The β-oxidation pathway is downregulated during diapause termination in Calanus copepods

Calanus copepods are keystone species in marine ecosystems, mainly due to their high lipid content, which is a nutritious food source for e.g. juvenile fish. Accumulated lipids are catabolized to meet energy requirements during dormancy (diapause), which occurs during the last copepodite stage (C5). The current knowledge of lipid degradation pathways during diapause termination is limited. We characterized changes in lipid fullness and generated transcriptional profiles in C5s during termination of diapause and progression towards adulthood. Lipid fullness of C5s declined linearly during developmental progression, but more β-oxidation genes were upregulated in early C5s compared to late C5s and adults. We identified four possible master regulators of energy metabolism, which all were generally upregulated in early C5s, compared to late C5s and adults. We discovered that one of two enzymes in the carnitine shuttle is absent from the calanoid copepod lineage. Based on the geographical location of the sampling site, the field-samples were initially presumed to consist of C. finmarchicus. However, the identification of C. glacialis in some samples underlines the need for performing molecular analyses to reliably identify Calanus species. Our findings contributes to a better understanding of molecular events occurring during diapause and diapause termination in calanoid copepods.

Of the two enzymes in the carnitine shuttle, which transports fatty acids into mitochondria (carnitine O-palmitoyltransferase, CPT, EC 2.3.1.21), only one form (CPT1) was present in the C. finmarchicus reference transcriptome. Similar searches of a second C. finmarchicus transcriptome, two transcriptomes for C. glacialis and a Eurytemora affinis transcriptome (see methods for NCBI accession numbers), revealed only a single CPT homolog in these calanoid copepod species. In contrast, both CPT1-like and CPT2-like transcripts were identified through searches in the harpacticoid copepod Tigriopus japonicus. A maximum likelihood analysis confirmed that all copepod species contained CPT1 orthologs, but CPT2 was only present in T. japonicus (Fig. 1). The analysis correctly recovered the CPT1 and CPT2 clades with 100% bootstrap support. Within the CPT1 clade, the mammalian (Homo sapiens, Mus musculus) sequences fell into a moderately supported (74%) monophyletic clade, with multiple forms in each species due to diversification early in the deuterostome lineage. The arthropod sequences (C. finmarchicus, C. glacialis, E. affinis, T. japonicus, D. pulex, Tibolium castaneum, and Drosophila melanogaster) did not form a monophyletic group, and the calanoid copepod sequences were positioned on well-supported clade (97%) with relatively long branches.
Lipid fullness ratio. Mean lipid fullness ratio, measured as (lipid sac volume/prosome volume) * 100, was highest in the reference group (early diapause C5s), followed by the experimental C5s sampled on days 0, 5, 13 and 20 ( Table 2). The lipid fullness ratio was lower in adults than in C5s at all time points. ANCOVA results showed that the lipid fullness ratio declined with time (days since intitation of experiment) in the experimental C5s, but not in the adults (not present on day 0, P = 0.029, Fig. 2). The C5s (n = 185) showed a decline in lipid fullness ratio of 0.20% ± 0.09 (s.e.m.) per day, while the adults (n = 136) had a stable lipid fullness ratio over time (estimate ± s.e.m. = −0.01% ± 0.08 per day). transcriptomic temporal variation. A principal components analysis (PCA) of all genes in all samples (C5s and adults) showed a clear separation along principal component 1 (PC1, 22.18% variation explained) between the reference group (early diapause C5s, see methods) and adults on day 13 and 20 ( Fig. 3). In contrast, the experimental C5s scattered along PC1 between the reference group to the right and the adults to the left. These early and late C5s appear somewhat more differentiated from the reference group and had higher PC2 values (PC2: 11.56% variation explained).  Table S8 for accession numbers.
Differentially expressed genes (DEGs) between all compared sampling days were analyzed with generalized linear models (GLM, see methods). The number of DEGs (P > 0.05) ranged from 911 to 10580 (Fig. 4, number of all DEGs listed in Supplementary Table S3). The number of DEGs between the reference group and the adults sampled on days 13 and 20 was higher than 9000 (Fig. 4). There was also a large number of DEGs between the reference group and the experimental C5s (3000-6000 DEGs). Comparisons within the experimental C5s (days 0-20) had generally lower numbers of DEGs (900-2050 DEGs), while comparisons between experimental C5s and adults were around 6500 DEGs (Supplementary Table S3).
