binding (Fig. 4b, Additional le 5b). The interaction between temperature and pCO 2 factors explained 7.1% of the variance observed at the gastrula stage, but the interaction was not signi cant (p = 0.440) (Fig. 1d).
Temperature was the primary factor affecting prism gene expression Similar to the gastrula stage, the PCA of only the prism stage showed a separation of samples by experiment treatment with sample replicates grouping together (Fig. 1e). PC1, which captured 27.6% of the variance, was found to have highly signi cant negative correlations to the temperature treatment, prism body size, and thermotolerance (Fig. 2b). Prism body size for each sample was estimated using average embryo length and LT 50 (i.e., the temperature at which 50% mortality occured) measurements from this experiment that were previously reported in [46]. PC2 captured 9.9% of the variance (Fig. 1e) and had a signi cant negative correlation with the pCO 2 treatment (Fig. 2b).
Loadings within PC1 showed that genes responsible for most of the variation included heme-binding protein 2-like (loading = -0.0545) and a putative tolloid-like protein 1 gene (loading = -0.0483) (Additional le 3). Genes contributing variance to PC2 included an elongation factor 1-alpha (loading = 0.111), a Fbox/WD repeat-containing protein 7-like (loading = -0.0657), and a FK506-binding protein 5-like gene (loading = -0.0575) (Additional le 3). Using GO_MWU and the loading values for the complete set of genes, enriched GO categories for PC1 were identi ed. These included GO terms related to ion transport, regulation of gene expression, methylated histone binding, RNA methyltransferase, pseudouridine synthesis, antioxidant, cellular response to DNA damage stimulus, and response to oxidative stress (Additional le 4c). Genes contributing variation to PC2 enriched GO categories associated with ion binding, oxidoreductase, and metabolic process (Additional le 4d).
A permutational multivariate ANOVA revealed that at the prism stage, 27.2% of the variance was explained by the temperature treatment (p = 0.001) (Fig. 1f). DE analysis showed a total of 3842 genes were up-regulated in embryos raised at 17 °C relative to those raised at 13 °C (Fig. 3c). These up-regulated genes included a heme-binding protein 2-like (log 2 FC = 3.67, adj. p < 0.001), a putative tolloid-like protein 1 (log 2 FC = 3.21, adj. p < 0.001), and a heat shock 70 kDA protein 12A-like gene (log 2 FC = 0.91, adj. p < 0.001) (Additional le 3). GO analysis identi ed GO categories enriched in these up-regulated genes, which included oxidoreductase, response to oxidative stress, ion transmembrane transporter, and ATP metabolic process (Fig. 5a, Additional le 5c). A total of 3434 genes were down-regulated in embryos raised at 17 °C relative to those raised at 13 °C ( The pCO 2 treatment only explained 9.3% of the variance at the prism stage and was not signi cant (p = 0.091) (Fig. 1f). In fact, only 4 genes were up-regulated when comparing the 1050 matm to the 475 matm pCO 2 treatment at the prism stage (Fig. 3d), including a hypoxia-inducible factor 1-alpha-like isoform X1 gene (log 2 FC = 0.84, adj. p = 0.005) (Additional le 3). GO categories enriched in up-regulated genes were related to ATPase coupled to transmembrane movement of ions, and regulation of gene expression (Fig.  5b, Additional le 5d). A total of 64 genes were down-regulated in prism embryos raised at 1050 matm pCO 2 relative to those raised at 475 matm (Fig. 3d), and enriched GO terms related to oxidoreductase and G protein-coupled receptor (Fig 5b, Additional le 5d). Lastly, the interaction between temperature and pCO 2 factors explained only 7.2% of the variance at the prism stage and was not signi cant (p = 0.362) ( Fig. 1f).

