Ethical statement

The study was performed on a sibling group of Atlantic salmon (Salmo salar) from Mowi, Tveitevågen, Norway. Euthanasia and subsequent sacrifice of fish and the provision of eggs and sperm were conducted by an on-site aquaculture farmer according to the Regulations on Slaughterhouses and Production Facilities for Aquaculture Animals (FOR-2014-12-15-1831). Unfertilized eggs and sperm were brought to the High Technology Center, Bergen, Norway, where both fertilization and fish rearing occurred in a facility (University of Bergen) approved by the Norwegian Food Safety Authority (VSID2135). Fish did not undergo handling except for euthanasia and, therefore, special approvals were not required according to Norwegian national legislation as stated in the Norwegian Animal Welfare Act (LOV-2015-06-09-16-65) and Regulations on the Use of Animals in Experiments (FOR-2017-04-05- 451), given by the EU for animal experiments (Directive 2010/63/EU). All fish were euthanized on-site with an overdose of metacaine (MS-222, MSD Animal Health, Netherlands) before further handling.

Experimental design and sampling

Details of the 557 days experiment and sampling points are given in S1 Fig. The fertilized eggs were reared in egg incubators with a light:dark (LD) cycle of 14:10 (light switched off and on at 22:00 and 08:00, respectively) from fertilization to first feeding with white LED as described in [17]. Three months before sampling of the parr stage, the lights were changed to two ceiling lamps containing four fluorescent tube lights (GE Starcoat T5 Long Last 1449 mm Fluorescent Tube 49W, 3000K Warm white (General Electric, MA, USA)) for the rest of the experiment. The feeding fry to parr stage were kept in a LD cycle of 20:4 until they reached a size of approximately 11 cm. Photoperiodic induction of the parr-smolt transformation was conducted by exposing the parr to a LD 12:12 photoperiod (light switched off and on at 20:00 and 08:00, respectively), which mimics a winter light signal for 6 weeks, followed by a full 24 h daylength (LD 24:0) to resemble the summer light environment. This latter light period of LD 24:0 was subsequently maintained until the end of the experiment. During the experiment the fish were kept in black tanks (60 cm in diameter x 70 cm high). The fish were feed with Ewos (Norway) freshwater diet with the appropriate feed according to the size of the fish, by automatic feeders following the light period. The temperature from fertilization to first feeding was 6.0 ± 0.5°C which was gradually increased after first feeding to 11.5°C. The temperature was held constant for the rest of the experiment, as the temperature for avoiding the temperature as a zeitgeber, and it was determined to be 11.5 ± 0.3°C prior to seawater transfer, 11.1 ± 0.13°C after this transition. The transition from fresh to sea water was done gradually by a change of the inlet water from fresh to seawater, generating full seawater within a few hours. Five samplings were conducted through the parr-smolt transformation with the four first stages in freshwater and the fifth stage in seawater. At each stage, four fish were euthanized, weighed and the fork length measured, further, brains and left eyes were carefully dissected and individually snap frozen in liquid nitrogen, before storage at – 80°C. The following stages were sampled starting at 12:00: i) parr before the application of a winter signal (Parr, zeitgeber time (ZT) ZT8); (ii) parr reared and exposed to a winter signal for 6 weeks (ParrWS, ZT4); (iii) mid-smolt 22 days after transition to the summer signal (MidSmolt, no ZT); (iv) peak-smolt 51 days after applying the summer signal (PeakSmolt, no ZT); and (v) post-smolt approximately one month in seawater (PostSmolt, no ZT) (see S1 Fig). Note that eyes were unfortunately not sampled at i) parr before the application of a winter signal (Parr), only brains. Weights (g), fork lengths (cm) and condition factors (where K-factor = (100*weight)/(length^3)) were plotted with standard deviation (S2 Fig) to show that the K-factor values, a well-established indicator of smolt development, were consistent with the literature and decreased during smoltification [39,40].

RNA extraction

Four brains and eyes per stage were immersed in prechilled (−80°C) RNA later ICE (Invitrogen, Carlsbad, CA, USA) for at least 48 h at −20°C. After treatment, brains were individually inspected to confirm that the pineal and pituitary glands were intact, and the lenses of the eyes were removed. Total RNA was isolated from individual whole brains and eyes using TRI-reagent (Sigma, St. Louis, MO, USA) and the brains and eyes were divided into pieces less than 100 mg before total RNA isolation, following manufacturer’s instructions. The isolation was performed separately on each piece and prior to the precipitation step, lysates originating from the same brain or eye were pooled to ensure that the final RNA sample represented the entire individual brain or eye. After purification, a nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) was used to measure the RNA quantity and quality before and after DNase I treatment (TURBO DNA-free Kit, ThermoScientific, Waltham, MA, USA). Prior to RNA sequencing, RNA integrity was checked using an Agilent2100 Bioanalyzer with RNA 6000 Nano Kit (Agilent Technologies, CA, USA). All RNA integrity numbers (RIN) were determined to be between 8.4-10 and deemed valid for further RNA processing.

