Functional divergence of thyrotropin beta-subunit paralogs gives new insights into salmon smoltification metamorphosis

Smoltification is a metamorphic event in salmon life history, which initiates downstream migration and pre-adapts juvenile salmon for seawater entry. While a number of reports concern thyroid hormones and smoltification, few and inconclusive studies have addressed the potential role of thyrotropin (TSH). TSH is composed of a α-subunit common to gonadotropins, and a β-subunit conferring hormone specificity. We report the presence and functional divergence of duplicated TSH β-subunit paralogs (tshβa and tshβb) in Atlantic salmon. Phylogeny and synteny analyses allowed us to infer that they originated from teleost-specific whole genome duplication. Expression profiles of both paralogs in the pituitary were measured by qPCR throughout smoltification in Atlantic salmon from the endangered Loire-Allier population raised in a conservation hatchery. This revealed a striking peak of tshβb expression in April, concomitant with downstream migration initiation, while tshβa expression remained relatively constant. In situ hybridization showed two distinct pituitary cell populations, tshβa cells in the anterior adenohypophysis, and tshβb cells near to the pituitary stalk, a location comparable to the pars tuberalis TSH cells involved in seasonal physiology and behaviour in birds and mammals. Functional divergence of tshβ paralogs in Atlantic salmon supports a specific role of tshβb in smoltification.

A histological study by Hoar first reported an activation of thyroid follicles in Atlantic salmon during smoltification 9 . An increase in T4 plasma levels was then shown during smoltification in various salmonids (coho salmon, Oncorhynchus kisutch 10,11 , masu salmon, Oncorhynchus masou 12 ; Atlantic salmon 13 ), even though changes in T4 plasma levels were relatively limited as compared to other hormones such as cortisol, insulin-like growth factor-1 and growth hormone 3,5 . Experimental hormonal treatments suggested that TH may be responsible for smoltification-related change in rheotaxis and swimming behaviour 14 and olfactory imprinting 15 , but would be insufficient to induce preadaptation to osmoregulation in seawater 16 .
While many studies addressed TH, little is known on TSH and smoltification, and contradictory data have been obtained. In the Atlantic salmon, pituitary TSH cells were more numerous and had increased activity in presmolts and smolts than in parr 17 , but no ultra-structural changes of TSH cells were observed in coho salmon during smoltification 18 . No change or a slight decrease in pituitary tshβ mRNA levels were measured in smolts as compared to parr in Atlantic salmon 19,20 and coho salmon 11 . No variations in pituitary and plasma TSH protein levels were reported in coho salmon throughout smoltification 11 .
While extant amphibians, birds and mammals possess only a single TSH (a single tshβ gene), recent studies revealed the presence of duplicated tshβ paralogs in some other vertebrates 21 . Duplicated tshβ genes originated from whole genome duplication events that occurred in early vertebrates ("1R/2R", for "1 st and 2 nd rounds of whole genome duplication") and in early teleosts ("3R", for "3 rd round of whole genome duplication") 21 . Thus, chondrichthyans, such as the elephant shark, Callorhinchus milii, and basal sarcopterygians, such as the coelacanth, Latimeria chalumnae, have two tshβ paralogs issued from 2R (tshβ and tshβ2), while tetrapods have lost the tshβ2 paralog and conserved only a single tshβ gene 21 . Tshβ2 has also been lost in the actinopterygian lineage, but various teleost species possess two tshβ paralogs as a result of 3R-duplication of tshβ 21,22 , named in the present study tshβa and tshβb, according to the most common nomenclature "a and b" for teleost 3R-paralogs. In the present study, we searched for tshβ paralogs in the Atlantic salmon, considering also the additional genome duplication that occurred in the salmonid lineage ("4R") 23 .
We revealed the presence and expression of two tshβ paralogs in Atlantic salmon and brought new knowledge on the evolutionary history of tshβ genes in salmonids. We investigated the potential involvement of the tshβ paralogs in smoltification, using the Atlantic salmon from the Loire-Allier basin as a model. This population is the last extant salmon population able to migrate long rivers in Western Europe and is currently endangered. Samplings were performed at the "Conservatoire National du Saumon Sauvage" (CNSS), Chanteuges, France, whom breeds wild brood stock and produces juvenile salmon which are released at different developmental stages, as part of a conservation programme. Quality and timing of smoltification are key issues for this population, as smolts need to achieve a 900 km-downstream migration before reaching the Loire estuary in a narrow window of suitable physiological and environmental conditions. This requires new research advances on environmental and neuroendocrine regulatory mechanisms of smoltification and initiation of downstream migration. With the demonstration of a striking expression of one of the tshβ paralog, this study provides the first evidence of a peak expression of tsh during smoltification.