Discussion
In this study, we examined how the gene expression patterns of gastrula and prism embryos varied by the developmental temperature and pCO 2 conditions under which they were raised. We also assessed whether the transcriptomic results aligned with the morphometric and physiological results previously reported in [46]. Although both temperature and pCO 2 can in uence rates of sea urchin development [34,53], any potential differences in developmental timing should not have impacted the results of this study because samples were collected based on developmental progression to the desired embryonic stages as detailed in the Methods, rather than by hours post-fertilization. Overall, we found that while transcriptomic patterns varied by developmental stage, temperature had a dominant effect on changes in gene expression while pCO 2 elicited a more subtle transcriptomic response that was largely limited to the gastrula stage. Experimental conditions impacted genes related to the cellular stress response, transmembrane transport, metabolic processes, and the regulation of gene expression.
In terms of experimental design, embryos were obtained by evenly pooling eggs from ve females and fertilizing them with sperm from a single male to produce all full or half siblings. Admittedly, there are caveats to this approach. The results presented here may only be representative of a small subset of the population, or they may be driven by the quality of the particular male selected to fertilize the eggs. Upon including data from our previous study that examined gene expression patterns during M. franciscanus early development [48], a PCA showed that, although samples primarily grouped by developmental stage, there is a clear distinction between the embryos of the two studies (Additional le 6). This is likely due to a combination of genetic and environmental differences between the two source populations, as the adult urchins were collected from different sites and during different years. Indeed, in the purple sea urchin S. purpuratus, genetic variation has been shown to in uence transcriptomic responses to temperature and pCO 2 stress during early development [36,54]. Given that the data presented here represents a limited selection of the genetic variation that exists in this species, the results should be interpreted with caution. We therefore recommend that additional studies be performed within other M. franciscanus populations and with multiple male-female crosses to determine if our results are unique to this study. Nevertheless, this approach was implemented in an effort to limit genetic variability and malefemale interactions that may have otherwise confounded the molecular results.
All samples used for RNA extractions were each composed of a pool of 5000 individuals and should thus represent the same mixture of genotypes. Therefore, we do not expect differences in gene expression patterns to be due to genetic variability between embryo cultures, particularly because a low incidence of mortality was observed during the experiment, although it was not directly measured. In the absence of selection, the observed variability in gene expression, body size, and thermotolerance between embryos raised under different experimental treatments re ect plasticity exhibited by M. franciscanus during its early development. We discuss this plasticity, and how it may relate to embryo performance under different conditions that M. franciscanus are likely to experience in their natural environments currently and in the future under ocean change scenarios.
Gene expression varied by developmental stage: General patterns Developmental stage (gastrula or prism embryos) was the primary factor driving differences in gene expression patterns across samples ( Fig. 1a and 1b). In a past study, we raised cultures of M. franciscanus embryos in a single laboratory environment that mimicked average, non-stressful conditions in situ (i.e., 15 °C and 425 matm pCO 2 ) and documented signi cant transcriptomic differences between gastrula and prism stages [48]. Therefore, there are many alterations in gene expression between these stages that occur as a result of development and are independent of differences in environmental temperature and/or pCO 2 conditions. This is also evident in Fig. 1a in which gastrula samples do not cluster with prism samples that share the same experimental treatment.
Because comparing gastrula versus prism gene expression patterns was not a goal of this study, no direct differential expression analyses were performed between stages, although gene expression analyses performed independently of stage (i.e., without analyzing gastrula and prism stages separately) are reported in Additional le 2. Nevertheless, embryos at each developmental stage exhibited different transcriptomic responses to temperature and pCO 2 treatments. For instance, many more genes were differentially expressed due to temperature at the prism stage than at the gastrula stage (Fig. 3).
Additionally, the pCO 2 treatment explained a signi cant amount of variance in gene expression in gastrula embryos, but not later at the prism stage ( Fig. 1d and 1f). Similarly, the morphometric response to temperature and pCO 2 treatments varied by stage, in which only pCO 2 affected gastrula embryos by reducing body size under elevated pCO 2 conditions (i.e., 1050 matm) [46]. On the other hand, temperature was the dominant factor at the prism stage, with warmer conditions (17 °C) increasing body size, offsetting the stunting effect of high pCO 2 [46]. The observed patterns between gene expression and body size will be described in greater detail later in the Discussion.
Different life stages are predicted to have different sensitivities to stress [23]. The variability between gastrula and prism stress responses may be explained by a difference in stage-speci c vulnerability.
During the gastrula stage, the archenteron is formed from invagination of the embryo's vegetal plate [55], a fundamental process known as gastrulation that is essential for successful development in metazoans [56]. At the prism stage, the embryo develops its digestive tract and skeletal rods, which are vital structures required for the embryos to eventually become feeding, planktotrophic larvae [57,58]. Accordingly, differences in responses to environmental conditions between these two stages are likely re ective of the distinct processes undergone by these embryos to ensure their continued developmental progression.
The variability between stages could also be due to the timing and duration of exposure to stress. The effects of a stressor can become increasingly deleterious as the length of exposure continues, and organisms not permitted adequate time to recover may exhibit increasingly poor performance. Furthermore, during development there may be negative carry-over effects that persist into later life stages [59,60]. Alternatively, organisms may acclimate to stressful conditions over time, and are therefore less adversely affected by a stressor following the initial exposure. For example, in the coral Acropora hyacinthus, the immediate transcriptomic response to heat stress was much higher than the transcriptomic response following 20 hours of exposure to warmed conditions [61]. Thus, it remains important to acknowledge that organisms may responds differently to various environmental stressors depending on their life history as well as the timing and duration of the exposure. was signi cantly correlated to the temperature treatment (Fig. 2a). In general, the observed temperature effects on gene expression at the gastrula stage were approximately akin to those reported for the purple sea urchin S. purpuratus [36], whose biogeographical distribution overlaps with that of M. franciscanus.
Here, DE analysis revealed that gastrula raised in the higher temperature treatment (i.e., 17 °C) expressed genes associated with a cellular stress response. Gastrula embryos of S. purpuratus that were raised under an 18 °C temperature treatment exhibited a comparable cellular stress response by up-regulating genes associated with cellular responses to reactive oxygen species and unfolded proteins [36]. We also found that M. franciscanus gastrula embryos raised under the warmer treatment exhibited transcriptomic patterns indicative of increased transmembrane transport, while embryos from the colder treatment appeared to decrease metabolic processes. Similarly, S. purpuratus increased expression of ion channel, cell-cell signaling, and metabolism genes at warmer temperatures [36]. Lastly, temperature appeared to impact the regulation of gene expression in M. franciscnus gastrula embryos, including genes related to epigenetic mechanisms. Temperature also appeared to in uence how gene expression was regulated in S. purpuratus gastrula embryos with higher temperatures leading to a down-regulation of genes related to transcription and RNA processing [36].
Despite similarities in how temperature in uenced gastrula gene expression, unlike in M. francicanus, there was an effect of temperature on S. purpuratus morphology. Speci cally, S. purpuratus gastrula embryos raised at warmer temperatures were signi cantly smaller in size [36], whereas M. franciscanus gastrula embryos did not signi cantly differ in size as a result of temperature [46]. This could re ect differences in experimental design between the studies (e.g., treatment temperatures, breeding designs, and urchin collection sites), or it could re ect differences that exist at the species level. Unlike S. purpuratus, the temperature response of M. franciscanus gastrula embryos that occurred at the molecular level was not re ected at the organismal level. We postulate that the transcriptomic differences between gastrula raised at 17 °C and 13 °C served to compensate for direct temperature effects and allowed the embryos to maintain the same size despite the temperature treatments. Below, we explore with greater detail how temperature affected the expression patterns of genes associated with the cellular stress response, transmembrane transport, metabolism, and gene expression regulation in M. franciscanus gastrula embryos.
Cellular stress response (Gastrula temperature) Response to stress at the cellular level often includes processes to mitigate or remove cell damage [62]. Cell death protein 3, which encodes a protease involved in apoptosis [63], was up-regulated in gastrula embryos raised at 17 °C, indicating that the warmer temperature may have caused stress-induced programmed cell death. DE analysis also provided evidence of DNA damage and repair. This is similar to observations in Acropora corals in which heat stress caused an up-regulation of DNA replication and repair genes [61,64]. Here, GO terms related to DNA metabolic process and DNA recombination were enriched with up-regulated genes under warmer temperature conditions (Fig. 4a). DNA metabolic processes can include both DNA synthesis and degradation for the purposes of replication and repair.
Furthermore, DNA recombination in somatic cells has been identi ed as a critical mechanism for DNA damage repair [65,66]. Taken together, 17 °C temperature conditions appear to induce stress within the gastrula embryos, which undergo response mechanisms to combat cellular damage.
Transmembrane transport (Gastrula temperature) Gastrula embryos raised under warmer temperatures also increased expression of genes related to transmembrane transport, potentially both within and between cells (i.e., cell-cell communication). For instance, genes related to G protein-coupled receptor, cation channel, cell surface receptor signaling pathway, and vesicle-mediated transport were up-regulated in gastrula embryos raised under 17 °C relative to those raised under 13 °C (Fig. 4a). This is also supported by PC loadings, in which genes related to ion binding and cation channel contributed variance to PC1 (Additional le 4a), which was signi cantly correlated to the temperature treatment (Fig. 2a). Increased transport of materials, particularly ions, across cell membranes may indicate osmoregulation and maintenance of homeostasis. This aligns with reports in juvenile sea urchins of the species Loxechinus albus, in which gene expression alterations under elevated temperatures provided evidence of increased active transmembrane transport of sodium and potassium ions [67].
Here, gastrula raised under the lower temperature treatment expressed genes associated with metabolic depression. Speci cally, GO terms identi ed as negative regulation of biological process and negative regulation of metabolic process were signi cantly enriched by genes down-regulated in gastrula embryos raised under 17 °C relative to those raised under 13 °C (Fig. 4a). In this study, metabolic rates of embryos raised under different treatments were not measured at the gastrula stage, but we may expect that, given the effect of temperature on biochemical reaction kinetics, metabolic rate should increase predictably with temperature [68]. Generally, higher metabolic rates have been recorded at warmer temperatures in marine ectotherms [69][70][71][72]. This positive correlation between temperature and metabolism has been observed M. franciscanus at the adult stage [73].

