Demonstration of the Coexistence of Duplicated LH Receptors in Teleosts, and Their Origin in Ancestral Actinopterygians

Pituitary gonadotropins, FSH and LH, control gonad activity in vertebrates, via binding to their respective receptors, FSHR and LHR, members of GPCR superfamily. Until recently, it was accepted that gnathostomes possess a single FSHR and a single LHR, encoded by fshr and lhcgr genes. We reinvestigated this question, focusing on vertebrate species of key-phylogenetical positions. Genome analyses supported the presence of a single fshr and a single lhcgr in chondrichthyans, and in sarcopterygians including mammals, birds, amphibians and coelacanth. In contrast, we identified a single fshr but two lhgcr in basal teleosts, the eels. We further showed the coexistence of duplicated lhgcr in other actinopterygians, including a non-teleost, the gar, and other teleosts, e.g. Mexican tetra, platyfish, or tilapia. Phylogeny and synteny analyses supported the existence in actinopterygians of two lhgcr paralogs (lhgcr1/ lhgcr2), which do not result from the teleost-specific whole-genome duplication (3R), but likely from a local gene duplication that occurred early in the actinopterygian lineage. Due to gene losses, there was no impact of 3R on the number of gonadotropin receptors in extant teleosts. Additional gene losses during teleost radiation, led to a single lhgcr (lhgcr1 or lhgcr2) in some species, e.g. medaka and zebrafish. Sequence comparison highlighted divergences in the extracellular and intracellular domains of the duplicated lhgcr, suggesting differential properties such as ligand binding and activation mechanisms. Comparison of tissue distribution in the European eel, revealed that fshr and both lhgcr transcripts are expressed in the ovary and testis, but are differentially expressed in non-gonadal tissues such as brain or eye. Differences in structure-activity relationships and tissue expression may have contributed as selective drives in the conservation of the duplicated lhgcr. This study revises the evolutionary scenario and nomenclature of gonadotropin receptors, and opens new research avenues on the roles of duplicated LHR in actinopterygians.


Introduction
In teleost species, as in other vertebrates, reproduction is controlled by the brain-pituitarygonad axis. The pituitary gonadotropins, follicle stimulating hormone (FSH) and luteinizing hormone (LH), are key hormones in the regulation of gametogenesis and steroidogenesis in both males and females [1]. Together with the pituitary thyrotropin (TSH), and the chorionic gonadotropin (CG) in primates, the gonadotropins are heterodimeric glycoproteins, composed of a common alpha subunit and a beta subunit that confers the biological specificity of the hormone [1][2]. After their characterization in mammals, LH and FSH have been demonstrated in the other tetrapods, as well as in teleosts and chondrichthyans, indicating the presence of the two gonadotropins in early gnathostomes [3].
Previous phylogeny studies of teleost gonadotropin receptor sequences, indicated that teleost LHR may separate into two clades [20][21]. Comparative genomics has revealed that a whole-genome duplication event (named 3R, for third round of genome duplication) occurred specifically in the teleost lineage [22]. The two LHR types may thus have resulted from the teleost 3R. As only one or the other of the two LHR types was found in given extant teleost species [21], this may suggest that the 3R would have been followed by alternative losses of one or the other lhcgr duplicated gene. Based on the finding that teleost dual lhcgr were mutually exclusive, and located on the same locus, another hypothesis was also proposed, that they may not result from gene duplication, but from interallelic conversion [21].
Our studies of the draft genomes of the European eel (Anguilla anguilla) and Japanese eel (Anguilla japonica), representative species of a basal group among teleosts (Elopomorphs) [23], revealed a larger conservation of duplicated genes in the eels, as compared to other teleost species. This is illustrated by a larger number of hox genes in the eels than in the other, more recently emerged, teleosts [24][25]. This also applies to genes involved in the neuro-endocrine systems. For instance, eels are so far the only teleost species to have conserved up to three kiss receptor genes [26]. Therefore in the present study, we investigated whether the eels may have conserved putative duplicated genes for gonadotropin receptors.
Previous studies in the Japanese eel, reported the partial cDNA cloning of a single fshr and a single lhcgr (lhr) [27], as well as the full-length cloning of cDNA encoding a single functional FSHR and a single functional LHR, respectively [28][29][30]. In the European eel, one full-length cDNA encoding a functional FSHR was also cloned and characterized [31]. Japanese eel fshr and lhcgr transcripts are expressed in female ovary and male testis and show expression changes during induced sexual maturation in both females and males [27,30].
In the present study, we successfully identified within the draft genomes of the European and Japanese eels, a single fshr gene, but two lhcgr genes, named lhcgr1 and lhcgr2, providing the first evidence of the conservation of coexisting duplicated lhcgr in vertebrates. We isolated cDNA sequences of the single fshr and of the duplicated lhcgr1 and lhcgr2 in the European eel. Quantitative PCR (qPCR) revealed differential tissue expression patterns of fshr and of the two lhcgr1 and lhcgr2 mRNA, which might have contributed to the maintenance of the duplicated lhcgr genes in the eels.
Based on our first evidence of two lhcgr in the eel, we reinvestigated the presence and number of gonadotropin receptor genes in other teleost species and found the conservation of two lhcgr in some other extant teleosts. These data open new research avenues for basic and applied fish reproductive endocrinology. To further elucidate the evolutionary history of gonadotropin receptors, we also identified new fshr and lhcgr sequences in genomes of key representative gnathostome species, including two chondrichthyans, the little skate (Leucoraja erinacea), and the elephant shark (Callorhinchus milii), a non-teleost actinopterygian which has diverged before the teleost-specific 3R, the spotted gar (Lepisosteus oculatus), and a representative species of an early diverging lineage among sarcopterygians, the coelacanth (Latimeria chalumnae). Gene prediction, phylogeny and synteny analyses allowed us to propose a new scenario, suggesting that a single copy of fshr gene would have been conserved throughout vertebrate evolution, while duplicated lhcgr might have originated early in the actinopterygian lineage, before the teleost-specific third round of whole-genome duplication.