Results
two thyrotropin beta subunit (tshβ) paralogs in the Atlantic salmon. Using the recent Atlantic salmon genome assembly (GCA_000233375.4) 24 , we retrieved two genes with two exons each, encoding complete TSH β-subunit sequences: one gene (named in the present study tshβa) located on the chromosome ssa22 and corresponding to the tshβ sequence previously isolated 19 , and the second gene (named in the present study tshβb) located on ssa15 and coding for another tshβ; this tshβb sequence encompassed two exons, as all vertebrate tshβ, and included a previously identified exon 1 21 . In order to assess that the two tshβ paralogs are transcribed, cDNA sequences were successfully cloned using pituitary RNA from Atlantic salmon sampled during the parr-smolt transformation. A partial tshβa mRNA sequence (401 bp) was cloned and its sequence shared 100% identity with already characterized tshβ (AF060566) 19 . A tshβb mRNA sequence (517 bp) including the full length CDS was cloned (MG948546, this study); it presented 100% identity with the corresponding predicted tshβb sequence in the genome. A third putative tshβ locus was identified on ssa12, corresponding to a second paralog of tshβa. This locus includes only exon 2 with a frameshift mutation resulting in an early stop codon. Therefore this locus was identified as a tshβa-pseudogene ( Supplementary Fig. S1A). In order to confirm this pseudogene, we cloned and sequenced the genomic sequence using DNA extracted from testis of an Imsa River (Norway) Atlantic salmon. The cloning confirmed the loss of exon 1 and the presence of the deletion in exon 2 leading to an early stop codon ( Supplementary Fig. S1B). In addition, using Illumina short reads from an on-going sequencing project of the genome from a Loire-Allier Atlantic salmon, we aligned short reads against the two Atlantic salmon tshβa loci. For the pseudogene, read alignments confirmed the sequence of exon 2 with the presence of the deletion leading to the early stop codon. These results support the conclusion that the second tshβa paralog in Atlantic salmon is a pseudogene and that this loss is a common feature of Atlantic salmon populations.
Comparison of TSHβ deduced amino-acid sequences ( Supplementary Fig. S2) showed that Atlantic salmon paralogs TSHβa and TSHβb shared 31.5% identity and 47.5% similarity and that both have conserved the twelve cysteine residues that have been shown to be required for proper folding and functional activity of TSH in mammals 25 . Both paralogs also shared the typical N-glycosylation site conserved among vertebrate glycoprotein hormone beta subunits; in addition, Atlantic salmon TSHβb presented a second N-glycosylation site located between the two first cysteine residues, as with the other teleost TSHβb 21 . phylogeny analysis of tsHβ. Molecular phylogeny analysis was performed on 38 gnathostome TSH β-subunit sequences and using lamprey, Petromyzon marinus, glycoprotein hormone β-subunit (GpHβ) as outgroup ( Fig. 1; Supplementary Fig. S2). As previously shown 21 , most sequences grouped into a "classical TSHβ" clade, whereas a few chondrichthyan and basal sarcopterygian sequences formed a small TSHβ2 sister clade.
Among the TSHβ clade, the analysis also supported the two sister clades for teleost sequences resulting from teleost 3R, named here TSHβa and TSHβb following the teleost 3R paralogs nomenclature (previously named TSHβ and TSHβ3 21 ). The two Atlantic salmon TSHβ branched into the two teleost TSHβa and TSHβb clades, respectively ( Fig. 1) allowing us to classify and name the Atlantic salmon TSHβ paralogs, TSHβa and TSHβb. The pike, Esox lucius, representative species of a sister group of Salmoniforms, the Esociforms which have not undergone the salmonid 4R, also possessed the two teleost tshβ 3R paralogs, encoding for TSHβa and TSHβb ( Fig. 1). In contrast, we identified up to three tshβ genes in the genomes of rainbow trout, Oncorhynchus mykiss (GCA_002163495.1) and coho salmon (GCF_002021735.1): a single tshβb and two tshβa genes (tshβaα and tshβaβ according to the current nomenclature "α and β" for salmonid 4R paralogs 23 ). They encode for three putative TSHβ (TSHβaα, TSHβaβ and TSHβb, Fig. 1, Supplementary Fig. S2). synteny analysis of tshβ genomic region. To further assess the origin and nomenclature of salmonid tshβ paralogs, we performed a synteny analysis (Fig. 2) of the tshβ genomic region of the pike and of two salmonids, the Atlantic salmon and the rainbow trout. We used as a reference the tshβ genomic region of the spotted gar, Lepisosteus oculatus, holostean basal actinopterygian, which has not undergone the teleost 3R. Synteny analysis confirmed that the tshβ genomic region has been duplicated into two paralogons in the pike, in agreement with the teleost 3R, and further duplicated into four paralogons in the Atlantic salmon and rainbow trout, in agreement with the salmonid 4R. As an example, kcna10, one of tshβ neighbouring genes, was present as a single gene in the spotted gar, as two paralogs in the pike, and as four paralogs in the Atlantic salmon and rainbow trout, reflecting full conservation of 3R-and 4R-duplicated paralogs (Fig. 2). Other tshβ neighbouring genes, such as rplp2, tspan, or slc16a1, showed full conservation of 3R-duplicated paralogs, but incomplete conservation of 4R-duplicated paralogs leading to only three paralogs in salmonids (Fig. 2). Further paralog gene losses were observed for other neighbouring genes, and in particular for ap4b1, with a single gene present in the pike and salmonids, as in the spotted gar, indicating losses of 3R-and 4R-duplicated paralogs (Fig. 2). With regards to tshβ, synteny analysis confirmed that tshβa and tshβb paralogs arose from the teleost 3R, as shown in the pike (Fig. 2). Concerning tshβa, synteny analysis assessed that salmonid 4R further gave rise to tshβaα and tshβaβ paralogs, both conserved in the rainbow trout, while only tshβaα was conserved in the Atlantic salmon. Using Oxford Nanopore long reads from the on-going sequencing project of the genome from a Loire-Allier Atlantic salmon, we were able to further confirm the pseudogenization of tshβaβ. Long reads spanning the neighbouring genes of tshβaβ confirmed the lack of tshβaβ exon 1 and the presence only of tshβaβ exon 2 including the deletion leading www.nature.com/scientificreports www.nature.com/scientificreports/ to an early stop codon. For simplicity, in this study we named the tshβaα paralog conserved in the Atlantic salmon, tshβa. Concerning the 4R-duplicated tshβb paralogs, synteny supported that only one paralog was conserved in the rainbow trout and Atlantic salmon, named here tshβb. Differential tissue distribution of tshβa and tshβb transcripts in the Atlantic salmon. We developed specific qPCRs for each Atlantic salmon tshβa and tshβb paralogs and compared the tissue distribution of their expression in smolts (Fig. 3). Both tshβa and tshβb paralogs were mainly expressed in the pituitary. While salmon tshβa transcript was exclusively found in the pituitary, tshβb was expressed also at low levels in various brain regions, and at lower but detectable levels in some peripheral tissues such as gills, kidney, liver, muscle, fat and gonads; tshβb transcripts were not detectable in heart, spleen and skin (Fig. 3).