Regulation of gene expression (Gastrula temperature)
Temperature also had an evident effect on the regulation of gene expression in M. franciscanus gastrula embryos. GO terms identi ed as regulation of transcription by RNA polymerase II, translation regulator, and regulation of gene expression were enriched by genes relatively down-regulated in gastrula embryos raised under 17 °C relative to those raised under 13 °C (i.e., genes are comparatively up-regulated in the colder temperature treatment) (Fig. 4a). Histone binding, histone modi cation, chromatin remodeling, and chromatin organization genes were also associated with the 13 °C gastrula treatment. Histone and posttranslational modi cations are examples of epigenetic mechanisms. These, and other epigenetic modi cations can act to regulate gene function without altering the DNA sequence, promoting phenotypic plasticity and potentially modulating the response to different environmental conditions [74][75][76]. Histone variants and modi cations may activate or repress transcription processes by altering chromatin structures, impacting the regions of the genome that are available for transcription [77], and have been shown to mediate responses to changing environmental conditions in marine organisms [78][79][80].
Additional analyses such as ChIP-seq (i.e., chromatin immunoprecipitation sequencing) to locate regions targeted by histone modi cations and DNA-binding proteins [81,82] or ATAC-seq (i.e., assay for transposase-accessible chromatin with high-throughput sequencing) to assess genome-wide chromatin accessibility [83] are required to pro le speci c histone variants or modi cations and their impact on gene expression.
Although evidence of transcription regulation was observed, it is di cult to conclude if this led to an increase or decrease of gene expression, particularly with respect to the functional signi cance of histone modi cations and chromatin remodeling. These mechanisms are much less studied in marine invertebrates than other modi cations such as DNA methylation [74,75], and although our data support that these modi cations occurred in response to the environment, additional approaches are required to determine the precise modi cations, their locations, and their impact on gene expression. In this study, there were more up-regulated than down-regulated genes in the 17 °C versus the 13 °C gastrula treatment, and the up-regulated genes appeared to have a greater log 2 FC in expression (Fig. 3a). Nevertheless, future studies pairing comparative epigenetic analyses with transcriptomic approaches are required to elucidate how these various mechanisms in uence gene expression in response to different environmental conditions in M. franciscanus. pCO 2 in uenced gene expression of gastrula embryos The pCO 2 treatment in uenced gene expression patterns at the gastrula stage, although to a lesser degree than temperature, and the interaction between the two factors was not signi cant (Fig. 1d).
Gastrula pCO 2 conditions explained 13.2% of the observed variance (p = 0.021) with 9 genes up-regulated and 166 genes down-regulated in embryos raised under 1050 matm pCO 2 relative to those raised under 475 matm pCO 2 (Figs. 1 and 3). Additionally, PC2, which captured 12.0% of the gene expression variance at the gastrula stage, was signi cantly correlated to the pCO 2 treatment (Fig. 2a). In this study, we anticipated that the 475 matm pCO 2 treatment was not stressful, as it represented the average ambient pCO 2 levels M. franciscanus regularly experience in their natural habitat [44]. Evidence suggests that calcifying marine organisms such as M. franciscanus are sensitive to declines in ocean pH (i.e., increases in pCO 2 levels) [84][85][86], and while M. franciscanus may periodically experience elevated pCO 2 conditions in nature during upwelling events [15,43,45,87], the 1050 matm pCO 2 treatment was expected to induce a stress response.
While the effect of pCO 2 on gastrula gene expression patterns was less than the effect of temperature, there was a pronounced impact of pCO 2 conditions on gastrula body size. Gastrula raised under elevated pCO 2 conditions (i.e., 1050 matm) were signi cantly smaller than those raised under 475 matm [46]. Therefore, pCO 2 elicited a relatively muted transcriptomic response compared to temperature, but appeared to have a much greater in uence at the organismal level. Below we discuss the expression patterns of genes affected by the gastrula pCO 2 treatment, which included those related to metabolism and ion transport.