Ethics statement
Complementary DNA cloning, and transcript tissue distribution studies in the European eel (Anguilla anguilla) were performed using total RNA samples, which had been already isolated and used in our previous studies [26,[32][33], thus avoiding the sacrifice of additional eels, considered as endangered species.

Gene and protein nomenclature
Gene and protein nomenclatures were standardized in the present text for all species by gene symbol in lowercase italic and protein symbol in uppercase non-italic.
Gonadotropin receptors from other teleosts. Using the FSHR and the two LHR protein sequences identified in the European eel in this study as queries (tBlastn), fshr and lhcgr genes were sought and annotated (if not yet predicted) in the increasing number of teleost genomes available in Ensembl or NCBI (see S1 Table).
Gene coding sequences (CDS) were annotated or corrected by comparison with well-known orthologous CDS from other species, in respect with the canonical splice site rule using CLC Main Workbench (S1 Table).

cDNA cloning of European eel gonadotropin receptors
The predicted European eel fshr, lhcgr1 and lhcgr2 CDS sequences were completed and confirmed by cDNA cloning. Specific cloning primers were designed on both extremities and inside the predicted CDS sequences, using Primer3 browser [35] (S2 Table). First strand cDNAs were synthesized from total RNA samples previously prepared from testis and ovary of silver European eels [26,33]. One μg was reverse transcribed using superscript III (Invitrogen, Carlsbad, CA, USA), random primers (50 ng) (Invitrogen) and oligo(dT) 18 (50 μg) (Roche, Mannhein, Germany), at 50°C for 60 min, after an initial step at 25°C for 10 min. PCR were performed using GoTaq mix (Promega, Fitchburg, WI, USA) on ovary and testis cDNAs. PCR amplifications were performed on the Life Express Thermal Cycler (Hangzhou Bioer Technology Co., Ltd, Hangzhou, China) with the following thermal conditions: after an initial step of 95°C for 10 min, five cycles at 95°C for 30 sec, annealing temperature ranging from 65°C to 62°C for 30 sec, 72°C for 2 min, followed by 30-40 cycles with annealing temperature ranging from 55 to 53°C for 30 sec, and a final extension at 72°C for 5 min. Lhcgr1 cDNA was cloned from the ovary, fshr and lhcgr2 cDNAs from the testis. For lhcgr2 cloning, cDNA templates were previously denaturated during 10 min at 95°C. Amplified products were separated on agarose gel and purified using QIAquick gel extraction kit (Qiagen, Vento, Netherlands). Purified PCR fragments were sequenced directly, or after subcloning into a pGEM-T Easy Vector (Promega), by the sequencing service of GATC Biotech (Constance, Germany).

Prediction of gonadotropin receptor protein sequences and domains
Nucleotide sequence translation and first methionine were predicted using the package CLC Main Workbench. The signal peptide cleavage site was determined using SignalP [36] and putative transmembrane helices were predicted using TOPCONS [37]. Searches for motifs and sequence patterns were performed by comparison with protein families databases on ExPASy server [38]. Leucine-rich repeats were identified using Pfam database [39] and the recent crystallography study of human FSHR extracellular domain [40][41]. Potential tyrosine sulfation sites were predicted by sulfosite browser with 80% prediction sensitivity [42] and palmitoylation sites in the ICD were predicted using PalmPred browser [43].

Phylogeny analysis and domain comparison of gonadotropin receptors
Seventy eight gonadotropin receptor sequences of representative vertebrates were collected from the available databases of NCBI and Ensembl, including 33 sequences defined in the present study and 45 sequences previously published (S1 Table). Multiple alignments were performed using ClutstalW and were manually adjusted taking into account the receptor topology to improve the alignment accuracy. The phylogeny of gonadotropin receptors was inferred by Maximum Likelihood using PhyML with the WAG model of substitution [44] combined to the neighbor-joining interchange (NNI) method on Seaview browser [45]. A tree was generated and robustness of the phylogeny assumption was evaluated by the approximate likelihood test (aLRT) SH-like branch support and by bootstrapping procedure from 500 data set replicates. The sequence of the glycoprotein hormone receptor of an early chordate, the lancelet (Branchiostoma floridae), (XP_002610242) was designated to root the tree (S1 Table).

Synteny analysis of gonadotropin receptors
Flanking genes of the three gonadotropin receptors in European and Japanese eel draft genomes were sought by Blast searches. Only a limited number of flanking genes could be annotated due to the current size of the scaffolds. The Spotted gar, a non-teleost actinopterygian, which we showed to possess fshr, lhcgr1 and lhcgr2, was chosen as reference to study the genomic regions of the gonadotropin receptors. At the beginning of our study, the spotted gar genome was assembled at the level of chromosome but was non-annotated. The flanking genes of the spotted gar gonadotropin receptors were identified by blasting the contigs (AHAT01001660 to AHAT01001713) against the zebrafish protein dataset, using CLC Main Workbench. In the course of our study, an annotated version of the spotted gar genome became available in Ensembl and NCBI and was in agreement with our gene synteny predictions.
In addition to the spotted gar, fshr, lhcgr1 and lhcgr2 syntenic regions were compared using Genomicus (version 7.5) [46], Ensembl and NCBI genome visualization browser, in teleost species that have retained either one or both duplicated lhcgr1 and lhcgr2, in representative tetrapods, including human, and in a basal sarcopterygian, the coelacanth.