RNA sequencing

For RNA sequencing total RNA samples were submitted to the Genomics Core Facility at the University of Bergen (Bergen, Norway). Samples were processed using Illumina TrueSeq Stranded mRNA Sample Preparation Kits according to the manufacturers protocol and sequenced on the Illumina HiSeq 4000 platform (Illumina, San Diego, CA, USA. Sequencing generated a mean of 57 million 75 bp paired end reads per sample, which is within the recommendations from Illumina (https://knowledge.illumina.com/library-preparation/rna-library-prep/library-preparation-rna-library-prep-reference_material-list/000001243). All sequencing data were deposited to the European Nucleotide Archive (Accession number PRJEB103069). In addition, previously generated RNA sequencing data [17] were included and analyzed, to expand the timeline and give an initial developmental profile of Atlantic salmon. Specifically derived from this developmental series of whole Atlantic salmon individual samples were obtained at 255 dd (around the eye pigmentation stage (n = 6)), 379 dd (before hatching (n = 6)), 555 dd (after hatching (n = 6)) and 690 dd (before first feeding (n = 8)). The developmental series of larvae and alevins was sampled (ZT2) and snap frozen in liquid nitrogen and total RNA was extracted using the same protocols as described for brains and eyes. RNA sequencing of the developmental series was conducted at Genomics Core Facility at the University of Bergen (Bergen, Norway) as described above, see [17] for details.

RNA sequencing analysis

After RNA sequencing, all results were trimmed by applying Trimmomatic software version 0.38 [41] and aligned to the published Atlantic salmon reference genome (http://ftp.ensembl.org/pub/release-106/gtf/salmo_salar/) by using STAR version 2.7.0 [42]. Output files were processed by Samtools version 1.6 [43] and counts were generated using HTSeq version 0.11.2 [44]. Two arrays of normalized count files were generated by DESeq2 version 1.26.0 [45] namely (i) normalized counts of the eyes through the parr-smolt transformation (S1 Table) and (ii) normalized counts of the brains through the parr-smolt transformation (S2 Table). The variability in the data was visualized through two Principal Component Analysis (PCA) plots (S3 Fig) generated by DESeq2, using the default ntop = 500 top features by variance, showing eye and brain samples, respectively. DESeq2 was also used to perform differential gene expression analyzes applying the Wald statistical model with an adjusted p-value was set to < 0.05. In the analyzes genes with counts less than 10 per comparison were not included. The differential expression analysis compared all sampling stages against each other. For the eye samples six comparison were done (ParrWS vs MidSmolt, ParrWS vs PeakSmolt, ParrWS vs PostSmolt, MidSmolt vs PeakSmolt, MidSmolt vs PostSmolt, PeakSmolt vs PostSmolt) and for the brain samples ten comparisons were done (Parr vs ParrWS, Parr vs MidSmolt, Parr vs PeakSmolt, Parr vs PostSmolt, ParrWS vs MidSmolt, ParrWS vs PeakSmolt, ParrWS vs PostSmolt, MidSmolt vs PeakSmolt, MidSmolt vs PostSmolt, PeakSmolt vs PostSmolt). Annotated Ensembl GeneIDs of Atlantic salmon Ssal_v3.1 were used in the analyzes, see annotations in [11] for the visual opsin. Where possible, opsin gene names and their spectral sensitivities will be referred to by the use of standard photopigment nomenclature conventions and their specific wavelengths. In some cases to improve understanding, however, some opsins may be stated in terms of color when referring to the region of the light spectrum that their resultant photopigments are sensitive to, although it is acknowledged that color is perceived and is a concept that is strongly based on human psychophysical principles (e.g., blue, green and red opsins refer to photopigments that are short-, medium- and long-wavelength-sensitive). Further, see [46] for annotations of clock genes, circadian rhythm genes and enzymes of the melatonin biosynthesis pathway. Updated annotations of nonvisual opsins genes in Atlantic salmon, genes involved in the metabolism of TH and RA, and genes derived from the mammalian model of photoperiodism are listed in S3 Table. Graphs visualizing the normalized counts and PCA plots were generated using the ggplot2 v3.3.6 package [47] in Rstudio [48] and fold-changes were calculated by comparing logfold2 change between an average normalized count for one group against another.

Dynamics of visual opsins gene expression in early development

RNA sequencing was used to study visual opsin expression in whole Atlantic salmon larvae and alevins, ranging from the eye pigmentation stage to before first feeding. The results showed little or no expression of visual opsins genes before hatching, but the expression levels of all visual opsin classes started to increase during the transition of hatching (370 dd and 555 dd), with substantial expression towards first feeding (690 dd) (Fig 1). The expression of rod opsin (rh1–1) was most prominent (around 73% at 690 dd) (Fig 1A), followed by the subsequent primary cone classes: rh2–1 (a green or medium-wavelength-sensitive opsin) (Fig 1C), lws2 (red or long-wavelength-sensitive) (Fig 1D), and sws1–1 (UV-sensitive opsin) (Fig 1B). A single blue opsin class (sws2) (Fig 1B), three out of four green opsins (rh2-2, rh2–3 and rh2–4) (Fig 1C), and one out of four red opsins (lws4) (Fig 1D) were also expressed at a lower level, while two other red opsins (lws1 and lws3) (Fig 1D) were not expressed before first feeding.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Visual opsin gene expressions at early developmental stages before first feeding.