Metabolism and ion transport (Gastrula pCO 2 )
Gastrula raised under the lower pCO 2 treatment (i.e., 475 matm) expressed genes that signi cantly enriched GO terms related to several macromolecule biosynthetic processes (Fig. 4b). In contrast, those raised under the elevated pCO 2 treatment (i.e., 1050 matm) expressed genes that enriched GO terms associated with macromolecule catabolic processes (Fig. 4b). This may, in part, explain the difference in body size that was observed as a result of the pCO 2 conditions. Gastrula in the 475 matm pCO 2 treatment appeared to construct proteins and other macromolecules to maintain their growth and body size, while gastrula in the 1050 matm pCO 2 treatment underwent catabolic processes, possibly to obtain the energy required to respond to elevated pCO 2 levels. The pCO 2 stress response may include an increase of ion transport as a means of maintaining acid-base equilibrium given elevated H + concentrations under high pCO 2 conditions. GO terms related to ion binding, active transmembrane transporter, and ATPase coupled to movement of substances were enriched by genes up-regulated in gastrula raised in the 1050 matm treatment (Fig. 4b). Similarly, increased expression of ion transport genes has been observed in gastrula embryos of S. purpuratus exposed to moderately elevated pCO 2 levels (e.g., ~800 matm) [39], although the transcriptomic response of S. purpuratus embryos to pCO 2 stress can be in uenced by maternal effects [38].
Temperature was the dominant factor at the prism stage At the prism stage, temperature accounted for 27.2% of the observed variance (p = 0.001) with 3842 genes up-regulated and 3434 genes down-regulated in embryos raised under 17 °C relative to those raised under 13 °C (Figs. 1 and 3). Furthermore, PC1 for the prism stage captured 27.6% of variance in gene expression and was signi cantly correlated to the temperature treatment (Fig. 2b). Unlike at the gastrula stage, responses to temperature that were measured at the molecular level were also observable at the organismal level. Development at the warmer 17 °C treatment led to an increase in prism body size as well as a modest increase in prism thermotolerance [46]. Indeed, prism body size and thermotolerance variables were also highly correlated to PC1 (Fig. 2b). Therefore, the transcriptomic response to temperature appears to have in uenced both growth and resistance to heat stress in M. franciscanus prism embryos. Prism embryos raised at 17 °C exhibited increased expression of genes related to the cellular stress response, transmembrane transport, and metabolic processes, while genes related to DNA repair and the regulation of gene expression were associated with the colder temperature treatment (i.e.,