Tissue distribution of gonadotropin receptors transcript levels in the European eel
RNA samples and cDNA synthesis. Tissue distribution analysis was performed on RNA samples from 8 female silver migrating eels caught in the Loire River in France and previously prepared [32][33]. Total RNA extracted from ovary, olfactory bulb, mesencephalon and diencephalon, telencephalon, cerebellum, medulla oblongata, pituitary, eyes, liver, intestine, muscle, adipose tissue, gills, and thyroid follicles were used. Previously prepared RNA samples from testis [26] from 8 male silver eels were also analyzed. The silver stage is the last continental phase of the eel life cycle. The silver eels are ready to perform the oceanic reproductive migration, but are still immature, the ovary being at the pre-vitellogenic stage, and the testis at the spermatogonial stage. Four hundred ng of total RNA were reverse transcribed using Superscript III (Invitrogen) and random primers (50 ng) (Invitrogen), at 50°C for 60 min after an initial step of 25°C for 10 min.
Quantitative PCR primer design and analyses. Eel specific primer sets for each gonadotropin receptor were designed using Primer3 [35] on two successive exons, or on exon junctions, of the extracellular domain. Various primer set combinations were tested and compared. Efficiency and amplification specificity were checked for each primer set. The sets with the higher efficiency (100% for fshr and lhcgr1 and 87% for lhcgr2) were chosen for the following quantification experiment (S2 Table). Quantitative PCR (qPCR) product was sequenced to control the primer specificity.
Messenger RNAs were assayed using Light Cycler System (Roche) with the LightCycler Fas-tStart Master plus SYBR Green I kit (Roche) as recommended by the manufacturer. The final primer concentration used was 500 nM. Each sample was run in duplicate. The PCR conditions were 95°C for 10 min followed by 50 cycles at 95°C for 5 sec, 60°C for 10 sec and 72°C for 10 sec. The specificity of amplified qPCR products, was checked by melting curve analysis. Relative transcript levels were quantified using standard curves prepared with cDNA from tissue samples in which they were abundant, i.e. ovarian, telencephalon, or testis, for lhcgr1, lhcgr2, and fshr, respectively. Results were expressed as arbitrary units of gonadotropin receptor transcript level / total RNA level, and normalized to the mean expression value in the ovary, considered as 1.

Results, Discussion, and Conclusions
A single fshr but two lhcgr are present in basal teleosts, the eels Prediction of a single fshr gene in both European and Japanese eel draft genomes. Blast searches using the fshr cDNA sequence cloned in the Japanese eel (AB360713; [28]) allowed us to retrieve one scaffold covering the full-length fshr cDNA in the Japanese eel genome, as well as in the European eel genome (S1 Fig).
Prediction of two lhcgr genes in both European and Japanese eel draft genomes. Blast of the lhcgr cDNA sequences previously reported in Japanese eels (AY742795; [27] Jeng et al. 2007 and EU635883; [29]) against the Japanese eel genome, allowed the prediction of two lhcgr genes located on two distinct scaffolds (S1 Table). One lhcgr gene corresponds to the cDNA sequence identified by Kazeto et al. [29]. We named it lhcgr1 according to phylogenic and synteny analyses (see sections below). The other lhcgr showed high sequence identity with the partial cDNA sequence identified by Jeng et al. [27] and was named lhcgr2. Blast of the two Japanese eel predicted lhcgr sequences against the European eel genome also allowed the identification of two lhcgr genes located on two distinct scaffolds. European eel lhcgr2 sequence could be only partially predicted, due to low assembly resolution or ambiguities of the European eel draft genome. To confirm the presence of three gonadotropin receptors in the eels and complete the predicted coding sequences, the cDNA for fshr, lhcgr1 and lhcgr2 were cloned in the European eel.
Cloning of the cDNAs of the three European eel gonadotropin receptors. Fshr cDNA (2040 bp) was cloned using testis RNA (S2 Fig). The cloned sequence shared 99.7% identity with the predicted European eel fshr CDS and the recently published European eel fshr cDNA cloned from ovarian RNA by Minegishi et al. [31] (AB700600). Few mismatches observed between the predicted genes and the corresponding transcripts may reflect some polymorphism. These differencies concerned the third base position of the codons, with no impact on amino acid sequences. A partial cDNA of 1913 bp encoding lhcgr1 was cloned from ovarian RNA (S3 Fig). This sequence covered 90.4% of the predicted CDS (2115 bp) of European eel lhcgr1. It shared 99% identity with the European eel lhcgr1 CDS and 98.1% identity with the Japanese eel lhcgr1 cDNA (EU635883; [29]). The full-length coding sequence of the European eel lhcgr2 was cloned (2104 bp) using testis RNA (S4 Fig). It shared 98.5% identity with our predicted Japanese eel lhcgr2 CDS. It is the first time that duplicated lhcgr transcripts are isolated in a vertebrate species.
Prediction of the protein domains of the eel gonadotropin receptors. European eel fshr CDS encoded a 660 amino acids (aa) FSHR including a 17 aa signal peptide (S2 Fig). Lhcgr1 and lhcgr 2 CDS encoded a 704 aa LHR1 including a 26 aa signal peptide, and a 700 aa LHR2 including a 17 aa signal peptide, respectively (S3 and S4 Figs). The three eel gonadotropin receptors showed the typical topology of the glycoprotein hormone receptors with a long Nterminal extracellular domain (ECD), corresponding to about the half of the whole amino acid sequence (322 aa, 349 aa and 359 aa, for FSHR, LHR1 and LHR2, respectively) (Fig 1). The Nterminal cysteine box (cb1) of ECD of the eel receptors included 4 cysteines for both eel LHR as for mammalian glycoprotein hormone receptors, but only 2 cysteines for FSHR as in some other teleosts. The typical leucine-rich repeat domain (LRRD) of the ECD was composed of 11 leucine-rich repeats (LRR) for the three eel receptors as in mammals (Figs 1 and 2). The succession of LRR confers to the ECD a concave shape that is involved in the hormone binding [40][41][47][48][49][50]. The C-terminal part of the ECD, which corresponds to the hinge region, was delimited in the three eel receptors by two conserved cysteine boxes (cb2 and cb3) containing 3 cysteines each. The transmembrane domain (TMD) of the three eel receptors (263 aa, 267 aa, and 264 aa, for FSHR, LHR1 and LHR2 respectively) was composed, as for all GPCR, of highly conserved seven transmembrane alpha-helices (TH), connected by 3 extracellular (el) and 3 intracellular (il) loops. The receptors terminated with an intracellular domain (ICD) corresponding to a C-terminal cytosolic tail (58 aa, 62 aa, and 60 aa, for FSHR, LHR1 and LHR2 respectively). The ICD is involved in GPCR signal transduction and receptor intracellular trafficking and desensitization (for review [51]).
Comparison of the gene structure of the three eel gonadotropin receptors. Complementary DNA sequences allowed us to assess the intron-exon gene structure of the eel gonadotropin receptors. The European eel fshr gene consisted of 10 exons and showed the typical exon-intron structure described for the human fshr [52], (Fig 1). The coding part of first exon (156 bp) encoded the signal peptide and the N-terminal region of the ECD, including the first cysteine box (cb1) and one of the LRR. A succession of 7 small exons (69-81bp) and a larger 9th exon (188 bp) encoded the major part of the ECD, including all the other LRRs, and two cysteines of the second cysteine box (cb2). The large 3'-terminal "rhodopsin like" exon, covered the C-terminal region of the ECD, including the third cysteine of the cb2, the hinge region and   Table) showed a similar typical 10 exon structure, with the exception of salmonid and percomorph fshr, which present several additional exons as previously reported [21,53].
The European eel duplicated lhcgr genes were both organized in 11 exons as the human lhcgr (Fig 1) [4,54]. As compared to eel fshr gene, they included an additional exon between the 9th exon and the large terminal exon. As for human lhcgr, this additional small exon (72 bp and 69 bp for eel lhcgr1 and lhcgr2, respectively) encoded the third cysteine of the cb2, and a part of the hinge region. This typical 11-exon structure was also predicted for lhcgr from various vertebrates, including chondrichthyans, a non-teleost actinopterygian and various teleosts [21] (Fig 1 and S1 Table) (see next section).

