Background & Summary

The copepod Calanus finmarchicus (Crustacea, Copepoda) is a key zooplankton species in the northern Atlantic food web as it converts sugars from algae into energy rich lipids that sustain higher consumers including marine fish larvae and seabirds1,2,3. Its high abundance and biomass also makes it an important contributor to ocean carbon flux4. The species inhabits a large latitudinal range from ~40° up to 80° N5. However, recent findings show that C. finmarchicus is undergoing temperature driven geographical shifts northwards because of climate change6,7,8, the effects of which are at their most extreme in the Northern Atlantic and Barents Sea. Therefore, the copepods will experience a change between the photoperiods they are adapted to at lower latitudes and the extreme high-latitude photoperiods. Photoperiodic variation is particularly pronounced in the Arctic with rapid change over short latitudinal ranges. The impact of such extreme photoperiods on non-endemic species is unknown, and the northward expansion of organisms at high latitudes may be limited by the adaptive capacity of their endogenous timing systems to extreme photoperiods8,9.

Endogenous timing systems, or biological clocks, are ubiquitous ancient and highly adaptive mechanisms enabling organisms to track and anticipate environmental cycles and prepare biological processes accordingly10,11. Since the identification of circadian clock genes in C. finmarchicus12, studies have shown that this species possesses a functional clock that might be involved in the timing of both diel and seasonal events, such as the ecologically and biogeochemically important diel vertical migration (DVM)13 or diapause14. However, the Arctic environment is characterized by dramatic seasonality resulting in permanent illumination during Midnight Sun and permanent darkness during Polar Night15. As the circadian clock is entrained and synchronized by daily light/dark cycles, the persistence of daily biological processes in Arctic organisms during the absence of those remains uncertain16,17, as well as the consequences for newcomer species due to global warming8,9. Moreover, the Arctic is characterized by strong fluctuations in sea ice-cover, reflecting on biotic and abiotic factors, such as species communities and interactions or light penetration18,19.

Copepods are among the important non-model invertebrates for which genomic resources are still limited, one barrier being that many species, including C. finmarchicus, have large genomes20,21. The de novo transcriptome of C. finmarchicus22 represents a useful resource for assessing the impact of global warming in this species of high ecological interest. In addition to differential gene expression analyzes, RNA sequencing has increased the ability to study the expression of rhythmically expressed mRNAs23,24,25. Indeed, at the molecular level, the endogenous clock machinery drives the rhythmic expression of downstream genes whose rhythmic translation and function ultimately underlie daily oscillations at cellular and organismal levels25. Note that in the field, environmental cycles also directly generate rhythms independently from the clock. Thus, temporal transcriptomic studies allow a major breakthrough in the understanding of daily dynamics of biological processes in the field.

In this study, we performed RNA sequencing on temporally collected in situ samples to generate a daily transcriptome of C. finmarchicus in the high Arctic during summer solstice period when the sun remains high above the horizon with minimal altitude variation. Sampling of C. finmarchicus stage V copepodites was performed at 4 h intervals within a 24 h cycle at two ocean stations along a latitudinal gradient. The northern station (82.5 °N, Nansen Basin) was characterized by sea ice-coverage, while the southern one (74.5 °N, Barents Sea) was sea ice-free. In addition to providing the raw data, we describe its quality assessment and the alignment to the reference transcriptome to verify reliability and determine transcript quantification. Finally a complete annotation is performed and two normalized datasets are provided for further transcriptomic data exploration of this species.

Methods

Sampling design

The sampling strategy was specifically designed for the detection of rhythmic transcripts25,26 although it does not exclude classic differential expression analysis27. Sampling design and analysis strategy are presented in Fig. 1, Table 1 and Supplementary Table 1. Sampling covered a complete 24 h cycle at 4 h intervals, resulting in seven time points per station. At each station, samplings were performed at similar time intervals of: 14–15 h, 18–19 h, 22–23 h, 2–3 h, 6–7 h, and 10–11 h (all times noted in local time (UTC + 2)). Sampling at “North” station, JR85, started on 18th June (3 days before the summer solstice) at 10–11 h and ended on 19th June at 10–11 h. Sampling at “South” station, B13, started on 30th June (9 days after the summer solstice) at 14–15 h and ended on 1st July at 14–15 h. At each timepoint the water column was sampled from 200 m to the surface with vertical hauls of a WP2 plankton net (opening ∅: 57 cm, net length: 236 cm, mesh size: 200 µm) with a meshed bucket cod end (mesh size: 200 µm) at a speed of 0.5 m*s−1. Transferring the animals from the net into the stabilization solution was done within less than 12 minutes for all samplings. A 12 h period of incubation at 2–4 °C was allowed to soak the samples thoroughly with the RNAlater stabilization solution (Ambion, UK) before they were transferred to −80 °C for further transport and storage.