Cellular stress response (Prism temperature)
Environmental stress can lead to the production of reactive oxygen species (ROS), which can cause oxidative stress if ROS production exceeds the organism's antioxidant or damage repair capacity [88,89].
Oxidative stress, and the response to resulting cellular damage, due to elevated temperatures have been documented across a wide variety of taxa, including algae [90], plants [91], mollusks [92], and shes [93,94]. GO enrichment from the DE analysis identi ed terms including oxidoreductase, antioxidant, response to oxidative stress, and response to reactive oxygen species that were enriched with genes up-regulated in response to the warmer temperature treatment (Fig. 5a). At the gastrula stage, embryos in the 17 °C treatment also expressed genes associated with DNA repair, but there was no evidence of increased macromolecule repair gene expression at the prism stage. This could indicate that while both stages initiated a cellular stress response due to warmer temperatures, there was less cellular damage of nucleic acids incurred by the prism stage.
In general, global change biology research in marine systems has focused on the negative consequences of increasing temperatures associated with ocean warming [95]. However, given the expected rise in variable and extreme weather events, the impact of decreased temperatures is also an important consideration, especially in regions dominated by upwelling. DE analysis indicated that prism embryos in the 13 °C treatment increased expression of genes related to DNA damage and repair. Speci cally, identi ed GO terms included cellular response to DNA damage stimulus, DNA metabolic process, DNA binding, and catalytic, acting on DNA (Fig. 5a). Increased DNA damage as a result of low temperature stress has been recorded in the Paci c white shrimp Litopenaeus vannamei [96]. Although 13 °C is within the range of temperatures that M. franciscanus experience in the Santa Barbara Channel (SBC) where urchins were collected for this study, it is lower than the annual average of ~15 °C [44] and may have generated stress and cellular damage in the prism embryos, leading to the activation of repair mechanisms.
Transmembrane transport and metabolism (Prism temperature) Similar to the gastrula stage, warmer temperature conditions caused an increased expression of genes related to transmembrane transport in prism embryos. Speci cally, GO terms of ion binding, ion transmembrane transporter, cation channel, and sodium and potassium ion transmembrane transporters were enriched with up-regulated genes (Fig. 5a), which may indicate osmoregulation and the maintenance of homeostasis. Prism embryos in the 17 °C treatment also increased expression of genes related to energetic processes (e.g., ATP metabolic process, organic acid metabolic process, lipid metabolic process, cyclic nucleotide metabolic process, and carbohydrate metabolic process) possibly to generate the energy required to support active transmembrane transport of ions and other materials. The up-regulation of genes related to energy production may have also supported the increased growth of prism embryos under warmer temperatures. In contrast, DE analysis revealed an increase in expression of genes related to the negative regulation of biological and metabolic processes in prism embryos raised at 13 °C. This supports the predicted expectation that organisms exhibit decreased metabolism under colder temperatures [68][69][70][71][72].

Regulation of gene expression (Prism temperature)
At the prism stage, the colder temperature treatment exhibited an enrichment of genes related to translation regulator, transcription coregulator, regulation of gene expression, and regulation of transcription by RNA polymerase II (Fig. 5a). Additionally, GO terms identi ed as histone binding, chromatin organization, RNA modi cation, and methylation-dependent protein binding were enriched in genes down-regulated at 17 °C relative to 13 °C (Fig. 5a). Thus, gene expression in prism embryos raised at 13 °C appeared to be epigenetically regulated by histone and RNA modi cations. This differential expression of genes related to gene expression regulation (i.e., down-regulated at the colder temperature relative to the warmer temperature) is similar to the pattern observed at the gastrula stage.
The prism stage exhibited a limited transcriptomic response to pCO 2 In other studies, echinoderms raised under elevated pCO 2 conditions have exhibited altered expression of genes related to skeletogenic pathways, spicule matrix proteins, cellular stress response, ion regulation and transport, apoptosis, metabolism and ATP production [38,39,[97][98][99][100]. Here, the pCO 2 treatment had a relatively minimal effect on prism gene expression patterns. The pCO 2 treatment explained only 9.3% of the observed variance at this stage and was not signi cant (p = 0.091) (Fig. 1f). This contrasts with observations made at the organism level in which elevated pCO 2 resulted in smaller prism embryos, although this could be offset by the positive effect of temperature, which acted as the dominant factor in uencing body size [46]. It is interesting that the transcriptomic response to elevated pCO 2 was more evident in gastrula than in prism embryos particularly because at the prism stage, skeletal rod formation occurs. Although GO terms identi ed as ion binding and ATPase coupled to transmembrane movement of ions were enriched with genes up-regulated at the elevated pCO 2 level (Fig. 5b), we also expected to observe increased stress associated with prism embryos undergoing calci cation processes under lowered pH conditions, but we detected no evidence of this.
It is possible that while there was a clear phenotypic difference in prism embryos raised under high versus low pCO 2 conditions, the transcriptomic changes underlying this difference were much more subtle. Alternatively, the prism stage may simply lack a robust transcriptional response to the 1050 matm pCO 2 treatment. For instance, the Mediterranean sea urchin Paracentrotus lividus exhibits different transcriptomic responses depending on the magnitude of the pH stressor [101]. Decreased pH conditions caused P. lividus embryos to increase their expression of calci cation genes, but not once the pH dropped below a certain threshold [101]. A similar result was observed in S. purpuratus in which embryos raised under a high pCO 2 treatment designed to re ect near-future levels was relatively muted relative to those raised under a more moderate pCO 2 treatment designed to re ect present-day low pH conditions [39]. The authors speculated that the transcriptional response required for acclimating to a more extreme pCO 2 level was too metabolically expensive, and the embryos instead opted to conserve energy to ensure shortterm survival, perhaps until environmental conditions became more favorable [39]. While a failure of embryos to respond at the transcriptomic level may allow for continued successful development under high pCO 2 conditions, there may be important physiological consequences such as the observed reduction in body size [46]. Thus, the lack of a transcriptomic response to high pCO 2 may have important tness consequences for M. franciscanus.
In the green sea urchin Strongylocentrotus droebachiensis, a quantitative genetic breeding design implemented by Runcie and colleagues demonstrated that changes in gene expression as a result of differences in pH exposure were minor relative to gene expression differences as a result of parentage [54]. Thus, minimal transcriptomic responses to pCO 2 may be due to the genetic structure of the sea urchins used in this experiment. Furthermore, the environmental exposure history of the adult urchins may have generated non-genetic parental effects (i.e., transgenerational plasticity), which can also generate a limited transcriptomic response to high pCO 2 [38]. While all embryo cultures for this experiment were composed of the same mixture of progeny from a cross between one male and ve females, it is possible that the sea urchins collected for this experiment may be from a population with a relatively muted transcriptomic response to high pCO 2 conditions.
It has been proposed that selection and local adaptation acts on populations that are regularly exposed to high pCO 2 conditions, such as those that often experience upwelling conditions within the CCS [15], and that these populations may harbor genotypes that are resistant to low pH conditions [40,98,[102][103][104][105]. In S. purpuratus, transcriptomic responses to high pCO 2 levels can vary by the frequency in which the sea urchin populations are exposed to upwelling conditions [40]. In particular, urchins from populations frequently exposed to low pH have greater transcriptomic responses to high pCO 2 than those that experience low pH less often [40]. While the site where the adult sea urchins were collected does experience periods of low pH due to upwelling [43,45], low pH events occur less frequently than at more northern sites within the CCS [15,106,107]. Therefore, the urchins used in this study may be comparatively less adapted towards mounting a transcriptomic response to high pCO 2 .