Gene ontology (Go) enrichment analysis. To identify processes that are potentially associated with diapause termination, a comparative GO enrichment analysis was performed based on DEGs between the reference group and all experimental C5s combined (days 0, 5, 13 and 20). Among upregulated genes in the reference group, 646 GO terms were significantly enriched (P < 0.05), while 497 GO terms were enriched among the upregulated genes in the experimental C5s. The GO terms with the most statistically significant enrichment (lowest P-values, i.e. least likely to be occur by chance. See Supplementary Fig. S1) in the reference group include 14 processes involved in the immune and/or defense system (e.g. GO:0006955: immune response, GO: 0006952: defense response and GO: 0038061: NIK/NF-kappaB signaling), as well as developmental processes of cells, neurons and synapses and endopeptidase activity. In the experimental C5s, the most statistically significant enriched GO terms ( Supplementary Fig. S2) include GO:0007283: spermatogenesis and GO:0007281: germ cell development, along with cuticle, chitin and epidermal developmental related processes. Other terms that were upregulated in the experimental C5s were mostly related to cell division and development in general.  www.nature.com/scientificreports www.nature.com/scientificreports/ Differential expression of genes of interest. The GLM results showed that several genes in the β-oxidation pathway were significantly (P < 0.05) downregulated in the experimental C5s compared to the reference group (Fig. 5A, comprehensive GLM metrics in Supplementary Table S4). Among these, more genes were downregulated in late C5s (days 13 and 20, seven genes, logFC range: −1.5 to −3.1, P < 0.05) than in early C5s (days 0 and 5, three genes, logFC range: −1.3 to −2.7, P < 0.05). One β-oxidation gene was upregulated on days 0, 5 and 13 (logFC: ~4.0, P < 0.05). Several genes were also downregulated in adults (days 13 and 20, logFC range: −1.5 to −4.9, P < 0.05) compared to the reference group (Fig. 5B). One β-oxidation gene was upregulated on day 13 (logFC: 1.57, P = 0.03). However, a custom GO term manually assigned to all identified genes in β-oxidation pathway (GO:5000000, see methods) was not enriched among the upregulated (P = 0.815) or downregulated (P = 0.373) genes.
Ferritin and hsp22, two proposed markers of Calanus diapause 30,31 , both exhibited significantly higher expression in the reference group compared to experimental C5s and adults at all time points (Fig. 6, and see    Supplementary Table S3. The start and end of each line indicates which comparison the line refers to. The thicker the line, the more DEGs between the compared groups, e.g. >9000 DEGs between the reference group and adults sampled on days 13 and 20. The thinner line, the fewer DEGs, e.g. <999 DEGs between experimental C5s sampled on days 5 and 13. In C5s, more β-oxidation genes were downregulated compared to the reference group on days 13 and 20 (seven genes) than on days 0 (three genes) and 5 (two genes). (B) Adults appeared in sufficient numbers for sampling on day 13. Six and five genes were downregulated in adults on days 13 and 20 compared to the reference group, respectively. Differences in expression is shown as log2 fold change in each group compared to the reference group. For example, a log2 fold change of -2 signifies a 4 fold decrease from the reference group. The bar at the top indicates day of sampling. The X-axis shows the gene ID (see Supplementary Table S4 for annotation and GLM metrics). . Hypothetical overall trend for expression of master regulators and lipid utilization during diapause, diapause termination and developmental progression in Calanus finmarchicus. The dashed line indicates the rate of lipid utilization. According to our observations, early Calanus spp. C5s have large stores of wax esters in their lipid sac during the intitation phase of diapause, and lipid catabolism (i.e. expression of β-oxidation genes) is low. Expression of ferritin, hsp22 and other genes related to increased stress tolerance is high, while expression of torso-like is low, and remains low in the beginning of the maintenance phase. Further into the maintenance phase, lipid catabolism increases. This coindices with high expression of torso-like and the master regulators of lipid metabolism SREBP, TAp63, HFN4 and NHR-E75. Closer to the termination of diapause, lipid catabolism declines and expression of these genes showed declining expression, reflecting lower energetic metabolism as the C5s are ready to molt into adults. When adulthood is reached under unfed conditions, both lipid catabolism and expression of the master regulator genes are low, and protein catabolism increases in order to maintain energy requirements.
www.nature.com/scientificreports www.nature.com/scientificreports/ of torso-like did not differ in C5s on day 0 compared to the reference group, nor among any of the time points in the experimental C5s (P > 0.05).