The graphs show normalized counts in the whole alevin with standard deviation at four developmental stages before first feeding. A) All 11 functional rod (black line) and cone (colored lines) opsins plotted together, showing that rod opsin (rh1−1) was expressed at the highest level compared to all cone opsin subclasses. B) Plotting of the short-wavelength-sensitive opsins shows that the expression of sws1−1 increased during hatching, with sws2 being substantially expressed only before first feeding at 690 dd. C) Medium-wavelength-sensitive opsins are all expressed before first feeding, with rh2−1 expression increasing during hatching, while the other rh2 genes showed substantial expression levels around 690 dd. D) Among the long-wavelength-sensitive opsins, both lws2 and lws4 were expressed before first feeding, with lws2 being the most prominent. Both lws1 and lws3 exhibited no expression during early development.

https://doi.org/10.1371/journal.pone.0349748.g001

Dynamics of visual opsins expressed in the eyes during the parr-smolt transformation

Differential gene expression analyzes showed that many genes were differentially expressed in the eye through smoltification, with an adjusted p-value set to < 0.05. Table 1 shows the total number of differentially expressed genes (DEGs) across the various stages of smoltification, with visual opsin genes being indicated. Also, see S4 Table for details, listing DEGs (between all stages) among the visual and nonvisual opsin genes and genes important in the metabolism of TH and RA, the latter being two molecular components that are involved in regulating the differential expression of tandem duplicated opsins, such as rh2 and lws subclasses [28].

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. The number of differentially expressed genes (DEGs) (adjusted p-value < 0.05) in the eye of Atlantic salmon.

https://doi.org/10.1371/journal.pone.0349748.t001

All 11 functional visual opsins were expressed in the eye during the parr-smolt transformation, with many exhibiting dynamic expression profiles through smoltification (Fig 2). In addition, these visual opsin smoltification expression patterns were compared to selected genes important in the metabolism of RA and TH to determine if these latter ventralizing factors coincided with the differential expression of tandem duplicated rh2 and lws opsin gene classes. As expected, rod opsin continued to be the most prominently expressed opsin and held around 85% of the total normalized counts for the 11 visual opsin genes in the eye through smoltification (Fig 2A). However, rh1−1 was not differentially expressed during smoltification (S4 Table) and had an average normalized count of over 600 000. By contrast, sws1−1 exhibited an over 1-fold decrease in expression from ParrWS to PostSmolt and was differentially expressed between these two stages (Fig 2B and S4 Table). Blue (sws2) opsin showed a relatively constant expression through smoltification (with an average of approximately 14 000 normalized counts between the stages) and was only differentially expressed between PeakSmolt and PostSmolt with a small fold decrease in expression (Fig 2C and S4 Table). Among the green opsins (Fig 2D), the expression of rh2−1 differentially increased greatly during smoltification and by 7-fold from ParrWS to PeakSmolt (S4 Table), with an average normalized count of over 40 000 at PeakSmolt. Compared to the other green opsins, rh2−2 exhibited low expression levels with an average of around 2000 normalized counts for the four stages but was differentially expressed between several sampling conditions (Fig 2D and S4 Table). For example, rh2−2 was lowly expressed at ParrWS, and had an almost 5-fold increase in expression between ParrWS and PeakSmolt. For both rh2−3 and rh2−4 the expression decreased though smoltification, with high expression levels in ParrWS (around an average of 20 000 and 30 000 normalized counts, respectively) and low in PostSmolt (around an average of 7000 and 8000 normalized counts, respectively), totaling over 1-fold (rh2−3) and 2-fold (rh2−4) decrease between these two outer stages (Fig 2D), however the rh2−3 was not significantly differentially expressed at any stage (S4 Table). Regarding red opsin expression, lws2 showed the highest level of expression, which was also observed at early developmental stages, with the greatest level of lws2 observed at PeakSmolt (approximately 35 000 normalized counts), upregulated approximately 1-fold from ParrWS to PeakSmolt (Fig 2E). Although not expressed before first feeding, the expression of lws1 increased differentially through smoltification and had almost a 3-fold increase from ParrWS to PostSmolt (from around an average of 1300–9700 normalized counts), yielding a 2-fold increase between MidSmolt and PeakSmolt (Fig 2E and S4 Table). In contrast to lws1 and lws2, the expression levels of lws3 and lws4 were very low throughout smoltification. Among the genes that function as ventralizing factors (see Fig 2E for selected genes based on expression levels and S4 Fig for all members), the deiodinase members, which are involved in the metabolism of TH, peaked at MidSmolt (i.e., dio3a2) and PeakSmolt (specifically dio2b were differentially expressed with an over 1-fold change from ParrWS to PeakSmolt) stages. Whereas genes that are involved in the metabolism of RA showed an increase during smoltification, for example, retinaldehyde dehydrogenase (aldh1a3.2) had highest expression at PeakSmolt with a significant increased from ParrWS to PeakSmolt, and a member of the cryptochrome P450 class, cyp26b1.1 was significantly peaking at PostSmolt with a 1,5- fold change from PeakSmolt to PostSmolt (Fig 2E and S4 Table). Interestingly, peaks of dio2b and aldh1a3.2 at PeakSmolt coincided with peaks in rh2−1 and lws1 (Figs 2D-2F).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Visual opsin expression in the eye during the parr-smolt transformation.