HSP gene expression
Heat shock proteins (HSPs) act as molecular chaperones in the cellular stress response by assisting in protein transport, protein folding and unfolding, stabilization of denatured proteins, and degradation of misfolded proteins [108,109]. Higher levels of HSPs have been shown to confer increased thermotolerance across a variety of marine taxa [110][111][112][113]. Therefore, we may have expected an upregulation of HSP genes linked with the slight increase in thermotolerance measured at the prism stage [46]. However, there were mixed results regarding differential expression patterns of HSP genes. At both stages, the heat shock 70 kDA protein 12-A like gene was up-regulated in the 17 °C treatment relative to the 13 °C treatment, whereas heat shock 70 kDa protein cognate 5 and heat shock 70 kDA protein 14 were down-regulated. No Hsp70 genes were differentially expressed due to pCO 2 , and no Hsp90 genes were differentially expressed due to temperature or pCO 2 .
Our results contrast with a study in S. purpuratus that found expression of Hsp70 and Hsp90 increased at higher temperatures [36]. However, in other investigations of sea urchin early development, increased Hsp70 expression was generally not observed under moderate warming scenarios. One study found that Hsp70 was not transcriptionally up-regulated in M. franciscanus until larvae were exposed to temperatures at or above 20 °C [32]. A study in S. purpuratus only found induction of Hsp70 occurred at temperatures above 21 °C [37]. Therefore, the 17 °C treatment may not have been extreme enough to induce a clear differential expression of HSP genes. Furthermore, in the green sea urchin Psammechinus miliaris, expression of HSP genes were low during early development relative to expression in adults [114]. The authors suggested that HSP expression was limited during this time [114] because overexpression of HSPs could have negative consequences for successful early development [115].
Therefore, large increases in HSP expression may be restricted during M. franciscanus early development.

Performance under current and future ocean conditions
Moderate ocean warming may be favorable for M. franciscanus early development by providing larger body sizes and increased thermotolerance at the prism stage [46]. The warmer temperature treatment could even mitigate the stunting effect of elevated pCO 2 on prism body size [46]. This effect of warmer temperatures, however, may only be bene cial on a short-term basis. Gene expression analyses indicated that embryos raised under 17 °C responded to cellular stress, and while there were no indications of negative impacts at the phenotypic level, there may be trade-offs and consequences to developing under warmer temperatures such as increased incidences of disease [116]. Prolonged heat exposure may eventually become detrimental, and negative carry-over effects can arise at later life stages [59,60].
Additionally, the observed plasticity at 17 °C may not extend to more severe warming scenarios. For example, a study in adult M. franciscanus found that although mortality did not vary between urchins acclimated to 15 °C or 18 °C, mortality was signi cantly higher at a more extreme temperature of 21 °C [31]. Nevertheless, our study revealed that M. franciscanus exhibited a sizeable transcriptomic response to 17 °C at both developmental stages. At the urchin collection site, temperatures of 17 °C are currently recorded during the summer months [44], and in the future, this temperature is likely to be reached more often given unmitigated climate change. More research is required to determine how M. franciscanus will be impacted as ocean warming continues, particularly as marine heatwaves increase in frequency [117,118].
During early development, M. franciscanus appear to be more susceptible to rising pCO 2 levels than to rising temperatures. The lack of a large transcriptional response, particularly at the prism stage, paired with a decrease in body size indicates that exposure to elevated pCO 2 is detrimental to developing embryos. Continued ocean acidi cation is therefore likely to have adverse impacts on future M. franciscanus populations, although this could be offset somewhat by the positive effects of simultaneous ocean warming [46,[119][120][121]. However, during seasonal upwelling events, M. franciscanus are exposed to corrosive pH conditions that lack the mitigating effects of warmer temperatures [15,16]. Spawning that occurs during or immediately prior to an upwelling event will subject developing embryos to high pCO 2 conditions paired with colder temperatures. 1050 matm pCO 2 levels have been measured at the urchin collection site during upwelling events [44], so populations of M. franciscanus are likely already impacted by elevated pCO 2 . Given ocean acidi cation and the increase in upwelling frequency and intensity that is predicted with continued climate change [122][123][124], the likelihood of M. franciscanus developing under stressful pCO 2 conditions should rise in the future.