The "master regulators" of lipid metabolism SREBP, TAp63, HNF4 and NHR-E75 were generally expressed at higher levels in C5s than in adults (Fig. 6, also see Supplementary Table S5 for GLM metrics). For example, the GLM results suggest that NHR-E75 was higher in the reference group (logFC = 1.563, F = 15.03, P = 0.023), and in C5s on day 0 (logFC = 2.191, F = 28.12, P = 0.002), day 5 (logFC = 2.28, F = 27.46, P = 0.002), and day 13 (logFC = 1.819, F = 17.97, P = 0.013) compared to in adults on day 13. In order to assess whether the copepods needed to commence protein catabolism in addition to lipid catabolism 23 , we investigated gene expression of the core proteasome. Most of the core proteasome subunits were upregulated in adults compared to the reference group (Fig. 6, also see Supplementary Table S6). For example, proteosomal subunits β1-4 all exhibited elevated expressed levels in adults on day 13 (P > 0.05).

Discussion
The primary objective of the present study was to investigate energetic metabolism in diapausing Calanus spp. C5s during progression to active adults. This was performed by manually annotating genes in the β-oxidation pathway within the C. finmarchicus transcriptome, and using RNA sequencing to investigate changes in expression of these genes during diapause termination and subsequent molting. Additionally, we investigated the expression of potential master regulators of lipid metabolism and molecular markers of diapause. General transcriptomic variation during the progression was also assessed.
The species determination analysis that was performed after sampling, pooling and gene expression analyses revealed the presence of C. glacialis together with C. finmarchicus in the Trondheimsfjord, at the depth of 400 m below the surface. The presence of both species at 400 m depth suggests that they may spend diapause together in this fjord. As will be discussed below, the patterns observed in the present study concerning lipid fullness ratio, energy metabolism and expression of target genes, do not seem to differ due to the species factor. Even though it has been recognized that the two species are genetically distinct 35 , C. finmarchicus and C. glacialis have similar life history traits, e.g. they both have wax esters as their primary storage lipids, and they have similar overwintering strategies 36 . It is therefore likely that similar gene expression patterns can be expected from these species. Thus, we can make cautious conclusions from these observations. Nevertheless, our results underline the importance of molecular species determinations for field-collected copepods and the need for future comparative studies.
All genes in the β-oxidation pathway in the KEGG database identified in D. pulex were successfully identified in C. finmarchicus, with the exception of CPT2 and BCD. In other organisms, the CPT2 enzyme functions along with CPT1 within the mitochondrial membrane 21 . The main function is to transport fatty acyl-CoA into the mitochondria for β-oxidation. Within each of three examined species of calanoid copepods (C. finmarchicus, C glacialis and E. affinis), we detected only a single CPT gene that was strongly supported as a CPT1 ortholog. The absence of a gene in a single transcriptomic database may reflect limitations in sequencing and assembly, but the absence of CPT2 from three species is more consistent with gene loss, some time after the divergence of the calanoids and harpacticoids. The calanoid CPT1 sequences were divergent and fell on long branches, which may reflect an expansion of CPT1 function in these species in association with the lack of CPT2; however, this hypothesis requires functional testing. Alternatively, other mitochondrial transporter enzymes may perform the functions usually performed by the carnitine shuttle (transport of fatty acids into mitochondria) in calanoid copepods. The presence of other possible alternative transporters should be subject for investigation in future studies of lipid catabolism in calanoid copepods.
We were unable to identify a BCD (EC 1.3.8.1) homolog in C. finmarchicus. This enzyme catalyzes the oxidation of a short-chain acyl-CoA to a short-chain trans-2,3-dehydroacyl-CoA using flavoprotein as electron acceptor 37 , and is involved in one of the later steps within the β-oxidation pathway in other species. Within the C. finmarchicus transcriptome, the transcript sequence most similar to BCD was annotated as a short-chain acyl-CoA dehydrogenase (EC 1.3.99.12). These two enzymes (BCD and EC 1.3.99.12) share an overlapping set of substrates and are involved in the same steps in β-oxidation, as well as other metabolic pathways. It is unknown what role, if any, the C. finmarchicus enzyme (EC 1.3.99.12) plays in β-oxidation, and further investigation is needed.
Mean lipid fullness ratio ranged from 22% in early diapause C5s (the reference group) to 12% in adults at the end of the experiment. As expected, the lipid fullness ratio was highest in the reference group as these C5s were in an early phase of diapause compared to the experimental copepods. The lipid fullness ratio was generally lower than otherwise reported in the literature for Calanus spp. C5s and adults [38][39][40] . The lower lipid fullness ratio may be explained by variation in individual feeding success 7,41 , as well as species-specific, geographical and seasonal variability in environmental conditions.