The graphs show normalized counts with standard deviation at four sampling points through smoltification. A) All rod (black line) and cone (colored lines) visual opsins are plotted together showing that the expression of rh1-1 was dominant in the eye through smoltification. B) Short-wavelength-sensitive-1 opsin (sws1-1) significantly decreased in expression from ParrWS to PostSmolt. C) Expression of sws2 was relatively stable through smoltification. D) Medium- wavelength-sensitive opsins demonstrated dynamic expression patterns through smoltification, with rh2-1 being the gene with prominent expression in the seawater phase at PostSmolt, with rh2-2 being lowly expressed at all stages, with rh2-3 and rh2-4 being dominantly expressed at ParrWS. E) Among the long-wavelength-sensitive opsins, lws1 increased in expression towards the seawater phase, lws2 was highly expressed throughout smoltification with a peak at PeakSmolt, whereas lws3 and lws4 were both lowly expressed. F) Selected ventralizing factors involved in the metabolism of thyroid hormone (TH) (namely dio2b and dio3a2) and retinoic acid (RA) (specifically rdh10a2, aldh1a3.2 and cyp26b1.1) were dynamically expressed with maximum expression levels towards the seawater phase.

https://doi.org/10.1371/journal.pone.0349748.g002

The dynamics of visual opsin expression reflect significant spectral shifts in ocular photopigments of Atlantic salmon

The predicted absorbance maxima (i.e., λ_max values) of visual photopigments found in Atlantic salmon were previously been determined using manual examination of known and well-understood tuning sites and molecular dynamics simulations-based modeling approaches [11].

As such, the spectra for visual photopigments expressed at different developmental as shown in Fig 3. In general, the total number of normalized counts for visual opsins in the eye during smoltification was highest in ParrWS exposed to a photoperiod of 12:12 (i.e., 12 h less of simulated daylight compared to the other stages), especially the rh1−1 gene expressed in rods responsible for dim light vision, and the expression decreased as the photoperiod was changed to continuous light (LD 24:0) (Fig 3E). At the alevin stage just before first feeding, all visual photopigment genes from sws1−1 (with a λ_max at 360 nm) to lws2 (with a λ_max at 553 nm) were expressed, except for lws1 (560 nm) and lws3 (543 nm), however note the low number of normalized counts apparent in a whole alevin compared to eyes through smoltification (Fig 3C). Among the normalized counts for visual opsins rod opsin counted for around 73%, followed by rh2−1, lws2 and sws1−1, visualized in Fig 3D by percentage of the total normalized counts of visual opsins not including rh1−1. Overall, both ParrWS and MidSmolt stages exhibited similar expression profiles, notable with expression of sws1−1 (λ_max of 360 nm) and both rh2−3 (λ_max of 511 nm) and rh2−4 (λ_max of 506 nm) were highly expressed in ParrWS and MidSmolt, together with lws2 (λ_max of 553 nm) (Fig 3A, 3F, 3E). Surprisingly, in addition to the declining sws1−1 expression, at both PeakSmolt and PostSmolt stages, there were shifts in green photopigment gene expression, from rh2−3 (λ_max of 511 nm) and rh2−4 (λ_max of 506 nm) to rh2−1 (λ_max of 472 nm), accounting for peak spectral sensitivity change predicted to be almost 40 nm towards the shorter wavelengths (Fig 3B, 3E, 3F). Despite the significant short-wavelength shift in rh2-based photopigments, Atlantic salmon exhibited a small long-wavelength shift of predicted 7 nm in the lws class of photopigments, a change derived from an increase in expression of lws1 (λ_max of 560 nm) in addition to the relatively high and stable expression of lws2 (λ_max of 553 nm) during smoltification (Fig 3D, 3E, 3F). Of note, blue opsin (sws2) had a relatively stable expression during smoltification, while lws3 (λ_max of 546 nm) was not substantially expressed at any of the developmental stages examined.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Spectral sensitivity dynamics of visual photopigments.

The spectral profiles of visual photopigments are based on previously determined absorbance maxima (i.e., λ_max values) based on inferred predictions [11]. Note that these λ_max values applied a standard A₁-based rhodopsin template in the evaluation as salmonid ocular photoreceptors (and the photopigments expressed therein) predominantly utilize a retinal chromophore derived from vitamin A₁ [26]. A) Both ParrWS and MidSmolt had tetrachromatic vision with expression of sws1−1 (λ_max of 360 nm), sws2 (λ_max of 420 nm), rh2−3 (λ_max of 511 nm) and rh2−4 (λ_max of 506 nm) and lws2 (λ_max of 553 nm). B) In PeakSmolt and PostSmolt the vision became trichromatic as sws1−1 expression diminished. Great shifts in expression levels among the green photopigments were observed, expression levels of rh2−3 (λ_max of 511 nm) and rh2−4 (λ_max of 506 nm) were downregulated while rh2−1 (λ_max of 472 nm) was greatly upregulated, accounting for peak spectral sensitivity change of up to almost 40 nm. In addition, the red photopigment lws1 (λ_max of 560 nm) had an increase in expression level shifting the spectral sensitivity 7 nm towards the longer wavelengths. C) The total number of normalized counts of visual opsins in whole alevins (data including whole body) and the eyes through smoltification visualized as bars, illustrating high expression levels of rod opsin, accounting for over 80% of the total counts for visual opsin genes during smoltification, with the highest level in ParrWS (approximately 89%). F) The dynamics of cone opsin genes calculated for each stage, given by the percentage of the total normalized counts of visual opsins not including rh1−1.

https://doi.org/10.1371/journal.pone.0349748.g003

Transcriptomic changes of nonvisual opsins in the eye and brain during parr-smolt transformation