Identification of additional gonadotropin receptors in representative gnathostome genomes
To investigate the evolution of the gonadotropin receptors in gnathostomes, we conducted a search for gonadotropin receptor sequences within 30 vertebrate genomes including two chondrichthyans, one basal sarcopterygian fish, 12 tetrapods, 1 non-teleost actinopterygian and 14 teleosts.
A single fshr and one or two lhcgr are present in teleosts. The screening of the available representative teleost genomes revealed, as in eels, the presence of one fshr and of both lhcgr1 and lhcgr2 genes in seven other species, e.g. in a characiforme, the Mexican tetra, Astyanax mexicanus, in a salmonid, the Atantic salmon, Salmo salar, in two atherinomorphs, the platyfish, Xiphophorus maculatus and the guppy, Poecilia reticula, and two cichlids, the Nile tilapia, Oreochromis niloticus, and zebra mbuna, Maylandia zebra, and in a pomacentridae, the bicolor damselfish, Stegastes partitus (S1 Table). The coexistence of duplicated lhcgr in various teleost species clearly refutes one previous assumption of two mutually exclusive lhcgr in this lineage [21]. In Atlantic salmon, in addition to the lhcgr2 gene, we succeeded to annotate duplicated lhcgr1 genes (S1 Table and S5 Fig). The large number of lhcgr1 in this species may result from the recent salmonid tetraploidization (4R) [55].
In contrast, we found a single fshr gene and alternatively only one or the other of the two lhcgr genes, in seven other teleost annotated genomes. One lhcgr homologous to eel lhcgr2 (see section Phylogeny) was identified in a cypriniforme, the zebrafish, Danio rerio, while lhcgr homologous to eel lhcgr1 (see section Phylogeny) was identified in a gadiforme, the Atlantic cod, Gadus morua, in a beloniformes, the Japanese medaka, Oryzias latipes, in a gasterosteiforme, the stickleback Gasterosteus aculeatus, and in two tetraodontiformes, the pufferfish, Tetraodon nigroviridis and the fugu, Takifugu rubripes. In another cypriniforme, the common carp, Cyprinus carpio carpio, two putative lhcgr2 were identified (one full-length and one partial with potential frameshifts), that may result from the recent cyprinid tetraploidization (4R) [56][57] (S1 Table).
A single fshr and two lhcgr are present in a non-teleost actinopterygian. We could predict one fshr and two lhcgr (S1 Table), homologous to eel lhcgr1 and lhcgr2, respectively (see section Phylogeny), in the genome of a holostean actinopterygian, the spotted gar. The fshr consisted of a succession of 10 exons, as human and eel fshr (Fig 1), and encoded a 719 aa receptor. Both lhcgr were organized in 11 exons as eel lhcgr1 and lhcgr2, and human lhcgr (Fig  1), and encoded a LHR1 (696 aa) showing 53% identity with eel LHR1, and a LHR2 showing 59% identity with eel LHR2, respectively ( Table 1). Considering that the holostean lineage has diverged before the teleost-specific 3R, the duality of lhcgr in the gar indicates that a gene duplication event has already occurred before the teleost 3R.
A single fshr and a single lhcgr are present in sarcopterygians. In a basal sarcopterygian, the coelacanth, one partial lhcgr could be predicted on scaffold JH127072 (S1 Table). This partial sequence consisted of 10 exons giving a 632 aa protein, the first exon encoding for the signal peptide being missing. In addition, we could identify one fshr gene (S1 Table) on scaffold JH127072, positioned in tandem with the lhcgr. The fshr sequence consisted of 10 exons and encoded for a 690 aa receptor. As in the coelacanth, a single fshr and a single lhcgr are present in the other sarcopterygians, including tetrapods. No additional lhcgr could be found in any sarcopterygians. In the course of our investigation of genome databases, we observed that some sauropsid species possess fshr but may have lost lhcgr. In birds, fshr and lhcgr are still present and positioned in tandem, as in coelacanth and mammals. In contrast, we could not find any lhcgr in the genome of three squamates, the anole lizard, the Burmese python and the king cobra. Furthermore, we found several premature stop codons and frameshifts in the putative lhcgr sequences of a chelonian, the painted turtle and of crocodilians, the Chinese and American alligators (S1 Table). This suggests that a loss of a functional LHR would have occurred several independent times during the sauropsid radiation, in squamate, chelonian and crocodilian lineages [58][59]. Similarly, our recent studies on the kisspeptin system revealed multiple and independent losses of kiss and kiss receptor genes during the sauropsid radiation [60].
A single fshr and a single lhcgr are present in chondrichthyans. In the elephant shark genome, a single fshr and a single lhcgr could be predicted on scaffold 134 and scaffold 49, respectively (S1 Table). They showed the typical gene organization consisting of 10 exons and 11 exons (Fig 1), respectively. Fshr encoded a 741 aa receptor, and lhcgr a 709 aa receptor. From the little skate contig data, we succeeded to assemble a single partial fshr encoding 570 aa and a single lhcgr encoding 691 aa (S1 Table).
In conclusion, genome searches indicated that among all gnathostomes investigated, the actinopterygians, including teleost and non-teleost species, were the only group with two lhcgr. This suggests that duplicated lhcgr may represent a specific feature of the actinopterygian lineage.