Conclusions
The early developmental stages of M. franciscanus exhibited transcriptomic plasticity under different temperature and pCO 2 conditions. The extent of this plasticity, however, varied by developmental stage and by stressor. Although higher temperatures appeared to induce cellular stress, M. franciscanus exhibited a robust transcriptional response to temperature. The transcriptomic response to pCO 2 was much more limited, and may therefore increase the vulnerability of this species to ocean acidi cation. We caution against the overgeneralization of these ndings because they represent a small portion of the genetic variability that exists in M. franciscanus. Future studies that examine different populations and incorporate more genetic diversity will facilitate our understanding of how this species responds to its changing environment.
Accurately predicting how organisms will respond to future ocean conditions is necessary for the implementation of effective conservation and sheries management strategies. This study provides essential molecular-level insight towards understanding how an ecologically valuable shery species is affected by climate change. Although we characterized the transcriptomic response of M. franciscanus to ecologically relevant temperature and pCO 2 conditions, more work is required to determine how environmental stressors will impact the overall tness and adaptive potential of natural populations. Detrimental impacts that occur during early development can reduce the quantity or quality of juvenile recruits. Poor recruitment will likely not be immediately evident to the M. franciscanus shery until there is a marked decrease in the number of mature, harvestable adults. This delay in noticeable population declines may cause substantial alterations to coastal macroalgae ecosystems and nancial losses to the shing industry. Studies such as the one presented here are vital for developing proactive and adaptive management strategies to ensure climate-ready sheries [42,125].

Animal collection and culturing
Red sea urchins were collected and spawned as described in [46]. Brie y, adults were collected from  [20]. Eggs from ve individual females and sperm from a single male were collected. A subsample of eggs from each female was fertilized with sperm from the male and high fertilization success was examined for each cross (i.e., visually con rming the formation of fertilization envelopes). These subsamples were only used to verify suitable male-female compatibility and were discarded prior to the experiment. An approximately equal number of eggs from each of the ve females were gently pooled together. The pool of eggs was fertilized by slowly adding dilute, activated sperm from the male until approximately 98% fertilization success was reached. Performing crosses with a single male ensured that all cultures were composed of full-or half-sibling embryos. This approach was selected in an effort to limit paternal genetic variability and differences in male-female interactions that could otherwise impact the results of the study.
As described in [46], the newly-fertilized embryos were raised in three replicate culture vessels for each of four different combined temperature and pCO 2 treatments: 1) 17 °C and 1050 matm pCO 2 , 2) 17 °C and 475 matm pCO 2 , 3) 13 °C and 1050 matm pCO 2 , and 4) 13 °C and 475 matm pCO 2 . These treatments represent current conditions measured in the SBC as well as projected future ocean conditions given continuing warming and acidi cation. It should be noted that the SBC is within a highly dynamic region in which water conditions can vary greatly. For example, the temperature recorded where and when the M. franiscanus were collected for this study was 13.3 °C, but throughout the year prior, the temperature at the collection site uctuated between 10.0 -20.5 °C, with an average temperature of 15.1 °C [44]. Static treatments were selected, however, because the early development of M. franciscanus occurs rapidly (e.g., <2.5 days in this study) so unlike their adult counterparts, the embryos are likely to experience relatively stable conditions. The treatment of 13 °C and 475 matm pCO 2 represents conditions that are regularly measured in the SBC [44]. 13 °C is on the lower end of the range of temperatures observed in this region, but is regularly observed during seasonal upwelling events. During upwelling events, combinations of decreased temperatures and elevated pCO 2 levels are common [15,43,45,87]. These upwelling conditions are represented by the combined 13 °C and 1050 matm pCO 2 treatment. Water conditions in this area also currently reach elevated temperatures of 17 °C on occasion [44], but as ocean warming continues, these conditions are expected to increase in frequency [126]. The combined 17 °C and 1050 matm pCO 2 treatment represents future conditions as both ocean warming and ocean acidi cation progress. The temperature, pH, salinity, and total alkalinity were measured in each culture vessel daily as detailed in [46].