Fig. 1
figure 1

Overview of the experimental workflow used to generate the transcriptomic data output (all times noted in local time (UTC + 2)). Sample details are available in Supplementary Table 1.

Full size image
Table 1 Summary of sampling and sequencing strategy. Details are available in Supplementary Table 1.
Full size table

Sites description

Sampling has been conducted during Cruise JR17006 of the RRS James Clark Ross in summer 2018 at two stations along a latitudinal gradient. The station “North” was sea ice-covered and located in the Nansen Basin (JR85; 82.56°N, 30.85°E). The station “South” was sea ice-free and located in the southern Barents Sea (B13; 74.5°N, 30°E). Water depth at “North” was 3700 m and at “South” was 360 m. The sun’s altitude was always above the horizon but still showed diel oscillations of altitude above the horizon from 16 ° at midnight to 30.9 ° at midday at “North”, and from 7.7 ° at midnight to 38.6 ° at midday at “South” at the times of sampling (all times noted in local time (UTC + 2)). Sites were exposed to semidiurnal tide regimes, i.e., 2 tides per day, with a maximum amplitude of ± 0.47 m at JR85 and ± 0.36 m at B13 at times of sampling. Maps with the location of the sea ice edge at the time of sampling at “North” are available from the meereisportal28 (https://data.meereisportal.de/gallery/index_new.php?active-tab1=method&ice-type=satellite&satellite=A&region=n&resolution=daily&minYear=2018&minMonth=6&minDay=18&maxYear=2018&maxMonth=6&maxDay=19&showMaps=y&dateRepeat=n&submit2=display&lang=en_US&active-tab2=satellite). Modeled data of sun altitude were obtained from the United States Naval Observatory (https://aa.usno.navy.mil/data/docs/AltAz.php, USNO, USA). Information on the tidal dynamics have been drawn from the TPX08 model29 by using the OTPS package (Tidal Prediction Software, http://www-po.coas.oregonstate.edu/~poa/www-po/research/po/research/tide/index.html), via the mbotps program30 (MB-System). Solar altitude, tidal height and sea-ice cover during the sampling campaign at both latitudes are detailed in Supplementary Table 2. Temperature, pressure (depth), conductivity (salinity), oxygen saturation (SBE 43, Sea-Bird Electronics) and Chlorophyll a (Chl a) fluorescence (Aquatracka III fluorometer, Chelsea Technologies Group, UK) were measured from the surface to 200 m depth and are available in Hueppe et al.31.

Copepod sorting

Copepods were sorted at 2 °C under a stereo microscope for species (C. finmarchicus) and stage (CV). To distinguish C. finmarchicus from its closely related congener C. glacialis, morphological indicators were used, in particular the redness of the antenna, which has been shown to be a good indicator in the regions of sampling32; see also the molecular validation of morphological identification, below. For each timepoint and station, 3 replicates of 15 C. finmachicus CV were sorted. The choice to pool 15 individuals was made to (1) get the sufficient amount of RNA required for RNA sequencing and quantitative real-time PCR analyses and (2) increase the number of individuals (315 copepods per station in total) thereby decreasing the effect of individual variability.

RNA extraction

Each replicate was distributed to a 2 ml Precellys® homogenization tube (Bertin Instruments, France), containing a mix of 1.4 mm and 2.8 mm ceramic beads and homogenized in 600 µl of TRIzol® reagent (ThermoFisher Scientific, USA) with a Precellys® 24 Tissue Homogenizer (Bertin Instruments, France), using two times 15 sec. of homogenization at 5000 rpm with a 10 sec. break between. For RNA extraction, a Phenol/Chloroform based single-step extraction in combination with a spin column based solid phase extraction (Direct-zol™ RNA MiniPrep Kit, Zymo Research, USA) was used. Genomic DNA was removed by DNase I digestion on column as part of the RNA extraction kit and total RNA was eluted in ultra-pure water. A portion of the RNA of each of the samples was used to investigate relative expression of 8 candidate genes with SYBRGreen based quantitative real-time PCR (qPCR) on candidate genes, using the 2−∆Ct method32 and the geometric mean of elongation factor 1α and 16 s rRNA as reference, as described by Hueppe et al.31. Another portion of each samples was send to GeT-PlaGe core facility in dried-ice for RNA sequencing.

RNA sequencing

RNA sequencing was performed at the GeT-PlaGe core facility, INRAE Toulouse. The 42 RNA sequencing libraries were prepared according to Illumina’s protocols using the Illumina TruSeq Stranded mRNA sample prep kit to analyse mRNA. Briefly, mRNA were selected using poly-T beads. Then, RNA were fragmented to generate double stranded cDNA and adaptors were ligated to be sequenced. 11 cycles of PCR were applied to amplify libraries. Library quality was assessed using a Fragment Analyser (Advanced Analytical Technologies, Inc., Iowa, USA) and libraries were quantified by qPCR using the Kapa Library Quantification Kit (Roche). RNA sequencing experiments have been performed on a NovaSeq S4 lane (Illumina, California, USA) using a paired-end read length of 2 × 150 pb with the Illumina NovaSeq Reagent Kits.