Mean lipid fullness ratio decreased over time during progression through the C5 stage (Fig. 2), indicating catabolism of lipids in the lipid sac to fuel metabolic functions. This is expected as most crustaceans preferentially catabolize lipids during periods of fasting or starvation 23 , and is similar to previous observations of diapausing C. finmarchicus 42 . There was no change over time in lipid fullness ratio in adults. This suggests that the adults may be using other sources to fuel energy requirements, or that adults have an overall lower requirement for energy. The adults may also have reduced their metabolism due to starvation as previously reported in copepods 43 .
Transcriptional profiles varied according to the developmental progression of the copepods with time, as shown by the first axis of the PCA (PC1, Fig. 3). In this PCA, the C5s in the reference group clustered to the far right, and adults to the far left. The intermediate and somewhat variable positioning of the experimental copepodites between the reference group and the adults along PC1 may suggest some asynchrony in development and diapause phase. It should again be noted that some of the C5 experimental samples contained C. glacialis. However, these samples were all positioned between the reference group and the adults, and did not appear to www.nature.com/scientificreports www.nature.com/scientificreports/ differ from the C. finmarchicus C5 experimental samples along PC1 and PC2. Furthermore, the samples containing C. glacialis do not diverge from the other samples with respect to expression of lipid catabolism genes or master regulators, as will be discussed below.
Developmental progression of the copepods was also reflected in the number of DEGs between the copepods on the different days (Fig. 4). The adults were the most different from the reference group, and the experimental C5s were intermediary between the reference group and adults. Together, these results illustrate the developmental progression occurring from C5 copepodites in early diapause to active adult copepods, i.e. the more time between the different sampling points, the more different the copepods are, with respect to number of DEGs.
When compared to the reference group, the most significantly enriched GO terms in the experimental C5s were dominated by terms related to development of gametes, cell division and molting. These were expected findings as these processes are important in preparations for adulthood and reproduction, and indicates that the experimental C5s were in various stages of development within the maintenance and termination phases of diapause. Similar processes were enriched in N. flemingeri in the weeks after emergence from diapause in a recent study 18 . The reference group had enriched GO terms mostly related to immune responses, suggesting a high resistance towards stressful stimuli, which is recognized as a diapause characteristic 7,18,30 . Put into the context of developmental progression, the reference group which was collected directly on the research vessel in August seems, as expected, to be in an earlier diapause stage, i.e. early maintenance phase, than the experimental C5s which were collected in November. Large variation in development between C5 copepodites has been previously observed in C. finmarchicus in a study by Tarrant et al. 17 . It is likely that some asynchrony in development and diapause termination must also occur in nature, to spread predation risk and to prolong phytoplankton feeding opportunities for offspring.
In this study, we treated the reference group as an initial developmental point ("time zero") that could be compared with all the experimental copepods. Collection during two different seasons with potentially differing environmental conditions could be problematic. However, conditions such as temperature, salinity and pressure at 400 m depth remains very similar throughout the year at the sampling location (CTD-data, long term survey, Trondhjem Biological Station, database under contruction). Diapause itself can lead to a relatively homogenous metabolic state of the Calanus population at 400 m, with minimal access to food and predators for an extended time period. Thus, we argue that the results are likely not severely affected by sampling during different seasons, though it is desirable to repeat the experiment and make comparisons within a single season.
Although there was a linear decrease in lipid fullness ratio of C5s, the expression pattern of genes in the β-oxidation pathway (Fig. 5) suggests a non-linear utilization of lipids. Late C5s (days 13 and 20) had a higher number of downregulated β-oxidation genes compared to the reference group than early C5s (days 0 and 5). This suggests reduced lipid catabolism towards the end of diapause, i.e. close to diapause termination and molting (see Fig. 6 for hypothesized lipid utilization). During starvation, most crustaceans preferentially catabolize neutral lipids while conserving polar lipids, which have an essential role in maintaining cell membrane structure and integrity 23 . When the neutral lipid storage becomes depleted, proteins are increasingly utilized as an energy source 23,44 . Our results, which suggests a reduction in lipid catabolism towards diapause termination, could be explained by an adjustment during diapause from mainly lipid catabolism to a higher degree of protein degradation in late C5s. Similar distinct phases of biomass degradation has previously been reported in crusteceans 45 . Downregulation of β-oxidation genes was also observed in the adults (Fig. 5), and the adults sampled early in the experiment had similar mean lipid fullness ratio as those sampled late (Fig. 2). This suggests that the adjustment from primarily lipid catabolism toward increasing protein catabolism continued after the copepods reached adulthood. Indeed, several genes encoding the subunits of the proteasome were upregulated in the adults compared to the reference group, indicating increased protein catabolism. The lack of upregulation of these proteasome genes in the late C5s may indicate that low lipid stores could be an important endogenous cue for the initiation of moulting and development into adults, as has been previously suggested 46 .