Phylogeny of gonadotropin receptors
To better understand the relationships between gnasthostome gonadotropin receptors, a molecular phylogeny (Fig 2) analysis was conducted on 78 gonadotropin receptor protein sequences from 35 species (S1 Table). The glycoprotein hormone receptor of a cephalochordate, the lancelet, was used to root the tree. The gnathostome FSHR and LHR formed two distinct monophyletic groups (Fig 2).
The FSHR from the two chondrichthyan species clustered together at the base of the gnathostome FSHR group. The osteichthyan FSHR subdivided in two clades, one formed by the sarcopterygian and the other by the actinopterygian FSHR. Among the sarcopterygians, the coelacanth FSHR branched at the basis of the tetrapod FSHR, in respect to its phylogenic position. Similarly, among the actinopterygians, the spotted gar FSHR branched at the basis of teleost FSHR, in agreement with its phylogenic position. The European and Japanese eel FSHR branched at the basis of the other teleost FSHR, consistent with their corresponding phylogenic position. The topology of the teleost FSHR clade followed the known phylogenetic relationships among teleosts.

Duplicated LH Receptors in Actinopterygians
The LHR from the two chondrichthyan species formed a distinct clade at the base of all gnathostome LHR, in respect to their phylogenic position. The osteichthyan LHR group subdivided in two clades: one clade of sarcopterygian LHR including the coelacanth LHR and the tetrapod LHR and one clade of actinopterygian LHR. The actinopterygian LHR clearly further separated into two clades, each well supported (99-100% bootstrap values), and including both gar and eel LHR1, or gar and eel LHR2. Spotted gar, European and Japanese eel LHR1 or LHR2 were at the basis of each of the two actinopterygian LHR clades, respectively, in agreement with their phylogenic positions. For other teleost species possessing two LHR, such as Atlantic salmon, Mexican tetra, Nile tilapia and platyfish, the duplicated LHR were distributed between actinopterygian LHR1 and LHR2 clades, respectively, as for eels and spotted gar. In the species where only a single LHR was found in the genome, such as zebrafish, carp, medaka, stickelback, tetraodon, fugu and cod, the LHR clustered in one or the other of the two clades (LHR2 for cyprinids, LHR1 for the other species). Previously characterized LHR from various teleost species (e.g., siluridae, pleuronectidae, sparidae), for which the genome has not yet been completed, were represented in one or the other LHR clade, with a topology congruent with the known phylogeny of the teleost radiation. One may highlight the striking case of some closely related species, such as siluriformes [20], with the LHR of the channel catfish, Ictalurus punctatus, branching within the LHR1 clade, while the LHR of the African catfish, Clarias gariepinus, within the LHR2 clade. A similar situation was previously reported in percomorphs, between two pleuronectiforms or between two sparidae [21]. Our finding of coexisting duplicated lhcgr should promote further investigation to decipher if each of these species possess in fact the two lhcgr receptors, or if alternative losses of one or the other lhcgr have occurred in closely related species.
In conclusion, the presence of duplicated lhcgr in the eel provides the first demonstration of the duality of lhcgr in teleosts, as previously suggested by Chauvigné et al. [21]. Furthermore, we showed the presence of duplicated lhcgr in various other teleosts, including characids, salmonids, cichlids, poecilids. The presence of duplicated lhcgr genes in teleosts may have led to the assumption that they are derived from the teleost-specific third whole-genome duplication (3R). We recently demonstrated a 3R-origin for other duplicated glycoprotein hormone receptors in teleosts, thyrotropin hormone receptors (tshra and tshrb) [33]. However, differently from the case of tshr, our present study revealed that the duplicate lhcgr1 and lhcgr2 genes are also present in a non-teleost actinopterygian, the spotted gar, suggesting that the lhcgr duality in teleosts does not result from the 3R but might be derived from an anterior duplication event.
Given the presence of a single lhcgr gene in chondrichthyans and sarcopterygians including the basal coelacanth, the most parsimonious assumption is that the lhcgr duplication event is specific to the actinopterygian lineage. In teleosts, the 3R whole-genome duplication event would have been expected to generate further duplicated fshr, lhcgr1 and lhcgr2 genes. Given the fact that, despite the 3R, we found no additional fshr and lhcgr genes in extant teleosts as compared to the gar, it could be assumed that a massive loss of duplicated gonadotropin receptor genes derived from the 3R, occurred in early teleosts. The impact of the recent tetraploidization events (4R) that occurred in salmonids, is still evident on the number of lhcgr genes in Atlantic salmon (Fig 2).