Reads alignment and quantification

42 RNA sequencing libraries were obtained (Fig. 1, Table 1, and Supplementary Table 1). The number of paired reads per library was between 74 million and 276 million with an average of 126 ± 5 million (mean ± SE) reads. The RNA sequencing libraries reads quality were evaluated using FastQC33. Contamination was checked by aligning reads against E. coli, Yeast and PhiX genomes.

The Calanus finmarchicus de novo transcriptome22, based on different life stages and deposited to Bioproject PRJNA236528, was used as the reference transcriptome. It is composed of 206,012 contigs and presents good results of quality assessment, with a nearly complete BUSCO set22,34. Reads were aligned to the de novo transcriptome with BWA-MEM (http://bio-bwa.sourceforge.net/bwa.shtml). Quantification was performed with SAMtools35 idxStats to generate the quantification matrix. The matrix was filtered with edgeR36 and only contigs with more than 1 CPM (Count Per Million) in at least one sample were kept, providing a matrix of 76,550 contigs. Information on the datasets resulting from this study is available in Table 2.

Table 2 List of available datasets related to the study (NCBI Bioproject PRJNA62888640 and figshare collection 512770441).
Full size table

Annotation

We provided different annotations for all further analysis. Contigs were aligned with DIAMOND37 on NR (2019-09-29), Swissprot and Trembl (2018-12) to retrieve corresponding best annotations. An annotation matrix was then generated by selecting the best hit for each database if: i) the percent of the query length covered by the alignment was higher than 60%; ii) the percent of the subject length covered by the alignment was higher than 40%; iii) the percent of identity of the alignment was higher than 40%. Contigs were also processed with InterProScan38 to scan InterProScan signatures. A GO was assigned to each contig with an InterProScan hit containing a GO annotation. Information on the datasets resulting from this study is available in Table 2. Note that a previous annotation of Calanus finmarchicus reference transcriptome22 against Non-redundant (NR) protein database is also available at https://doi.org/10.5061/dryad.11978.

Normalization

Two normalizations are proposed (down-sampling normalization and RLE normalization) but the choice of normalization depends on the analysis required downstream. For a rhythmic analysis, we suggest down-sampling the mapped reads to the lowest number among the 42 samples (down-sampling normalization), i.e. to 70.4 million properly mapped reads per sample for all samples (after filtering), in order to adjust for differences in sequencing depth among samples23,25,39. This was performed with StreamSampler.jar (https://github.com/shenkers/sampling). EdgeR36 was used to perform RLE normalization, since it is more appropriate for differential expression analysis. Information on the datasets resulting from this study is available in Table 2.

Data Records

Raw reads were gathered in the NCBI BioProject PRJNA62888640 which includes all BioSamples used for the study (Table 2, Supplementary table 1). We also provide the following in figshare collection 512770441 (Table 2): the quantification matrix for the 206,012 contigs; the list of identifiers corresponding to the 76,550 contigs after filtering; the two suggested normalization matrices (down-sampling and RLE) and; the datasets annotations (DIAMOND annotation matrix, InterProScan annotation, GO association).

Technical Validation

Molecular validation of morphological identification

Since C. finmarchicus’ Arctic congener C. glacialis also occurs in the region of sampling and differences between the species can be very subtle42, morphological identification was validated by molecular species identification on a subset of samples from the same stations21,43. DNA was extracted from individual copepods using the HotShot method44, and the species-specific nuclear insertion/deletion (InDel) marker G-150 was amplified using a modified protocol from Smolina et al.45. Identification was done by accessing the size of the resulting amplicon via electrophoresis on a 2% agarose gel. Results have shown that 99% of the individuals identified as C. finmarchicus by the morphological identification method were also clearly identified as C. finmarchicus by the molecular identification method, while 0.1% were not clearly identified and 0.7% were identified as the Arctic congener Calanus glacialis (n = 305 individuals).

Extraction and RNA integrity

RNA extraction procedures were performed with randomization of samples to ensure reliable and unbiased data production. RNA purity was assessed by OD measurements with a NanoDrop 8000 spectrophotometer (ThermoFisher Scientific, USA), and all 260/280 and 260/230 OD ratio was superior to 1.9. RNA integrity was evaluated with a Fragment Analyzer (Advanced Analytical Technologies, Inc., Iowa, USA; RNA Kit (15nt) Standard Sensitivity, Agilent). Due to a non-conventional 28 S/18 S ribosomal ratio in this species, sample quality was evaluated on the electropherogram46. No degradation in the inter region was observed. Total RNA samples were stored at −80 °C.