Sample generation and sequencing
The early gastrula stage was designated by the formation of secondary mesenchyme cells and the extension of the archenteron to approximately one-half the embryo's body length (~23.5 hours postfertilization (hpf) at 17 °C and ~32.5 hpf at 13 °C). The prism stage was designated by the tripartitioning of the archenteron, the formation of skeletal rods, and the pyramid-like body shape of the embryo (~44 hpf at 17 °C and ~55.5 hpf at 13 °C). Embryos at both developmental stages were sampled from each culture vessel by gently concentrating the embryos onto a submerged 35-mm mesh lter and transferring them into a 15-mL Falcon tube using a plastic transfer pipet. The concentration of embryos per mL of FSW was calculated by counting three aliquots of embryos so that a coe cient of variance (CV) of less than 10% was reached. At both the gastrula and prism stages, 5000 embryos from each culture vessel were transferred into a 1.5 mL microcentrifuge tube. Embryos were quickly pelleted by centrifugation, excess seawater was removed, and the samples were ash frozen in liquid nitrogen. The samples were stored at -80 °C until RNA extractions were performed.
Total RNA extractions were performed on samples from each culture vessel at both the gastrula and prism stages for a total of 24 RNA extractions. RNA extractions were performed by adding 500 mL of Trizol Ò reagent (Invitrogen) to each sample, and passing the entire contents through needles of decreasing sizes (i.e., three passes through a 21-gauge needle, a 23-gauge needle, and then a 25-gauge needle) until homogenized. The RNA was isolated using a chloroform addition. The RNA was precipitated in 100% isopropyl alcohol, washed in ice cold 80% ethanol, and resuspended in DEPC-treated water. RNA purity, quantity, and quality were veri ed using a NanoDrop â ND100, a Qubit â uorometer, and a TapeStation 2200 system (Agilent Technologies). The RNA samples were submitted to the Genome Center at the University of California, Davis where 24 libraries were generated using poly(A) enrichment.
The 24 libraries were pooled and sequenced across two lanes on an Illumina HiSeq 4000 with 100 basepair (bp) single reads.

Data processing
Adapter sequence contamination and base pairs with quality scores below 30 were removed from the raw sequence data using Trim Galore! (version 0.4.1) [127]. FastQC (version 0.11.5) [47] was used to verify sequence quality. The trimmed sequence data were mapped onto a developmental transcriptome for M. franciscanus [48] and expression values were calculated using RSEM (version 1.3.0) [49] and bowtie2 (version 2.3.2) [128]. The limma package [52] in R (version 3.4.4) was used to lter the sequence data to those that had at least 0.5 counts per million mapped reads across at least three of the samples. A scale normalization was applied to the read counts using a trimmed mean of M-values (TMM) normalization method [129]. The counts were converted to log-counts per million (logCPM) using edgeR's cpm function in limma.

PCA
Principal component analyses (PCAs) were performed on logCPM data to examine distances between the samples using the PCAtools package [50] in R. Because gene expression patterns in M. franciscanus vary signi cantly by developmental stage [48] and because comparing gene expression patterns across development was not a goal of the current study, both principal component and differential expression analyses were executed separately for each stage. Using the eigencorplot function, PCs were correlated to metadata variables, which included the experiment treatment factors (i.e., temperature and pCO 2 ).
Additional metadata variables were included using data from [46]. Speci cally, gastrula and prism body size were included using average length data (mm) for each sample. Thermotolerance, which was only measured for the prism stage, was included using the average LT 50 value (i.e., the temperature at which 50% mortality occured) of each treatment. The methods of how body size and thermotolerance values were obtained are described in detail in [46]. PCAtools was used to examine the loadings of the PCs with signi cant correlations to metadata variables. Permutational multivariate ANOVAs were performed using the adonis function in the package vegan [130] to establish the proportion of variance explained by xed factors (i.e., developmental stage, temperature treatment, and pCO 2 treatment).

DE Analysis
Differential expression (DE) analyses were performed using the limma-trend approach in limma by making pairwise comparisons between temperature treatments (17 °C versus 13 °C) and between pCO 2 treatments (1050 matm versus 475 matm) for each developmental stage. Moderated t-statistics and pvalues adjusted by the Benjamini and Hochberg's method were used to control the false discovery rate [131]. Differentially expressed genes (genes relatively down-or up-regulated) were determined for each pairwise comparison of interest (adjusted p-value < 0.05). Gene ontology (GO) analyses were performed following the GO_MWU package [51] in R. This package uses a rank-based Mann-Whitney U approach to determine whether GO categories are signi cantly enriched by up-or down-regulated genes. This study was conducted on a marine invertebrate that does not require Institutional Animal Care and Use Committee (IACU) protocols or approval. However, the specimens were collected in accordance with the requirements of a California Scienti c Collecting Permit granted to GEH (SC-1223) that has been noted in the Methods. In addition, no human research subjects were used for this study, and IRB approval was not required.

Consent for publication
Not applicable.

Availability of data and material
The datasets generated and/or analyzed during the current study are available in the NCBI Short Read Archive under Bioproject accession number PRJNA637102. Additional R data and analysis scripts are available at a Github repository (https://github.com/julietmwong27/Mfranciscanus_RNAseq_Ranalysis).

Competing interests
California NanoSystems Institute. This work also used computing resources supported by NSF under   Correlations at a the gastrula stage and b the prism stage between PC1-PC8 (columns), which contribute >80% of the explained variation in gene expression, and metadata variables (rows) of the experiment treatments (i.e., temperature and pCO2), body size (i.e. embryo length in mm), and thermotolerance (i.e. LT50 in C, prism stage only). The orange-purple color scale represents the strength of the Pearson's correlation (1 to -1). *p < 0.05, **p < 0.01, and ***p < 0.001 Figure 3 Temperature, and to a lesser degree, pCO2 treatments caused differential gene expression at a the gastrula stage and b the prism stage of early development. Genes that were not differentially expressed are displayed in gray while signi cant DE genes (adjusted p-value < 0.05) are displayed in color.