Synteny of gonadotropin receptors
To further elucidate the gonadotropin receptor evolutionary history, we characterized and compared the adjacent genomic regions of each fshr and lhcgr loci, in sarcopterygians including representative tetrapods and the basal coelacanth, in a non-teleost actinopterygian, the spotted gar, and in teleosts including species like the eels and the tilapias which have conserved both lhcgr types, and species like the medaka and the zebrafish that have conserved only a single lhcgr (lhcgr1 or lhcgr2) (Fig 3).
In human genome, which contains only a single fshr and a single lhcgr as for all sarcopterygians, both receptors are positioned in tandem within the same locus (locus 2p21) [61] (Fig 3). The arrangement in tandem led to the hypothesis that fshr and lhcgr arose from a local, smallscale duplication of an ancestral gene, rather than from the whole-genome duplication events that occurred in early vertebrates (2R) [52]. This tandem position was shown to be conserved in all tetrapods [21-52-62]. In the present study, we observed a similar tandem position for fshr and lhcgr in a basal sarcopterygian, the coelacanth (on scaffold JH127072) (S3 Table).
In the spotted gar genome, fshr and duplicated lhcgr genes were localized on the same chromosome LG16. We named the spotted gar duplicate lhcgr genes according to their genomic position: the lhcgr positioned in tandem with fshr, as in the sarcopterygians, was called lhcgr1, while the other lhcgr, separated by around 47 genes (3Mb) from the fshr-lhcgr1 tandem, was Duplicated LH Receptors in Actinopterygians called lhcgr2 (Fig 3). Lhcgr2 was in a specific genomic region, sharing no paralogs with the genes neighbouring fshr-lhcgr1 tandem. The presence of the fshr-lhcgr(1) tandem in an actinopterygian as in sarcopterygians, suggests that the local duplication event that gave rise to the tandem, would have occurred early in the gnathostome lineage. We hypothesize that a second local duplication occurred specifically in the actinopterygian lineage, producing the second lhcgr (lhcgr2) (Fig 4).
Synteny analysis indicated that the genomic region of the three actinopterygian gonadotropin receptors has been duplicated in teleosts, likely as a result of the teleost 3R (Fig 3). Conserved 3R-duplicated genes in this region include for instance thbs2, psme4, nrxn1, fbxo11, prcke, ppm1 and prph2 (Fig 3). Synteny analysis confirmed that despite the 3R, teleost genomes contained no additional gonadotropin receptor genes as compared to spotted gar, due to selective gene losses. These losses have concerned not only duplicate gonadotropin genes but also some other duplicate neighbouring genes (e.g. erlec1, msh2, ston1, kcnk12, stf2d1, mertk) ( Fig  3). Noticeably, as a result of these gene losses, the fshr-lhcgr1 typical tandem position was not conserved in extant teleosts, in which fshr and lhcgr1 are located on separate chromosomes (e. g. in Nile tilapia, medaka) (Fig 3). This is in agreement with previous synteny analysis of teleost gonadotropin receptors [21], and with an early study by Oba et al [53] who first indicated by RFLP analysis that lhcgr and fshr were not genetically linked in tilapia. Teleost fshr conserved some neighbouring genes including at the proximal 3' position the 3R-duplicated nrxn1, and at 5' position the 3R-duplicated fbxo11 (Fig 3), in agreeement with previous studies [21].
In Nile tilapia, a species that has conserved the duplicated lhcgr, both lhcgr1 and lhcgr2 were maintained co-localized on the same chromosome (LG13), as in the spotted gar, and were separated from each other by a distance of 350 kb covering 14 predicted genes. Each lhcgr had conserved some of the syntenic neighbouring genes, as compared to the spotted gar. In particular, lhcgr1 was flanked at the proximal 3' by the 3R-duplicated nrxn1 and at 5' by the single ston1, and 3R-duplicated fbxo11, while lhcgr2 at 3' by the single sft2d1 and at 5' by the single mertk.
Synteny analysis confirmed that, due to additional losses of lhcgr genes, some teleost species, despite the 3R, have conserved even less receptors than the gar, as illustrated by the loss of lhcgr1 in zebrafish and of lhcgr2 in medaka (Fig 3). In zebrafish, the syntenic blocks including sft2d1 at 3', and mertk at 5', were found flanking the lhcgr2, as in Nile tilapia and spotted gar. Similarly, medaka lhcgr1 conserved typical neighbouring genes such as 3R-duplicated nrxn1 and single ston1, as in Nile tilapia and spotted gar.