Evidence for the parr-smolt transformation during the experimental periods. Three independent experiments were performed in 2013, 2014 and 2016 where fish were sampled from December/January until June in order to cover the smoltification period that occurs in early spring for the Loire-Allier Atlantic salmon population 26 . Smoltification is classically described through typical morphological, behavioural and physiological changes. These series of changes were used to assess the occurrence of the smoltification.
Morphological changes. All juvenile salmon had visible parr marks at the first sampling time which progressively regressed throughout the parr-smolt transformation. Conversely, body silvering increased and darkening of the pectoral fins took place throughout the smoltification period. Photos were taken of each fish sampled in 2016 and representative photos are displayed in Supplementary Fig. S3. These changes are characteristic of the parr-smolt transformation (see review 3 ).
Behavioural changes. Rheotactic behaviour, positive or negative, was observed every day during the daylight hours throughout the experimentations. All parr exhibited positive rheotaxis at the beginning of the experiments and maintained this behaviour until the end of March. The inversion of rheotaxis from positive to negative, typical of smoltification and which triggers the onset of downstream migration, was observed during early April in all experiments. All fish had inverted to negative rheotaxis by April 4 (2013), April 9 (2014) or April 8 (2016). They maintained negative rheotaxis until the end of the experimental period (end of June). This inversion timing in early April is in agreement with the previous reports for the Loire-Allier population 26 .
Physiological changes. Gill Na + , K + -ATPase (NKA) activity was measured in fish sampled in 2013 and 2014 (Fig. 4). Gill NKA activity was low in parr (February), increased from March to reach a peak in smolts in April/ May and decreased in post-smolts in June. The increase of gill NKA activity is a typical physiological characteristic of smoltification which prepares smolts to transition from fresh water to sea water (for review 3   Tshβ genomic region of a non-teleost actinopterygian, a holostean, the spotted gar (Lepisosteus oculatus) was used as a reference. Tshβ genomic region was duplicated by teleost-specific whole genome duplication (3R) as seen in the pike (Esox lucius), resulting in two paralogons (a and b), and further duplicated by salmonid-specific whole genome duplication (4R) as seen in the Atlantic salmon (Salmo salar) and rainbow trout (Oncorhynchus mykiss), resulting in four paralogons (aα, aβ, bα, bβ). For each species, chromosome or scaffold number is indicated. Gene positions are given (in Mega base) below the genes. Full names and references of tshβ and neighbouring genes are given in Supplementary  Table S2. Red cross indicates loss of tshβ paralog; dotted red cross indicates tshβ pseudogene.
www.nature.com/scientificreports www.nature.com/scientificreports/ expression profile of tshβa remained relatively constant throughout the sampling period, a dramatic peak in the expression of the other paralog, tshβb, was measured in March-April, during the smoltification period (Fig. 4). Notably, the tshβb expression peak was concomitant with the timing of rheotaxis inversion in both experiments (Fig. 4).
In order to confirm the differential regulation of the pituitary expression of tshβa and b paralogs, we performed a third experiment from December 2015 to June 2016, with high frequency sampling during the smoltification period (Fig. 5). The results were in full agreement with 2013 and 2014 experiments. Pituitary expression of tshβa remained stable throughout the experiment, while a large expression peak of tshβb was recorded during the smoltification period. Pituitary tshβb transcript levels started to rise during February, reached a dramatic peak in early April and then dropped at the end of April until June, to reach lower levels than in December (Fig. 5). It is noteworthy that qPCR results (Figs 4 and 5) were expressed as arbitrary units for each paralog; however comparison of quantification cycle values (Cq) between paralogs suggested lower pituitary levels of tshβb than tshβa at the beginning (December/February) and end (June) of the experiments (8 Cq mean difference), and even in April at the time of tshβb peak (4 Cq mean difference). We observed again that the change in rheotaxis from positive to negative, a characteristic of smoltification, was concomitant with the expression peak of tshβb paralog (Figs 4 and 5).