Ferritin has been suggested to have a role in protection against oxidative stress and in delaying development in C. finmarchicus, and its expression has been hypothesized to be highest during the early stages of diapause 31 . Indeed, the expression of ferritin was highest in the reference group, and lower in C5s on days 0, 5, 13 and 20, and continued to decrease in adults (see Supplementary Fig. S3). As previously suggested, this implies that the reference group was in earlier stages of diapause than the experimental C5s, i.e. early maintenance phase. Similarly, hsp22 has been shown to be elevated in C. finmarchicus in diapause 30 , and suggested to be involved in stress tolerance. In the present study, this gene had a similar expression pattern as ferritin, and showed a trend of reduction during termination of diapause. Torso-like protein has been suggested as a good discriminator between early and late C5s regarding development, with increasing expression with progression throughout the C5 stage 17 . This protein helps to regulate both developmental timing and body size 17,47 . Torso-like expression was low in the reference group and in C5s on day 0, and higher in later C5s, before dropping in the adults. This implies that after day 0 the experimental C5s were further along in development.
Expression of the master regulators was generally upregulated in early C5s (the reference group, and days 0 and 5) and downregulated in adults and/or late C5s (days 13 and 20). SREBP is a membrane-bound transcription factor that can stimulate lipid biosynthesis, as shown in insects, where it has an essential role in membrane production 26 . TAp63 is closely linked to lipid metabolism in mice 25 . HNF4 belongs to a subfamily of nuclear receptors (NR2A), which is represented by a single gene in Drosophila and has further duplicated within vertebrate and nematode lineages. In Drosophila, HNF4 regulates genes needed for β-oxidation and lipid mobilization 28 . NHR-E75 is a member of the same nuclear receptor family as the ecdysone receptor (EcR; NR1D and NR1H, respectively), E75 has been identified in many insect species 48 . E75 is involved in molting and developmental progression in arthropods 27,48 , and could therefore be expected to be upregulated in the late C5 which were close to the terminal molt. It has also been suggested that in D. melanogaster E75 may be a functional equivalent of the vertebrate PPARγ, which is a master regulator of lipid metabolism 49  www.nature.com/scientificreports www.nature.com/scientificreports/ linked to the lipid metabolism in copepods, and that the downregulation in the late C5s is related to the downregulation of β-oxidation genes in these copepods observed in the present study.
Together, the expression profiles of the molecular markers of diapause/development and the master regulators suggest that the reference group was comparatively early in the maintenance phase, while the other early C5s (i.e. days 0 and 5) were in the second half of the maintenance phase, where development is gradually initiated 15 and fueled by catabolism of wax esters in the lipid sac. Late C5s were closer to adulthood, which was reflected by the expression of master regulators of lipid metabolism that did not differ compared to the adults (see Supplementary  Table S5). Hypothesized trends based on the observed gene expression and lipid catabolism profiles are illustrated in Fig. 6. Following our observations, early Calanus spp. C5s have large lipid sacs during the initiation phase of diapause, as well as very low lipid catabolism and high expression of ferritin, hsp22 and other genes related to increased enzyme and nutrient conservation and increased stress tolerance. Torso-like expression is low in early diapause (the initiation phase) and in the beginning of the maintenance phase (reference group and C5s from day 0 in the present study). Later in the maintenance phase, the C5s catabolize wax esters from their lipid sacs. This catabolism declines with developmental progression during diapause. High lipid catabolism is expected to occur during energetically-demanding stages of development, one of which, according to our observations, seems to transpire relatively early during the maintenance phase (early C5s), and not during preparation for molting during the termination phase. Elevated expression of torso-like, SREBP, TAp63, HFN4 and NHR-E75 was observed in the early C5s, and can thus be interpreted to coincide with increased energetic metabolism and intiation of developmental processes. Closer to the termination phase (late C5s), these genes showed declining expression, reflecting lower energetic metabolism and lipid catabolism as the C5s are ready to molt into adults. When adulthood is reached under unfed conditions, both lipid catabolism and expression of the master regulator genes are low, and protein catabolism is increased in order to maintain energy requirements.