There were eleven nonvisual opsin genes among the DEGs in the eye during smoltification (S4 Table) and their dynamics are shown in S5 Fig. Notable, members of the retinal G protein-coupled receptor opsin (rgr) had high expression level compared to other nonvisual opsins in the eye during smoltification. The highest expression level was observed for rgrb2 with an average of 10 000 normalized counts at MidSmolt and a downregulation in expression at PeakSmolt, with an over 1-fold decrease in expression between MidSmolt and PeakSmolt and an over 1-fold increase between PeakSmolt and PostSmolt (see S4 Table for details). Further, opn6a1 had highest expression at ParrWS, before the expression level decreased towards the seawater phase by almost 1-fold change (S5 Fig). In addition, differential gene expression analyzes, with an adjusted p-value set to < 0.05, were done for the brain comparing all stages through smoltification and Table 2 reveals many DEGs in the brain, including the number of nonvisual opsins among the DEGs (see also details in S5 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. The number of differentially expressed genes (DEGs) (adjusted p-value < 0.05) in the brain.

https://doi.org/10.1371/journal.pone.0349748.t002

The expression profiles of 42 nonvisual opsins in the Atlantic salmon brain during smoltification showed that several of the opsins have increased expression towards the seawater phase (Fig 4A). However, only eight nonvisual opsins had a count over 100 and their expression dynamics were plotted in Fig 4B and among the plotted genes, tmtopsin3b was the only gene not significantly differentially expressed (S5 Table). Interestingly, the pineal-specific exorhodopsin had the most prominent expression level in the brain, followed by rgrb1, and had an almost 5-fold significant increase in expression level from Parr to PostSmolt (S5 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. The dynamics of nonvisual opsins during smoltification.

A) Heatmap of the normalized counts of all nonvisual opsins in the brain, scaled by row, demonstrates an increase in several of the opsins towards the seawater stage. B) The graphs show normalized counts with standard deviation at five sampling points in the brain through smoltification for the eight nonvisual opsins with normalized counts greater than 100. Exorhodopsin (exorh) had the highest expression among the opsins in the brain, followed by retinal G protein-coupled receptor opsin (rgrb1).

https://doi.org/10.1371/journal.pone.0349748.g004

Further, the distribution of normalized counts for nonvisual opsins were visualized by percentage to reveal the differential expression profile between the eye and brain over the course of smoltification as a total (Fig 5). The results showed that in the brain exorh and rgrb1 (with approximately 28% and 20% of the total nonvisual opsin counts, respectively) were the most abundant genes, followed by genes in the opn7 class (e.g., opn7b and opn7 cd1 at approximately 6% and 9%, respectively) and tmtopsin class (e.g., tmtopsin3b at approximately 5%). In the eye, genes in the rgr class held 65% of the normalized counts, with a high expression level of rgrb2 (around 45% of the counts). In addition, genes in the opn6 class held in total approximately 14% of the normalized counts, rrh approximately 6% and opn4x1a above 5%. Interestingly, rgrb2 and not rgrb1 was highly expressed in the eye, indicating a sub-functionalization of the two genes.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Differential expression profiles of nonvisual opsins in the eye and brain of Atlantic salmon.

Bar chart of the nonvisual opsins plotted by percentage in the eye and brain. The total number of normalized counts at the four sampling points (ParrWS to PostSmolt) for the nonvisual opsins was set to 100 percent for each tissue and the distribution among the different opsins was plotted back-to-back. In the brain exorh was the most abundant opsins, followed by genes in the rgr class. In the eye, genes in the rgr class were highly expressed, with rgrb2 counting for approximately 45%.

https://doi.org/10.1371/journal.pone.0349748.g005

Dynamics of clock genes and outputs of the circadian clock mechanism during smoltification

A total of 35 clock genes were among the DEGs detected in the brain sampled at the same time of the day during smoltification (see S5 Table for details on significant changes between timepoints). The heatmap of normalized counts of all clock genes shown in Fig 6A, reveal that several of the genes changed expression levels in the brain from Parr to ParrWS, when there was a change in the photoperiod from LD20:4 to LD12:12. Also, changes in expression levels of the clock genes were detected in ParrWS compared to MidSmolt, with the photoperiodic shift from LD 12:12 to LD24:0. Interestingly, the developmental stages MidSmolt, PeakSmolt and PostSmolt, all exposed to LD24:0, had differential expression of clock genes and analysis revealed nine DEGs comparing PostSmolt to PeakSmolt, with seven of the genes being upregulated in PostSmolt (S5 Table). Further, Fig 6B demonstrates detailed expression profiles of differentially expressed core clock genes, one member from each class, with expression profiles of all DEGs in each class in S6 Fig. Among the clock members, clock1a.2 was the only DEG (S5 Table), with a slight decrease in expression towards the seawater phase (Fig 6B and S6A Fig). All members of the arntl class were differentially expressed during smoltification and in general the members had the highest expression level in Parr and were downregulated in ParrWS (Fig 6A, S6A Fig and S5 Table). Illustrated by arntl1b.2 that had a decreased by approximately 0,5-fold between Parr and ParrWS, before a slight increase in expression level towards PeakSmolt, followed by a decrease in PostSmolt (Fig 6B). All period members, except per2a were among the DEGs and showed a similar trend in expression (S6B Fig and S5 Table). Highlighted by the per1b with the most prominent change, having an almost 2-fold increase between Parr and ParrWS, before a decrease of more than 1-fold between ParrWS and PeakSmolt and a subsequent 1-fold increase in expression level between PeakSmolt and PostSmolt (Fig 6B). Among members of the cryptochrome class, all genes except cry2 were differentially expressed and cry3b.1 had the greatest change in expression level, with an almost 1,5-fold increase between ParrWS and MidSmolt, followed by a decrease towards the seawater phase (Fig 6B, S6C Fig and S5 Table). In the nr1d class, all genes except nr1d4a.2 were differentially expressed (S5 Table). Notable, nr1d1a, nr1d2a.2, nr1d4b.1 and nr1d4b.2 had dynamic expression profiles that peaked at MidSmolt, decreased in PeakSmolt before increasing again in PostSmolt. For example, nr1d4b.1 had an almost 1,5-fold increase in expression level between Parr and MidSmolt, before a 0,5-fold decrease between MidSmolt and PeakSmolt and a 0,5-fold increase between PeakSmolt and PostSmolt (S6D Fig). In the ror class, roraa.1 and roraa.2 had the highest expression level with a significant increase from Parr to ParrWS, that counted for an over 0,5-fold change between the two stages, and the two DEGs in the cnsk class had an increase in expression from the Parr to PostSmolt (S6 Fig and S5 Table). In addition, the distribution of normalized counts for clock genes was visualized by percentage to reveal the differential expression profile between the eye and brain during smoltification (S7 Fig). In difference from the nonvisual opsin genes, the expression levels of clock genes were more evenly distributed. The genes with highest percentage in both tissues were per1b (around 8% in the eye and 5% in the brain) and rorcb.1 (6,5% in the eye and 6% in the brain) (S7 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Expression profiles of the clock genes in the brain of Atlantic salmon.