Distinct populations of tshβa-and tshβb-expressing cells in the Atlantic salmon pituitary.
In order to get more insight into the pituitary expression of the two paralogs, we compared the localization of tshβa Owing to the different nature of the tissues, transcripts levels were normalized to the amount of total RNA, and expressed as arbitrary units. Results are means ± s.e.m (n = 5 females for ovary; n = 5 males for testis; n = 10 mixed sex for the other tissues as there was no differences between sex). www.nature.com/scientificreports www.nature.com/scientificreports/ and tshβb transcripts by fluorescent in situ hybridization (FISH), using pituitaries from parr sampled in October and November, smolts sampled in April and post-smolts in June. Tshβa-expressing cells could be detected by FISH in parr and in smolts at all sampling times, while tshβb-expressing cells could be observed only in April. The lack of detection of tshβb-expressing cells by FISH in parr and in post-smolts is in agreement with the very low expression of this paralog outside of the smoltification period, as shown by qPCR. Tshβa cells were located in the anterior adenohypophysis, in the rostral pars distalis (RPD) close to the follicles formed by the prolactin cells, at the border with the proximal pars distalis (PPD) (Fig. 6). In contrast, tshβb cells were observed in the dorsal region of the PPD, bordering the dorsal region of the pars nervosa (PN) near the pituitary stalk (Fig. 6). No labelling was observed when using control sense probes ( Supplementary Fig. S4). FISH results demonstrated that salmon tshβa and tshβb paralogs are expressed by two distinct cell populations within the pituitary.

Discussion
The present study revealed that two paralogous genes, named tshβa and tshβb based on phylogeny and synteny analyses, are expressed in the Atlantic salmon. The deduced protein sequences TSHβa corresponded to the subunit previously characterized in Atlantic salmon 19 , while TSHβb corresponded to a novel not yet investigated subunit. A third homologous gene sequence was identified as a pseudogene. Atlantic salmon TSHβa and b sequences shared the 12 cysteine residues and the N-glycosylation site conserved among vertebrate TSHβ. Yet, they were largely different with only 47.5% amino-acid similarity reflecting their divergence since the teleost 3R genome duplication event. Differently to TSHβa, Atlantic salmon TSHβb possessed a second glycosylation site, as previously observed for other teleost TSHβb 21 , which may confer differential biological properties. Recent work in mice has shown that tissue specific glycosylation of TSH, produced in pars distalis versus in pars tuberalis, induces differential bioactivity in the blood 27,28 .
Phylogeny and synteny analyses brought new advances on the evolutionary scenario of tshβ genes, as illustrated in Fig. 7. As previously indicated 21 , the duplicated paralogs tshβ and tshβ2 originated from vertebrate 2R, but only tshβ was conserved in the actinopterygian lineage as observed in an extant holostean, the spotted gar.  www.nature.com/scientificreports www.nature.com/scientificreports/ Teleost 3R duplicated tshβ into tshβa and tshβb, both conserved in various extant teleosts. Salmonid 4R further duplicated tshβa into tshβaα and tshβaβ paralogs, both conserved in Oncorhynchus species; however, tshβaβ is undergoing a loss in Atlantic salmon, where it was detected as a pseudogene. We confirmed the tshβaβ pseudogene sequence using representatives from two wild Atlantic salmon populations (Loire-Allier, France and Imsa, Norway). This indicates that tshβaβ loss is a common feature among Atlantic salmon and does not represent a divergence related to domestication of the strain 24 used for the published genome. Salmonid 4R would have also duplicated tshβb into two paralogs, but a single tshβb gene has been conserved in the Oncorhynchus and Salmo species investigated in this study; this suggests that one of the tshβ 4R paralogs would have been lost in the salmonid lineage, shortly after the 4R, before the divergence between Oncorhynchus and Salmo species (Fig. 7).
Development of specific qPCR for Atlantic salmon tshβa and tshβb allowed us to show that both genes are mainly expressed in the pituitary, and suggested that tshβa transcript levels are higher than tshβb transcript levels. The expression of both tshβa and tshβb paralogs in the pituitary has been previously reported in two other teleost species, the stickleback, Gasterosteus aculeatus (named tshβ1 and tshβ2 by the authors 22 ) and the European eel, Anguilla anguilla (named tshβ and tshβ3 by the authors 21 ).
As measured by qPCR in the Atlantic salmon, tshβa was exclusively expressed in the pituitary, while tshβb was also expressed at lower levels in different brain regions, and in various peripheral tissues. A similar result was previously obtained in the European eel, with a pituitary-only expression of tshβa and a more ubiquitous tissue distribution of tshβb (tshβ and tshβ3 respectively 21 ). In amphibians, reptiles and birds, which possess a single tshβ gene, most studies have reported the expression of tshβ mRNA only in the pituitary (chicken 29 ; bullfrog 30 ; quail 31 ; turtle 32 ; duck 33,34 ) while investigating a wide range of other tissues (brain, gonads, liver, thyroid, muscle, lung, heart, intestine, kidney and spleen). However in mammals, which also possess a single tshβ mainly expressed in the pituitary, tshβ transcript or TSH protein were reported at low levels in the brain 35 and in some peripheral tissues, such as myometrium and skin 36 .