The current draft genomes of the European and Japanese eels allowed only a partial reconstruction of fshr, lhcgr1 and lhcgr2 genomic regions (S6 Fig). Fshr conserved synteny included 3R-duplicated nrxn1 and fbxo11, lhcgr1 synteny included single ston1, and lhcgr2 was surrounded by single sft2d1 and mertk, as in spotted gar and Nile tilapia.
In conclusion, synteny analysis supported the phylogeny analysis, indicating the presence of duplicated lhcgr in actinopterygians, and of their origin before teleost emergence, independently of teleost 3R (Fig 4). The synteny analysis revealed that only single genes for fshr, lhcgr1 and lhcgr2 were maintained after the teleost 3R, subsequent to multiple gene deletions, which also led to the physical separation of the fshr-lhcgr(1) tandem (Fig 4). As discussed by George et al. [52], fshr-lhcgr tandem in tetrapods forms, together with nrxn1, a highly conserved syntenic block, which might be associated with evolutionary constraint that retain regulatory sequences [52]. In teleosts, fshr and lhcgr1 were separated on different chromosomes, but both conserved a 3R-duplicated flanking nrxn1 gene. Additional studies on gene expression regulation in teleosts may provide information whether separate fshr and lhcgr1 genes are submitted to independent or parallel expression regulation mechanisms, as well as on the potential role of nrxn1. In the same perspective, future studies should investigate how differential regulation of duplicated lhcgr in actinopterygians, may contribute to the fine-tuning of LH action.
Proposed nomenclature for actinopterygian, including teleost, duplicated LH receptor genes Given our findings that two LH receptor genes coexist in holostean and teleost species, as a result of local gene duplication early in the actinopterygian lineage, we proposed a phylogenetically-founded nomenclature. Based on their orthology with the single sarcopterygian lhcgr, and on their origin from a local gene duplication, we classically named the duplicated actinopterygian lhcgr genes, lhcgr1 and lhcgr2. According to the synteny analysis in the gar, the duplicate lhcgr gene that had conserved the close tandem position with fshr (an ancestral feature in osteichthyans) was named lhcgr1. The other gar lhcgr gene was named lhcgr2. In teleosts, the whole genome duplication (3R) would have potentially generated a further doubling of LH receptor genes as compared to the gar. However, due to gene losses early after 3R, and before the emergence of the basal group of Elopomorphs, extant teleost species investigated so far do not possess more lhcgr genes than the gar. Furthermore, phylogeny analysis clearly clustered teleost and gar sequences in a single lhcgr1 clade, and a single lhcgr2 clade. Thus, based on the origin of duplicated lhcgr in actinopterygians before the 3R, and on their molecular phylogeny relationships, the same nomenclature, lhcgr1, lhcgr2, was applied in teleosts, as in the gar. A specific nomenclature had been previously developed for teleost dual LH receptor isoforms, "lhcgrba, lhcgrbb", the first letter "b" designing 3R-paralogon "b" and the second letter "a or b" Fig 4. Origin and evolution of duplicated lhcgr in actinopterygians. Fshr-lhcgr tandem was inherited by actinopterygians and sarcopterygians from a commun osteichthyan ancestor. The fshr-lhcgr tandem was conserved in sarcopterygians including tetrapod [52] and actinistian (coelacanth) lineages. In actinopterygians, a gene duplication event occurred before the emergence of teleosts, leading to two lhcgr genes, one maintained in tandem position with fshr, and named lhcgr1, and the other one named lhcgr2. The tandem fshr-lhcgr1 and the lhcgr2 were conserved on the same chromosome after the emergence of the holostean (spotted gar) lineage. Teleost whole-genome duplication (3R) potentially generated duplicated fshr, lhcgr1 and lhcgr2. Early after the 3R, multiple and selective gene losses led to the conservation of only single genes for fshr, lhcgr1 and lhcgr2, and to the separation of the fshr-lhcgr (1) tandem. Additional loss events of lhcgr1 or lhcgr2 have occurred independently throughout teleost radiation, so that some extant teleosts have conserved only a single lhcgr, lhcgr1 or lhcgr2, according to species. doi:10.1371/journal.pone.0135184.g004