A) The heatmap of normalized counts during smoltification is scaled by row and shows a dynamic expression of the clock genes. B) Normalized counts with standard deviation are plotted for selected differentially expressed genes within each class of the core clock genes, clock, arntl, cry and period. The profiles show prominent differences between the genes within each class and developmental stages, e.g., arntl1b and per1b have opposite peaks and troughs at ParrWS, PeakSmolt and PostSmolt.

https://doi.org/10.1371/journal.pone.0349748.g006

Further, melatonin is an important output of the circadian clock mechanism and genes coding for enzymes that dictate melatonin synthesis were analyzed. Most genes involved in the enzymatic steps were differentially expressed in the brain and had an upregulation from Parr to PostSmolts (Fig 7 and S5 Table). Interestingly, the two paralogues of the melatonin synthesizing enzyme arylalkylamine N-acetylserotonin 2 (aanat2.1 and aanat2.2) were significantly upregulated by approximately 6,5-fold and 4-fold, respectively, from Parr to PostSmolt, where about 2-fold and 1,5-fold of the increase took place in the transition from PeakSmolt to PostSmolt (S5 Table). Also, two paralogues of the last enzyme in the synthesis of melatonin N-acetylserotonin O-methyltransferase (asmt.2 and asmt2) had an approximately 4,5-fold and 1-fold, respectively, significant increase from Parr to PostSmolt (Fig 7 and S5 Table). The distribution of normalized counts for all important enzymes of the melatonin synthesis was visualized by percentage to reveal the differential expression profile between the eye and brain during smoltification (S8 Fig). Of notice, the expression levels of aanat1 and aanat2 were as previously reviewed [49,50] mainly distributed between the eyes and brain, respectively, with the three aanat1 genes counting for over 38% of the important melatonin synthesizing enzymes in the eye, while aanat2 counted for approximately 22% in the brain (S8 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Expression levels of important genes in the melatonin synthesis in the brain of Atlantic salmon.

The heatmap, scaled by row, shows the differentially expressed genes (DEGs) with a count more than 100 in the melatonin synthesis from tryptophan to melatonin, revealing in increased expression towards the seawater phase.

https://doi.org/10.1371/journal.pone.0349748.g007

Dynamics of genes in the mammalian model for photoperiodism in the brain

The expression profiles of the pituitary specific growth hormone (gh) and important elements of the mammalian model for photoperiodism were analyzed in the period of parr-smolt transformation (Fig 8 and S5 Table showing differentially expressed genes). A large increase in the expression level of gh was shown between MidSmolt and PostSmolt, with an over 2-fold increase between MidSmolt and PeakSmolt, indicating of a proper parr-smolt transformation (Fig 8A). The genes in the photoperiodic switch, arntl2 and dec1, together with markers for summer (eye absent, eya3) and winter (chromogranin A, chga) physiology were studied [36]. In addition, the expression patterns of the deiodinases, the gatekeepers for controlling the availability of active or inactive thyroid hormone, that are implied to be involved in developmental and seasonal physiology of salmonids [51,52], were studied. The expression levels of all genes were plotted in S9 Fig and selected DEGs are shown in Fig 8B. Among the dec1 genes, three genes (dec1.1, dec1.3 and dec1.4) were significantly upregulated in ParrWS compared to Parr, while the three arntl2 genes were downregulated in the same comparison (S5 Table). The photoperiodic switch is illustrated by plotting arntl2c and two dec1 genes (dec1.1 and dec1.4, illustrating the similar expression pattern between paralogues) (Fig 8B). Comparing Parr and ParrWS dec1.1 and dec1.4 had an approximately 1-fold and 0,5-fold upregulation, respectively, while arntl2c had a 1-fold decrease. Further, dec1.1 was significantly downregulated between ParrrWS and PeakSmolt, while the arntl2 genes were upregulated comparing PeakSmolt and ParrWS (S5 Table and S9 Fig). The marker for winter physiology, chga, was upregulated in ParrWS compared to Parr and downregulated in PeakSmolt compared to ParrWS, while the two paralogues of eya3, the marker for summer physiology, were not differentially expressed (S5 Table and S9 Fig). Note that the level of thyroid-stimulating hormone (tshβb) expressed in the pituitary had a low expression level throughout smoltification, counting less than 10 normalized counts in all samples. Among the deiodinases, dio2b was differentially expressed through smoltification (S5 Table), with high expression levels at PeakSmolt after an approximately 2,5-fold increase in expression from Parr to PeakSmolt, while both dio2a and dio3a2 were lowly expression in the brain (S9 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. Growth hormone and selected genes in the model of photoperiodism.