Future research regarding energetic metabolism during diapause in C. finmarchicus should include attempts to induce diapause in the laboratory, to circumvent the unavoidable disturbance from collection of copepods in diapause and to avoid the presence of other Calanus species. Despite the apparent similarities in genetic variation during diapause termination of C. finmarchicus and C. glacialis, targeted comparisons of diapause physiology between the two species should be performed. The improved annotation of the genes in the β-oxidation pathway provides a foundation for further exploration of basic physiology in these keystone species in northern marine ecosystems. Our findings also provide a basis for a better understanding of the development occurring during diapause and diapause termination, and can in this way contribute to understand how large-scale changes, such as climate change, can influence copepod population dynamics and whole marine ecosystems. Methods copepod collection. Calanus spp. copepodites of stage C5 were collected from sea bed depth (400 m) up to 200 m in the Trondheimsfjord, Norway (N63°29′, E10°18) using a Nansen net (mesh size 200 μm) with a closing mechanism and a non-filtering cod end 50 . In this fjord, C. finmarchicus C5s migrate to deep waters to diapause in May, and resurface to reproduce in March 51 . Collection of copepodites in early stages of diapause was performed directly on the research vessel in August 2017, and will henceforth be referred to as "the reference group". Collection of C5s ("the experimental C5s") later in the diapause phase was performed in November 2016 from the same vessel. During sampling of both groups, the net was hauled at a speed of approximately 0.3 m/s to reduce physical stress to the copepods. On deck, the cod end was quickly removed from the net after retrival, and the content was screened on a 1300 μm mesh sieve submerged in a bucket of seawater. In this process, chaetognaths, significant copepod predators 52 , were eliminated from the samples. The copepods were temporarily kept in filtered seawater (salinity: ~33‰, temperature range 4-9 °C) in 30 L buckets that were covered in three layers of dark plastic, until they were either sampled for RNA seq (the reference group) on the research vessel, or transported (the experimental group) to the laboratory facilities on shore (NTNU Sealab, Trondheim, Norway). During sampling, the copepods were euthanized with tricaine methane-sulfonate (Finquel, 1.5 g/L seawater, Argent Laboratories Redmond, WA, USA) and photographed using CCD still-video camera (Sony DWF-sx900, Sony Corporation, Tokyo, Japan) operated by Fire-i software (Unibrain, Inc., San Ramon, CA, USA) connected to a dissecting microscope (Leica Microsystems, Germany). The frames were stored on a PC for further processing. For sampling of the reference group, 10 individuals per sample (n = 2) were placed together in 1 mL RNA later (ThermoFisher, USA) in 1.5 mL microcentrifuge tubes, which were refrigerated at 4 °C for approximately 24 hours before they were moved to −20 °C.
After transportation to the laboratory facilities on shore, the experimental C5s were acclimated in two 250 L polystyrene tanks with constant flow-through of filtered natural seawater (170 mL/min, 8 °C, no light exposure) for 13 days before the experiment was initiated. This was to ensure acclimatization to laboratory conditions, and to minimize continuous stimulation prior to the onset of experiment. The copepods were not fed for the duration of the experiment. experimental setup. To initiate the experiment, the copepods were temporarily sorted from the 250 L tanks using ladles and a 64 μm sieve into plastic cups, and then into the experimental 5 L borosilicate jars (n = 4) with open tops and modified spouts, which ensures an even outflow of water matching the continuous inflow of ~12 mL/min of filtered natural seawater. All handling was performed at 8 °C, using a halogen flashlight with a red glass filter for illumination. The room was kept in the dark and at 8 °C, and the experimental containers were kept behind light-blocking curtains throughout the experiment.
Each sampling day, three C5s per replicate were sampled for RNA seq in the same way as described above for the reference group. The first sampling of C5s in the experiment was performed on the same day as the sorting (day 0, 13 days after collection, Table 1). The second was performed on day 5, the third and fourth sampling were www.nature.com/scientificreports www.nature.com/scientificreports/ performed on days 13 and 20. On days 13 and 20, adults appeared in sufficient number for sampling, so three adults per replicate (females only) were sampled in addition on these days.

Species determination.
Recently, the occurrence of the "true Arctic shelf species" C. glacialis 3 was documented in the Trondheimsfjord through molecular analyses 32,33 . This information became available after the sampling and sequencing for the present study had already been performed. Morphological criteria have recently been deemed insufficient to reliably distinguish between C. glacialis and C. finmarchicus 32 . Thus, molecular-based species identification was performed to identify the potential presence of Calanus species other than C. finmarchicus in aliquots of the RNA samples, which consisted of RNA from pooled individuals. Archived RNA aliquots from all samples sequenced were used to investigate the potential presence of C. glacialis among the samples.