We further investigated by FISH the respective localization of tshβa and tshβb in the Atlantic salmon pituitary and revealed that the paralogs were expressed by distinct pituitary cell populations. Numerous tshβa-expressing cells were located at the border between the RPD and antero-ventral PPD, while fewer tshβb-expressing cells were observed in the dorsal PPD close to the pituitary stalk. Early histological works in salmonids using radiothyroidectomy already reported the localization of TSH cells mainly in the RPD at the junction with PPD (Atlantic salmon 37 ; chinook salmon, Oncorhynchus tshawytscha 38 ). Immunocytochemical studies, using an antibody against human TSHβ, revealed TSH cells in the ventral PPD adjacent to the RPD (chum salmon, Oncorhynchus keta and rainbow trout 39,40 ). A similar localization was observed using an antibody raised against purified coho salmon TSH (rainbow trout 41 ; chinook salmon 42 ). In light of our present study, these previous investigations likely observed the localization of the abundant TSHβa.
Two distinct populations of TSH cells have also been described in the pituitary of birds and mammals, but both expressing the same single gene present in tetrapods (tshβ). The "classical" TSH cell population is located in the pars distalis (PD), while a less numerous TSH cell population is located in the pars tuberalis (PT) which surrounds the pituitary stalk (quail 43 ; Soy sheep 44 ; mice 45 ; European hamster 46 ). We may relate the localization of salmon tshβa-and tshβb-expressing cells to that of amniote PD-and PT-tshβ cells, respectively, and infer that the Tshβa and tshβb arose from the teleost-specific 3 rd round of whole genome duplication (3R) and were further duplicated by the salmonid-specific 4 th round of whole genome duplication (4R). One of the 4R paralogs , tshβbβ, was lost early in the salmonid lineage, before the divergence between the Oncorhynchus and Salmo lineages. Another 4R paralog, tshβaβ, still present in the rainbow trout (Oncorhynchus mykiss) and coho salmon ( Oncorhynchus kisutch), is undergoing a loss in Atlantic salmon (Salmo salar) where it is detected as a pseudogene.
www.nature.com/scientificreports www.nature.com/scientificreports/ specific expression of tshβa and tshβb in these distinct pituitary cell populations would represent a typical case of subfunctionalization of duplicated paralogs.
The potential involvement of tshβa and tshβb paralogs in smoltification was investigated by measuring their pituitary expression profiles in juvenile Atlantic salmon from the Loire-Allier basin. Three independent sampling experiments were performed in 2013, 2014 and 2016 at CNSS. Classical smoltification-related changes such as body colouration, rheotaxis inversion and increased gill NKA activity were observed which indicate that these fish have undergone complete smolt development. Remarkably, a striking peak in the expression of tshβb, with no change in tshβa, was recorded in April, at the period of smoltification, in each yearly experiment. This is the first demonstration of a surge in pituitary tshβ in relation to smoltification metamorphosis in salmonids. In contrast, previous studies reported slightly lower 11,19 or no change 20 in pituitary tshβ transcript levels during smoltification; these investigations were in fact targeting the tshβa paralog, the expression of which remains relatively stable as shown in the present study. The demonstration of a differential regulation of the pituitary expression of salmon tshβa and tshβb paralogs, with a specific peak of tshβb, revealed a marked functional divergence of the two paralogs, conferring a specific role in smoltification to tshβb paralog.
In birds and mammals, TSH produced by PT cells plays a key role in the seasonal regulation of major steps of life cycles, including reproduction, migration and hibernation 44,47,48 . PT-TSH stimulates the expression of type 2 deiodinase (DIO2), which catalyses the conversion of T4 into the more biologically active T3, thus leading to the activation of TH-regulated brain functions (Fig. 8). In a recent study 49 , a salmonid specific 4R-issued DIO2 paralog (dio2b) has been identified in the Atlantic salmon, the expression of which increases in circumventricular brain area of cell proliferation, during experimental photoperiod-induced smoltification. The authors proposed a specific role of DIO2b in promotion of TH-dependent brain development during smoltification 49 .
In the line of PT-TSH role and action mechanism in amniotes, we propose that TSHb produced by dorsal PPD cells in Atlantic salmon may stimulate brain DIO2b expression and promote TH-activated brain functions related to smoltification (Fig. 8). As a support to this hypothesis, in our three yearly experiments, the expression peak of tshβb paralog occurred simultaneously with smoltification-related changes in rheotaxis, which triggers the onset of downstream migration.
In the Atlantic salmon, tshβb was expressed not only in the pituitary but also at lower levels in various brain regions, so that additional regulatory pathways may also occur locally in some brain regions. Such a local signalling between light sensors, TSH and DIO2, has been proposed in the saccus vasculosus for the photoperiodic regulation of reproduction in masu salmon 50,51 . In our study, tshβb transcripts were detectable in the saccus vasculosus of the Atlantic salmon smolt but were at least 100 times less expressed than in the pituitary; tshβa transcripts  43,44,47,48 by increasing brain deiodinase 2 (DIO2) and stimulating the conversion of thyroxin (T4) into triiodothyronine (T3). Salmon tshβa and tshβb paralogs are expressed by distinct cell populations as shown by FISH. We propose that tshβb cells located in the dorsal region of the pars proximal distalis (PPD) may be related to the PT-TSH cells in birds and mammals, and infer a subfunctionalization of salmon tshβa and tshβb paralogs. We revealed a peak expression of tshβ at smoltification, concomitant to the inversion of rheotaxis initiating downstream migration. An increase in brain dio2b paralog was reported at smoltification in Atlantic salmon 49  www.nature.com/scientificreports www.nature.com/scientificreports/ were not detectable outside the pituitary. The expression of tshβb paralog, not only in the pituitary, but in various brain regions and peripheral tissues observed in the present study warrants future investigations on the possible variations of tshβb transcripts in these regions during smoltification.