The graphs show normalized counts in the brain with standard deviation, illustrating the dynamics through smoltification. A) The growth hormone (gh) increased towards the seawater phase, with a great increase from MidSmolt to PeakSmolt. B) The dec1 genes are represented by the expression patterns of dec1.1 and dec1.4, showing the similar expression pattern between paralogues, while the arntl2 genes are represented by arntl2c, showing opposing expression peaks at the different sampling points through smoltification. The expression level of dio2b was high at PeakSmolt.

https://doi.org/10.1371/journal.pone.0349748.g008

The tandem duplicated visual opsin genes have shifts in expression during smoltification

The transcriptomic profile of visual opsin genes in Atlantic salmon larvae and alevins confirmed previous findings, showing that the visual opsins are expressed before first feeding, with a duplex retina dominated by rhodopsin expression from the start [11,53]. The sampling points before (379 dd) and after hatching (555 dd), showed that the opsins were expressed after hatching with an increased level towards first feeding. All opsin classes were present before first feeding, including the four members of the rh2 class and two of the members in the lws class, providing the alevin with tetrachromatic vision. In accordance with previous studies, there was a decline in expression of sws1−1 in the parr-smolt transformation, causing the vision to become trichromatic towards the seawater phase [26]. Interestingly, the tandem duplicated medium-wavelength-sensitive and long-wavelength-sensitive opsin genes showed a dynamic expression during smoltification, not previously described. In freshwater, rh2−3 and rh2−4 dominated the expression among the green opsins, but at PeakSmolt there was a shift in expression to rh2−1, while the rh2−2 had low expression through smoltification. The change to rh2−1 counted for almost 40 nm shifts towards shorter wavelengths based on the predicted absorbance maxima (i.e., λ_max values), coinciding well with the change in light environment from river and ocean. Further, the tandem duplicated lws1 and lws2 counted for most of the expression in the lws class, with the lws1 appearing towards the seawater phase. Note that the changes in green and red opsins occurred under equal spectrum and intensities of light in this experiment and further studies are needed to verify additional changes in the visual system related to direct adaptation to the natural light in the marine environment. This will help to fully understand the photoreceptive adaptation and tuning under this life history transition. Thus, it is difficult to distinguish if the change in opsin gene expression is due to shifts in photic environment or an effect of the developmental transition. Differential expression pattern of the green and red opsin genes is well described in fish and has been shown both in time and space [20,27,54], although not described in a developmental transition such as smoltification. For example, during zebrafish development rh2−1 and rh2−2 were expressed earlier than rh2−3 and rh2−4 and in adult zebrafish rh2−1 and rh2−2 were located dorsally, while rh2−3 was expressed in a restricted area ventrally, next to the ventral-most rh2−4 expression. In addition, lws1 was found in the ventral region and lws2 in the dorsal region, and for both classes the paralogues with the highest absorbance maximum were expressed in the ventral region [20,28]. Zebrafish have been used as models for understanding the differential regulation of tandemly arrayed genes, and the varying spatial distribution of green and red opsins in the zebrafish eye has been shown to be determined by the gene order, proximal regulator regions and a common locus control region [28,55]. Further, the thyroid hormone was shown to be involved in regulation of green and red opsins in zebrafish, as abundance of TH shifts the expression to ventrally expressed opsins with sensitivities in the longer wavelengths (rh2−3, rh2−4 and lws1) [56]. Similarly, it has been shown that retinoic acid switches the expression from lws2 to lws1 and that the RA signaling domain coincides with a ventral zone of lws1 expression in zebrafish [57]. Analysis of the TH metabolism in the salmon eye during smoltification revealed that dio2b converting the inactive T4 to the active T3 [58] was upregulated at PeakSmolt. While the dio3a2 involved in the inactivation of TH by converting T4 to T3 and degradation of T3 to T2 [58] had a peak at MidSmolt. The upregulation of dio2b at PeakSmolt, promoting active TH, coincided with the upregulation of lws1 having the highest absorbance maximum. However, among the green opsins rh2−3 and rh2−4 decreased and rh2−1 increased at PeakSmolt, supporting a shift to shorter wavelengths, and expression of a paralogue known to be dorsally expressed in zebrafish. Evaluating the expression of important enzymes in the conversion of retinol to retinoic acid [59] showed an upregulation towards the seawater phase, with the retinaldehyde dehydrogenase (aldh1a1 and aldh1a3.2) peaking at PeakSmolt coinciding with the lws1 expression. However, a member of the cryptochrome P450 class involved in the degradation of RA [59] was peaking at PostSmolt, suggesting that a ventralizing signal of RA in PostSmolt was suppressed. Taken together, the conserved synteny of the tandem duplicated green and red opsins [60] and differential expression of the opsins suggest a similar regulation mechanism in Atlantic salmon as in zebrafish. Thus, elucidating the special distribution of opsins in the retina during smoltification would give an understanding of the dorsal-ventral patterning of opsins in salmon.