First, RNA extracts of pooled individuals were converted into first-stranded cDNA using the Invitrogen ™ SuperScript ™ II Reverse Transcriptase (Life Technologies) following the manufacturer's protocol. This cDNA was then used as template for molecular species identification following the procedure previously described 33 , based on 5 InDel-type markers 53 (only the transcriptome-based markers were used).
Biometry. The length, height and area of each copepod's prosome and lipid sac were estimated from the photographs using ImageJ software (version 1.48k) with a drawing tablet (Wacom, Co. Ltd. Japan). Developmental stage (C5 and adult) was also determined from the photographs. Lipid sac volume and prosome volume were estimated using methods similar to Miller et al. 41 . Lipid fullness ratio (%) was calculated as the ratio of (lipid sac volume/prosome volume) * 100.
Statistical analyses of biometrical data were performed in R (Version 3.3.1). Outliers were removed using Tukey's method where data points above or below the 1.5 x interquantile range are removed. Two outliers were removed, leaving n = 319. The effects of time (as days since start of experiment) and stage (C5, n = 185 and adults n = 136) on lipid fullness ratio were investigated using an ANCOVA with time and stage as explanatory variables. Visual inspection of the model residuals confirmed their normal distribution. The reference group was excluded from the ANCOVA as this group was sampled directly after collection in the field, and thus not subjected to acclimation like the experimental copepods were, leading to a much higher range of lipid fullness ratios.
RNA isolation, library preparation and RNA sequencing. Samples were selected for RNA sequencing based on the median lipid fullness ratio of all samples were chosen for RNA sequencing. This selection consisted of two or three replicates from each day that C5s were sampled (reference and days 0, 5, 13 and 20), and two replicates of adults from days 13 and 20. Prior to RNA extraction, the copepods from each sample were removed from RNAlater, gently dried on a paper towel and weighed. Total RNA extraction was performed using the Qiagen RNeasy Plus Universal Mini Kit (Qiagen Inc., Valencia, CA, USA) with additional use of a QiaShredder column, following the manufacturer's protocol. The RNeasy kit includes a genomic DNA removal step. Final eluation volume of RNA was 30 μL. RNA quality was assessed using a Model 2100 Bioanalyzer instrument (Agilent, Santa Clara, USA). Meaningful RNA Integrity (RIN) values could not be obtained because the hydrogen bonds in arthropod 28S rRNA are disrupted in the denaturation heating step prior to Bioanalyzer analysis, resulting in fragments that migrate closely with 18S rRNA 54 . RNA quality was therefore assessed using electrophoregram summary graphs and gel images. All analyzed samples were of high quality, containing a strong 18S band and little or no evidence of genomic DNA contamination (large bands) or degradation (smear of smaller bands).
For RNA sequencing, cDNA libraries were synthesized from total RNA (40 ng/μL input) using the Illumina TruSeq Stranded mRNA sample preparation kit (Illumina, San Diego, USA), with minor modifications adjusted for smaller volumes than in the manufacturer´s protocol. Final volumes of cDNA libraries were 22 μL. Prior to RNA sequencing, the cDNA libraries were pooled and normalized, and a quality control was performed on a Bioanalyzer instrument by the sequencing facility. Pooling was conducted in order to incorporate biological material from a larger number of individuals into a limited number of RNA-seq libraries 55 , as is common in gene expression studies with copepods (reviewed by Tarrant et al. 56 ).

Illumina sequencing and bioinformatic analyses.
Samples were sequenced at the Genomic Core Facility (GCF) at NTNU, Trondheim, with 75 bp paired-end reads on an Illumina HiSeq 500 HO flowcell. This produced ~130 million reads in total, and ~9 million reads per sample (see Supplementary Table S7). Each sample was sequenced on four lanes on the Illumina flow cell. Adapter sequences were trimmed at the GCF. Quality control was performed with MultiQC (version 1.2.dev0).