In mammals, PT-TSH has been suggested to act not only on the basal hypothalamus but also to exert some paracrine effects in the pituitary 52 , so that we may also hypothesize pituitary actions of tshβb paralog during salmon smoltification. Using a basal teleost, the European eel, as a model, we previously identified 3R-duplicated TSH receptor (TSHR) paralogs and revealed the specific expression of tshrb paralog in various brain regions as well as in the pituitary, supporting brain and pituitary actions of TSH in teleosts 21 . Further studies should aim at investigating TSH receptor paralog number and tissue distribution in the Atlantic salmon.
In the present study, we propose a specific role of Atlantic salmon tshβb paralog at smoltification, related to the change in rheotaxis and triggering of downstream migration. The remarkable expression peak of pituitary tshβb demonstrated in our study may be of high relevance for deciphering the internal and environmental regulation of salmonid smoltification and initiation of downstream migration; especially concerning the endangered long-river Loire-Allier salmon population, understanding these mechanisms may provide new basis for its conservation. A specific role in migration of tshβb paralog (named tshβ2 by the authors 22 ) was also suggested in the stickleback 22 , based on higher pituitary transcript levels of tshβb in populations migrating to the sea, as compared to stream-resident populations, with no difference in tshβa paralog (named tshβ1 by the authors 22 ) between the two ecotypes. The authors suggested that genetic differences in cis-regulatory regions of tshβb gene may convey this adaptive divergence in migratory behaviour during stickleback radiation 22,53 . Similarly, there is large diversity of migration strategies among salmonid species and populations, from long-river anadromous to landlocked. Our present findings open new research avenues for comparing tshβb expression during smoltification, as well as tshβb regulatory genomic sequences, between salmonid species and ecotypes.
In conclusion, two tshβ paralogs are expressed in the Atlantic salmon by distinct pituitary cell populations, and exhibit a striking functional divergence, with a large expression peak of tshβb, but not of tshβa, during smoltification. This is the first demonstration of a peak expression of tshβ in salmonid smoltification. This involvement of the thyrotropic axis complies with the endocrine regulation of vertebrate metamorphosis typically observed in amphibian and flatfish. A specific role of tshβb paralog is suggested in the onset of smoltification-related downstream migratory behaviour, possibly mediated by the stimulation of brain DIO2 and T3 production, as shown for PT-TSH involved in the seasonal regulation of life cycle traits in birds and mammals. The remarkable functional divergence of tshβa and tshβb in salmon may have represented selective forces for the conservation of these duplicated paralogs.

Methods
Fish. The study was carried on juvenile Atlantic salmon (Salmo salar) of the Loire-Allier population raised indoor under natural water, temperature, and photoperiod conditions, at the Conservatoire National du Saumon Sauvage (CNSS), Chanteuges, France (Agreement N° B43 056 005; according to the ARRETE N° DDCSPP/ CS/2016/40), which breeds wild returning adult Atlantic salmon genitors caught at the Vichy dam, 620 km from the Loire estuary. The research project was performed in accordance with guidelines and regulations according to the protocol approved by Cuvier Ethic Committee France.