Raw reads assessment and quantification overview

All samples passed the FastQC33 “base quality control”. No relevant contamination hit was found after the alignment against E. coli, Yeast and PhiX. The mapping rate against the reference transcriptome22 of 206,012 contigs was higher than 72.4% for properly paired reads and higher than 93.6% considering both paired and single mate reads, validating the raw reads quality (Fig. 2, Supplementary Table 3). Furthermore, over the 42 samples, the maximal percentage of multi-mapped alignment is of 3.31% (Fig. 2, Supplementary Table 3).

Fig. 2
figure 2

Mapping statistics of the 42 samples against the set of 206,012 contigs. Number of million (M) reads: paired, single mate, and unmapped. Details are available in Supplementary Table 3.

Full size image

For an overview of the quantification matrix, a principal component analysis (PCA) was generated on the raw pseudo-count (log2 (count + 1)) non-normalized matrix (Fig. 3). Results showed a clear separation between samples from “North” and “South” stations, indicating environmental variations that might be due to latitude and/or sea ice-coverage.

Fig. 3
figure 3

Principal component analysis of the 42 samples based on expression levels of the whole transcriptome (206,012 contigs). “Groups” grouping the 3 replicates per time and per station.

Full size image

Filtering

Of the 206,012 transcripts, 37% (76,550) were expressed above the threshold of 1 CPM. This result corroborates previously observed results on the C. finmarchicus transcriptome22. Thus a large proportion of the whole contigs (63%) exhibited an extremely low level of expression, representing only 1.32 ± 0.04% of total aligned reads at “North”, and 1.27 ± 0.05% at “South” (Table 3, Supplementary Table 4).

Table 3 Average number (million, mean ± SE) of alignments per station and percentage of alignments discarded by filtering contigs with very low expression.
Full size table

Contigs annotation

By selecting the best hit for each database, the annotation matrix generated with Diamond37 has led to 36,274 and 22,527 contigs with an annotation in at least one database out of the 206,012 and 76,550 contigs respectively (Table 4). Moreover, the number of unique hits for each database is always lower than the number of contigs annotated by the respective database, highlighting the contigs’ functional redundancies.

Table 4 Number of contigs annotated with DIAMOND against NR, TREMBL and Swissprot and number of unique hits in the target database.
Full size table

The InterProScan annotation provided annotations from many protein signature databases. The main results are presented in Table 5 and Supplementary Table 5. A GO was attributed to 65,924 contigs over the whole transcriptome (206,012 contigs), while 33,057 contigs out of the 76,550 contigs with an expression level higher than 1 CPM in at least one sample had a GO annotation (Table 5, Supplementary Table 5).

Table 5 Number of contigs with an InterProScan annotation and details on main features.
Full size table

Quantitative real-time PCR data for normalization verification

The relative expression of six core circadian clock genes (clock, cycle, period1, timeless, cryptochrome2, vrille) and 2 circadian clock-related genes (cryptochrome1 and doubletime2) was investigated by quantitative real-time PCR and are available in Supplementary Table 6, allowing the verification of RNA sequencing normalization for further investigations. Regarding the two normalizations, the down-sampling normalization was selected for a rhythmic analysis based on concordant temporal expression profiles with qPCR data (using RAIN algorithm47), while the RLE normalization has been validated for differential expression analysis of the mean level of expression between stations, using the 21 samples of each stations as replicates.

Usage Notes

We present here the first in situ daily transcriptomes from the high Arctic, where molecular investigations of biological rhythms are exceptionally limited15,16. The samplings have been realized during drastic Polar photic conditions, i.e. the summer solstice, when daily oscillations of the Sun are minimal, high in the sky and always above the horizon15. The proposed datasets are thus novel and of interest due to the unique geographical location and time of year, the ecological importance of C. finmarchicus, and the rigorous temporal sampling strategy. Another strength of this dataset is the high depth of the RNA sequencing, with an average of 126 ± 5 million of reads (mean ± SE) per sample, which optimizes the detection of rhythmic transcripts25 in a species with a large genome20,21. Finally, the elaborate annotation of the large transcriptome is now publicly available and is thus accessible for further research.

The sampling strategy is optimized for rhythmic analysis, and particularly adapted for RAIN algorithm analysis23,25,47. Moreover, dataset allows powerful differential gene expression analysis using the 21 samples per station as replicates providing time-integrated detection of differentially expressed genes in C. finmarchicus with latitude/sea ice-cover. With climate driven environmental changes, this dataset ultimately constitutes new insights into transcriptomic regulation in the northward migrating copepod C. finmarchicus.