The present study focuses primarily on C. finmarchicus, as this species has been considered to be the dominant zooplankton species in the Norwegian Sea 57 . Thus, a C. finmarchicus reference transcriptome (PRNJA231164) was used as the main genomic database for the bioinformatic analyses. The reference transcriptome was re-annotated in its entirety 58 62 . Read counts were normalized to trimmed mean of M-values (TMM) to account for difference in library sizes, and FPKM-normalized (fragments per feature kilobase per million reads mapped) when producing PCA plots, using scripts within the Trinity pipeline. Differentially expressed genes (DEGs) between stages (adult vs C5) on days 0, 5, 13 and 20 compared to the reference group were analysed with generalized linear models (GLM) with a negative binomial distribution using packages EdgeR (version 3.6.8) and limma (version 3.20.9) within Bioconductor in R Version 3.2.3 63 . Alpha level was set to 0.05. Expected counts from RSEM were used as input in www.nature.com/scientificreports www.nature.com/scientificreports/ the GLM. Genes with very low counts per million (CPM ≤ 1) were filtered out. Tagwise dispersion was calculated using the Cox-Reid profile-adjusted likelihood method, which allows for multiple factors in the GLM 64 . EdgeR uses the Benjamini-Hochberg method for adjusting p-values for multiple comparisons. In a separate analysis, the reference group was compared to all C5s of the experiment combined, using the exact test in EdgeR, where dispersion is calculated using the quantile-adjusted conditional maximum likelihood method. This analysis was the basis for the gene ontology (GO) enrichment analysis, which was performed on up-and downregulated DEGs separately, using the goseq (version 1.26.0) 65  To further improve the annotation of the C. finmarchicus transcriptome, we manually annotated genes in the β-oxidation pathway, using as a reference the amino acid sequences of enzymes in the β-oxidation pathway from D. pulex, the only crustacean in the KEGG database as of December 2017 (dpx00071). It was also the species with most top hit blastx searches for C. finmarchicus in the study by Lenz et al. 16 . A nucleotide database was made from the reference transcriptome, and each amino acid sequence was used in a tblastx (BLAST+, version 2.7.1) query (translated nucleotide sequence toward amino acid sequence). E-value cutoff was set at 10 −7 . Following the manual annotation of the genes in the β-oxidation pathway, we observed that the automatic blast-based annotation had not consistently assigned these genes to a common GO term or pathway (i.e., related to β-oxidation). Similar limitations of automated blast-based annotation of the C. finmarchicus transcriptome have previously been observed for the glycolysis pathway 67 . To assess GO enrichment of the β-oxidation pathway specifically, we manually assigned all the genes with a custom term, "GO:5000000, " and repeated the enrichment analysis on up-and downregulated genes separately, comparing the experimental C5s combined against the reference group.
The same method as described above was used (with tblastn) when identifying master regulators of lipid metabolism and protein degradation genes (proteasome). The amino acid sequence for SREBP was obtained from Daphnia magna (SI , Table S8). Two partial sequences encoding TAp63 were obtained from Mytilus galloprovincialis, and both sequences had the same top hit in C. finmarchicus. The amino acid sequence for HNF4 was obtained from T. japonicus. Due to the importance of the peroxisome proliferator activator receptor (PPAR) as a master regulator of energy and lipid homeostasis in many other species 68 , we investigated the presence of this receptor in C. finmarchicus. The top hit had a 35% identity with a sequence that was annotated as NHR-E75. Amino acid sequences encoding all subunits of the core proteasome were obtained from D. melanogaster.
A phylogenetic analysis was conducted to investigate the apparent loss of a CPT gene from C. finmarchicus. CPT1 and CPT2 amino acid sequences were retrieved from the KEGG database (EC: 2.3.1.21) for D. pulex, D. melanogaster, T. castaneum, H. sapiens, and M. musculus (accession numbers for amino acid sequences in Supplementary Table S9a). We used the D. pulex sequences to search transcriptomic databases for the copepod species C. glacialis (PRJNA237014, PRJNA274584), E. affinis (PRJNA242763), and T. japonicas (PRJNA274317) using tblastn with default parameters (nucleotide sequence accession numbers in Supplementary Table S9b). We also searched a second transcriptomic database that had been independently assembled for C. finmarchicus (PRJNA236528). The longest open reading frame was identified within each copepod transcript using a biopython-based script 69,70 (reported in Supplementary Table S10). All sequences were aligned using MUSCLE with default parameters 71 . A maximum likelihood analysis was conducted using RAXML-HPC version 8 with rapid bootstrapping options 72 , as implemented on the XSEDE platform through the CIPRES Science Gateway 73 . Trees were visualized and edited (font size, position of apparent root) using FigTree version 1.1.2 and Adobe Illustrator (version CS4, repositioning of labels for visibility, labelling of clades).

Data availability
Sequence data have been submitted to the National Center of Biotechnology Information (NCBI; www.ncbi.nlm. nih.gov) under the Bioproject PRJNA420690.