For each experimental year, 390 under-yearling fish were transferred in December into two circular tanks (3 m diameter; depth range 0.5 m) supplied with UV filtered natural running water from the Desges River (tributary of the Allier). Experiments were conducted from December to end of June, which spans the smoltification period that occurs in early spring for the Loire-Allier Atlantic salmon population 26 . Three independent experiments were performed (2013, 2014 and 2016). Water temperature was measured using probes (Johnson control, Colombes, France; TS 9101: accuracy ±0.2 °C). An anti-clockwise flow was achieved by a tangentially oriented water inlet at the periphery of the tank and a central drain as previously described 26 . Photoperiod regime mimicked the natural photoperiod by using an outside light sensor that controlled the light above each tank. Each tank had a LedBulb (D 14-75 W E27 827 A67; Philips, Amsterdam, Netherlands) 3 m above the water surface. Fish were fed automatically with a custom fish diet (Turbot label Rouge, Le Gouessant, Lamballe, France) in excess five times a day at equal intervals during daylight hours. Fish swimming behaviour was visually observed during daytime: positive rheotaxis for fish facing the water current versus negative rheotaxis for fish swimming with the current 26 . tissue collection. Fish were anesthetized with an overdose of ms222 (0.4 ml/l; Sigma-Aldrich, St Louis, MI, USA). Photos were taken with a Canon EOS 1200D Digital SLR Camera with EF-S 18-55 mm f/3.5-5.6 III Lens in order to follow colouration changes, characteristics of smoltification. Fish were killed by decapitation. For qPCR analyses of the tissue distribution of tshβa and tshβb transcripts, 10 fish (5 males and 5 females) were sampled in March 2015; the following organs were individually collected and stored in RNALater (Ambion Inc, Austin, USA) at −20 °C until RNA extraction: retina, brain (dissected into olfactory bulbs, telencephalon, epiphysis, optic lobes, hypothalamus, saccus vasculosus, cerebellum, medulla oblongata), pituitary, as well as samples of gill filaments, kidney, liver, spleen, muscle, skin, abdominal fat, testis or ovary. www.nature.com/scientificreports www.nature.com/scientificreports/ Measurement of Na + , K + -ATPase activity. Gill samples of 10 fish per sampling times were collected in experiments 2013 and 2014. For each fish, four to six primary gill filaments were placed into 100 µl of ice-cold SEI buffer (250 mM sucrose, 10 mM EDTA, 50 mM imidazole, pH 7.3) and frozen at −80 °C for measurement of gill Na + /K + -ATPase (NKA) activity. NKA activity was determined with a kinetic assay run in 96-well microplates at 25 °C and read at a wavelength of 340 nm for 10 min as described previously 54 . Gill tissue was homogenized in 150 µl of SEID (SEI buffer and 0.1% deoxycholic acid) and centrifuged at 5000 × g for 30s. Two sets of duplicate 10 µl samples were run, one set containing assay mixture and the other assay mixture and 0.5 mM ouabain. The resulting ouabain-sensitive ATPase activity is expressed as µmoles ADP mg protein −1 h −1 . Protein concentrations are determined using BCA (bicinchoninic acid) Protein Assay (Pierce, www.piercenet.com, Rockford, Il, USA). Both assays were run on a THERMOmax microplate reader using SOFTmax software (Molecular Devices, www. moleculardevices.com, Menlo Park, CA, USA).  Table S3). PCR was performed using cDNA of smolt pituitaries collected in April using Taq DNA polymerase (Invitrogen, Carlsbad, CA, USA) for tshβa and Platinum Taq DNA polymerase (Invitrogen) for tshβb according to the manufacturer's instructions. Purified PCR fragments were subcloned into PCRII vectors (Thermo-Fisher, Wahtham, MA, USA) before sequencing by GATC Biotech (Brussels, Belgium). Plasmids were used for preparing cRNA probes (section fluorescence in situ hybridization).

Identification of tshβ
Cloning and sequencing of genomic DNA sequence of the second Atlantic salmon tshβa. Cloning primers for the Atlantic salmon tshβaβ gene were designed, based on the alignment of salmonids tshβaβ, on both sides of the coding sequence of the functional salmonid tshβaβ using Primer3 (Supplementary Table S3, Supplementary Figure x). Hatchery-produced Atlantic salmon from wild caught broodfish from the Imsa River (Norway) were used. Genomic DNA from testis samples was extracted by alkaline lysis. Testis pieces were heated in alkaline buffer (25 mM NaOH, 0.2 mM EDTA) at 95 °C for 5 min with vortexing. Reaction was stopped by adding the neutralisation buffer (40 mm Tris-HCL). PCR was performed using Atlantic salmon genomic DNA, DNA polymerase (Invitrogen), according to the manufacturer's instructions with a thermal cycle protocol including a 10 cycle touchdown phase (from 60 to 50 °C) followed by 40 cycles with annealing temperature at 50 °C. Purified PCR fragments were subcloned into PCRII vectors (Thermo-Fisher) before sequencing by GATC Biotech.
Illumina libraries and sequencing. Genomic DNA (gDNA) was extracted from snap-frozen liver from juvenile Loire-Allier Atlantic salmon using the Genomic-tip 100/G (Qiagen). DNA was subsequently sheared using a nebulizer (Life Technologies). Paired-end libraries were prepared from 5 μg of sheared DNA using the Paired-End Sequencing Sample Prep kit (Illumina Inc., San Diego, USA). For the library size selection step, the 400 bp band was cut from the agarose gel, purified and amplified by 10 PCR cycles. The resulting library was analyzed with a Bioanalyzer 2100 DNA 1000 series II chip (Agilent, Santa Clara, USA). All libraries were sequenced using an Illumina HiSeq2500 instrument with a read length of 2 × 151 nucleotides to a total of ~103.75 Gb of sequencing data and up to ~686.5 million reads. Nanopore libraries and sequencing. Genomic DNA (gDNA) was extracted from snap-frozen liver, spleen and kidney from a juvenile Loire-Allier Atlantic salmon using the Genomic tip-100/G (Qiagen). gDNA size was inspected in the TapeStation Genomic DNA system (Agilent) and found to be >60 Kb. The gDNA was subsequently sheared to 10-20 Kb fragments using a g-tube (Covaris, Woburn, MA) before library preparation. The library preparation was performed using 1D Genomic DNA by ligation for either SQK-LSK108 or SQK-LSK109 kit (Oxford Nanopore technologies, Oxford, UK). All nanopore libraries were sequenced in a FLO-MIN106 R9.4.1 SpotON Flow Cell attached to either MinION or GridION devices (Oxford Nanopore Technologies), generating 28.3 Gb of nanopore sequencing data divided over ~3.21 million reads.