Dynamic expression of a rgr class member and the pineal specific exorhodopsin

Only a few nonvisual opsins showed a substantial shift in expression in the eye and brain during smoltification. In the eye, rgrb2 was highly expressed during smoltification but with a decrease in expression in PeakSmolt, while exorh showed an increased expression from the parr to the seawater phase in the brain. Comparing the expression of the nonvisual opsins in the eye and brain showed that members of the rgr class dominated in both tissues and the differential expression between rgrb1 and rgrb2 in the brain and eye, might account for a subfunctionalization of these salmonid specific paralogues. Members of the rgr class are suggested to be both opsins and photoisomerases and have been shown to be highly expressed in retinal pigmented epithelium of the eye, preferably binding to all-trans retinal [16]. A dynamic expression of rgrb2 in the eye during smoltification might be associated with shifts in expression of visual opsin genes probably with a function as a photoisomerase, photo-converting the chromophore from all-trans to 11-cis retinal [61]. The pineal-specific opsin exorh, expressed with a daily rhythmicity in zebrafish is suggested be involved in the melatonin synthesis as knockdown morpholino studies with exorh decreased the melatonin synthesizing enzyme (aanat2) [62]. Also, single cell RNA sequencing of the pineal organ of zebrafish has shown that the pineal photoreceptor cells express both aanat2 and the last enzyme in the synthesis of melatonin the N-acetylserotonin O-methyltransferase (asmt) [63]. Interestingly, the great upregulation of exorh during smoltification corresponded with an upregulation of genes in the synthesis of melatonin and might be related to growth of the pineal organ to provide a greater light sensitivity in an oceanic environment.

Clock genes showed dynamic expressions in the parr-smolt transformation

A total of 35 clock genes were differentially expressed in the brain with a dynamic expression during smoltification, including members of the core clock and the stabilizing loop. Similarly, many clock genes were shown to be dynamically expressed in the gills during smoltification, with a subset of clock genes suggested to be associated with the physiological preparation of seawater migration, independent of circadian regulation [38]. Further, earlier studies in the parr brain have shown daylength dependent expressions, by differences in expression between fish exposed to long or short photoperiods, with a 4 hour shift in acrophase for cry2, while other components of the core clock became arrhythmic under long photoperiods [37]. In the current study, analysis indicates that the differential expression was related both to daylength and the developmental transition from freshwater to seawater, as dynamic expression was detected between stages with different photoperiods, but also between stages in constant light. However, brains were sampled at the same time point at daytime at each stage and the change in expression between stages exposed to different photoperiods might be a result of shifted acrophase and should be interpreted with caution. For example, parr under LD12:12 with lights on at 08:00, had an acrophase at 08:30 for per1b [64] and the great shift in expression of per1b between Parr (LD 20:4 lights on at 04:00, ZT8) and ParrWS (LD12:12 lights on 08:00, ZT4) in the current study might be a result of a shift in acrophase due to changed light period. However, in MidSmolt, PeakSmolt and PostSmolt the fish were all held under continuous light, but three clock genes were significantly downregulated from MidSmolt to PeakSmolt and as illustrated by the per1b expression, a total of nine clock genes were either up- or downregulated between PeakSmolt and PostSmolt. The change in expression of clock genes under constant light conditions, might account for a shift related to the developmental transition from freshwater to seawater.

Photoperiodic regulation of smoltification

The level of plasma growth hormone is known to increase in Atlantic salmon as a response to a longer photoperiod in the spring [35,65,66]. The increase has been shown to be significant 20 days after the switch to longer days, consistent with a prior elevation of the tshβb, followed by an increase in dio2b in the brain [35]. However, the expression level of gh in the pituitary was not significantly increased after 20 days and has been suggested to be elevated later to compensate for a drop in the pituitary gh content [35,67]. The present results are consistent with these suggestions as the mRNA level of gh first increased at PeakSmolt, approx. 44 days after the switch to longer days. The transcriptome was also analyzed for elevation of tshβb and dio2b during smoltification and while dio2b increased in the brain at PeakSmolt as reported previously [35,52], the expected elevated level of tshβb at MidSmolt was not found. Though, the expression of tshβb in the few cells of the dorsal pituitary, near the pituitary stalk [68], is likely to have be masked by sequencing of the whole brain, and future studies should analyze the brain and pituitary separately. Factors in the mammalian model of photoperiodism were also analyzed [36] and chga the marker for winter physiology peaked in ParrWS while the two eya3 paralogues, marking long days in mammals, were not differentially expressed in the brain during smoltification. Interestingly, genes involved in the circadian clock mechanism suggested to function as a photoperiodic switch between the winter and summer physiology in mammals, showed differential expression during smoltification. The dec1 paralogues promoting winter physiology were significantly upregulated in ParrWS, coinciding with downregulation of arntl2 paralogues, followed by upregulation of arntl2 paralogues at PeakSmolt. Taken together, these results support a common photoperiodic regulation of seasonality in vertebrates and highlight a need for an in dept study of the pituitary transcriptome during smoltification.