Alignment of Illumina and nanopore reads to tshβ paralogs. Illumina reads were aligned to the tshb paralogs using bowtie2 (version 2.2.5) 56 , while nanopore reads were aligned using minimap2 (version 2.5-r572) 57 with the default Oxford nanopore parameters (map-ont). SAMtools (version 1.2) 58 was used to remove unmapped reads from the SAM file, convert to BAM, sort and generate the index. The resulting BAM files were visualized using Integrative Genomics Viewer (IGV) (version 2.3.83) 59 . phylogeny analysis. Phylogeny analysis of 39 vertebrate TSHβ amino-acid sequences was performed using a part of dataset from 21 , enriched with additional teleost TSHβ sequences, including TSHβ paralogs of salmonids identified in this study. New tshβ genes were either retrieved from GenBank or were identified by blasting (TBLASTN algorithm) genome assembly databases when genes were not annotated in GenBank. The amino-acid (2019) 9:4561 | https://doi.org/10.1038/s41598-019-40019-5 www.nature.com/scientificreports www.nature.com/scientificreports/ sequences of TSHβ were deduced and signal peptides were predicted using CLC Main Workbench 8 (Qiagen). The sequence alignment was performed on CLC Main workbench 8 and manually adjusted. Phylogenetic tree was constructed using Maximum Likelihood algorithm with PhyML:3.0 60 combined to the SMS model selection 61 and SPR as tree improvement on ATGC browser (http://www.atgc-montpellier.fr/phyml/). Tree topology was assessed by bootstrapping on 1000 replicates. synteny analysis. Synteny analysis was performed on tshβ genomic region in actinopterygians, using a holostean, the spotted gar as a reference (LepOcu1 (GCA_000242695.1)). Comparisons were made with tshβ paralogons in the pike (Eluc_V3 (GCA_000721915.3)), and in two salmonid representatives, Atlantic salmon (ICSASG_v2 (GCA_000233375.4)) and rainbow trout (Omyk_1.0 (GCA_002163495.1)). Neighbouring genes of tshβ loci were identified and compared manually using chromosome annotation. Blast analyses on the genomes were performed to search for un-annotated genes and additional paralogs. Genes fractionated, showing a frameshift mutation or missing exon were considered pseudogenes.
RNA extraction and cDNA synthesis. Total RNA was extracted by homogenizing tissues in TRIzol (Thermo-Fisher) according to the manufacturer's protocol, using TissueLyser II (Qiagen). After a chloroform separation step, RNA was precipitated in ice cold isopropanol with 1 µl of glycoblue (Ambion). Total RNA was treated with DNase I (Roche Diagnostics, Basel, Switzerland) according to the manufacturer's instructions. RNA concentration was measured using Nanopore 2000c/2000 (Thermo-Fisher). Reverse transcription was performed using 75 ng random hexamer primers (Invitrogen) and SuperScriptIII First Strand cDNA Synthesis Kit (Invitrogen) following the manufacturer's protocol. For pituitaries, 250 ng of total RNA were used and 750 ng for brain and peripheral tissues.
Quantitative Rt-pCR. Specific quantitative real-time PCR (qPCR) primers for tshβa and tshβb were designed using Primer3 (http://bioinfo.ut.ee/primer3-0.4.0/) 55 with forward and reverse primers on two different exons to prevent amplification of genomic DNA (Supplementary Table S3); primers were purchased from Eurofins scientific (Luxembourg). Specificity of the primers was controlled by sequencing PCR product. β-actin was used as reference gene using previously published primers 62 . Quantitative PCR assays were performed using LightCycler 1.2 (Roche Diagnostics) and LightCycler FastStart Master plus SYBR Green I kit (Roche Diagnostics). Each reaction contained: 4 µl of diluted cDNA template, 2 µl of SYBR green master mix and 1 µl of specific primers (500 nM final concentration). The following thermal cycling steps were used for each qPCR run: initial denaturation 94 °C for 10 min followed by 41 cycles of 10 s of denaturation at 95 °C, 5 s of annealing temperature (60 °C or 62 °C; Supplementary Table S3) and 6 s of elongation at 72 °C. The program ended by slowly increasing temperatures (0.1 °C/s) from 68-95 °C for amplification specificity controlled by melting curve analysis. Relative quantification was performed using standard curves created for each gene with serial dilutions of pooled pituitary cDNA. One dilution of the cDNA pool was added in each run as a calibrator. All samples were analysed in duplicates and each qPCR run contained a negative control using water in substitute for template cDNA. Calculations of sample concentrations were made using the Roche LightCycler 1.2 manufacturer's software.
Fluorescence in situ hybridization (FIsH). Antisense and sense cRNA probes for FISH were synthesized by in vitro transcription from tshβa or tshβb plasmids using T7 and SP6 RNA polymerase (Promega, Madison, Wisconsin USA) and labelled with digoxigenin-11 UTP (Roche Diagnostics) at 37 °C for 2 h. Probes were purified using Nucleospin RNA clean-UP kit (Machere-Nagal, Hoerdt, France) and controlled by gel electrophoresis. Probe length for tshβa was 401 bp which nearly covers the 420 bp CDS sequence. Probe length for tshβb was 512 bp which covers the entire CDS sequence (Supplementary Fig. S5).