Duplicated LH Receptors in Actinopterygians
one or the other isoform [21]. This nomenclature was based on the hypotheses of two mutually exclusive LH receptor isoforms in teleosts, located on the same locus, and resulting from interallelic gene conversion and not from gene duplication [21]. These previous hypotheses are invalidated by the present findings. Our proposed nomenclature is based on the new evidences that two LH receptor genes coexist in actinopterygian species, are not located on the same locus, and result from local gene duplication in the actinopterygian lineage, before the emergence of teleosts. In teleosts, duplicated lhcgr1 and lhcgr2 genes correspond to previously designed lhcgrbb and lhcgrba isoforms, respectively.
Structure comparison of the duplicated LH receptors with special reference to the eel Amino acid sequence comparison revealed that European eel LHR1 and LHR2 have largely diverged (49.7% identity), almost as much as compared to FSHR (44.9 and 42.2%, respectively) ( Table 1). When comparing European and Japanese eel, a striking conservation was found for receptor orthologs (> 97%). A higher conservation was observed between each eel duplicate LH receptor and its respective orthologs from other actinopterygians, as compared to the duplicate paralogs: identity ranged between 54 and 64.2% for orthologs and between 41.5 and 47.9% for paralogs (Table 1). In order to further analyze the evolution of duplicated LHR, we compared the amino acid sequences of each functional domain (Table 1 and  Divergence in the extracellular domain between the duplicate LHR. Amino acid sequences of ECD, a region mainly involved in ligand binding, revealed some divergence between eel duplicated LHR (43.6% identity). As expected, further divergence was observed between each eel LHR and eel FSHR (34.1 and 32.1%, for LHR1 and LHR2, respectively). Among actinopterygians, ECD sequence identity was slightly higher between LHR orthologs (49.7 to 57.8% for LHR1, 46.5 to 52% for LHR2) than between LHR paralogs (39.4 to 44%). The conservation of the ECD between all actinopterygian LHR paralogs was still higher than between LHR and FSHR (30.1 to 34.5% identity) ( Table 1).
N-glycosylation sites could be predicted in the ECD of duplicate eel LHR, 3 for LHR1 and 5 for LHR2 (S7 Fig). These sites included for both LHR, a site on LRR7 common to all glycoprotein hormone receptors [19,63], and a site on cb2 conserved among all LHR, except spotted gar LHR2. A N-glycosylation site on LRR3 appeared specific to actinopterygian LHR2. In human, directed-mutagenesis of potential N-glycosylation sites of gonadotropin receptors, or disruption of glycosylation, demonstrated that glycosylation plays a role in the proper folding of the nascent protein into a mature receptor able of binding hormone and signaling [64][65].
An O-sulfation tyrosine site could be predicted in the hinge region of eel LHR2 but not eel LHR1 (S7 Fig). It corresponds to the sulfated tyrosine in the hinge region of human LHCGR (Tyr 331 ), FSHR (Tyr 335 ) and TSHR (Tyr 385 ), that has been postulated to be a prominent feature, conserved among all the GPHR, and involved in receptor activation [40][41][66][67]. This sulfation tyrosine site was predicted in some but not all teleost LHR2, and in none of teleost LHR1 nor teleost FSHR [68]. The lack of potential sulfation tyrosine site in most teleost gonadotropin receptors suggests alternative receptor activation mechanisms than the one requiring the sulfation of the hinge region, as described in mammals [41].
Amino acid comparison of the ECD revealed residues specific of one or the other duplicate LHR, which may constitute fingerprints of each LHR type (S7 Fig). These fingerprints were more marked in teleosts than in spotted gar, indicating that the duplicate LHR may have further diverged in teleosts. Some of these specific residues could be observed in the LRRD domain (S7 Fig). We also found specific features in the hinge region, the LHR1 showing a conserved motif negatively charged (AFHTWRR), that was deleted in the eel LHR2 and substituted in the other teleost LHR2 (S7 Fig). At the level of the Cb3, eel and other teleost LHR1 possess a proline residue (Pro 359 for eel LHR1), as sarcopterygian LHR as well as spotted gar LHR1 and LHR2, while proline is substituted by alanine (Ala 360 ) in eel LHR2, or by alanine or glutamic acid in other teleost LHR2 (S7 Fig). Minor residue replacements were also observed between eel duplicated LHR, as for instance in the LRR6, the substitution of Glu 160 in LHR1 by aspartic acid (Asp 161 ) in LHR2 (S7 Fig). In human LHCGR, replacement by site directed mutagenesis of the corresponding Glu 154 by Asp was shown to provoke an alteration of the response intensity [69].
All these variations in ECD sequences suggest that duplicated LHR may have diverged in their binding and activation properties in teleosts. So far functional studies on teleost recombinant gonadotropin receptors, using transiently transfected mammalian cell lines (Cos-7 or HEK), were performed either on FSHR, or on one or the other of the duplicate LHR, depending of the LHR type isolated [29][30][31]68,[70][71]. Data on ligand selectivity for eel LHR receptor are only available for eel LHR1 and not for LHR2. Japanese eel LHR1 is activated by eel recombinant LH (recLHR) and not to by eel recFSH, but can also be cross-activated by heterologous gonadotropins such as human chorionic gonadotropin (hCG), human FSH and trout LH [29][30][31]68]. Binding specificity toward homologous LH, with no cross-reaction of homologous FSH, has also been reported for LHR in other teleosts, whatever the LHR type: for LHR1 in channel catfish [15], Nile tilapia [68] and medaka [72]; for LHR2 in African catfish [73][74], zebrafish [75], salmon [13][14], European sea bass, Dicentrarchus labrax [76], and Senegalese sole, Solea senegalensis [21]. This indicates that both duplicate LH receptor types would be specific of LH in teleosts. Further functional studies, using recombinant receptors in species having conserved duplicated LHR, are still required to determine whether the dissimilarities between duplicated LHR ECD may affect ligand binding and receptor activation.
Conservation of the transmembrane domain between the duplicate LHR. The amino acid comparison of the TMD showed a high identity between the duplicated eel LHR (68.4%) as well as with FSHR (69.9 and 63.9%) ( Table 1). A strong amino acid identity was particularly observed among gnathostome gonadotropin receptors for the transmembrane helices 2, 3, 6 and 7, which include key residues participating to the network involved both in the transition of the inactive toward active receptor conformation and in binding the protein G [77] (S7 Fig). Nevertheless, some specific residue divergences between the duplicate LHR could be also noticed within the TMD (S7 Fig). As discussed above for ECD, these fingerprints in the TMD were more marked in teleosts than in spotted gar, indicating that the TMD of the duplicate LHR have further diverged in teleosts (S7 Fig). The phosphorylation site (TVR) in the il3 of the mammalian FSHR was highly conserved in the other gnathostome FSHR including the eel FSHR ( S7 Fig). Noticeably, this phosphorylation site was retrieved in spotted gar LHR1, eel LHR1 and some other teleost LHR1, but not in any LHR2 nor in any sarcopterygian LHR (S7 Fig). In the rat, phosphorylation of FSHR il3 was identified as an important determinant in arrestin association [78][79]. Some differences between duplicate LHR TMD sequences might thus confer differential properties in the regulation of signal transduction.
High divergence in the cytosolic tail between the duplicate LHR. The sequence comparison of the ICD showed a very low identity between the duplicated eel LHR (16.9%). A higher conservation was noticed among teleost LHR1 (43.6-49.3% identity) as compared to teleost LHR2 (14.7-32%) ( Table 1). Specific signatures for teleost LHR1 included amino acids shared with other gnathostome LHR but not LHR2 (such as R 652 of eel LHR1), as well amino acids conserved only among actinopterygian LHR1 (Cys 697 ) or among teleost LHR1 (AYHIK) (S7 Fig). Nevertheless, some sites were highly conserved among all gonadotropin receptors, such as the characteristic cysteine palmitoylation sites of the ICD of the GPCR (e.g. Cys 643 -Cys 644 for human LHCGR), which are common to all LGR among GPCR, and involved in receptor internalization and recycling [80][81]. Gonadotropin receptor signaling in mammals is mainly mediated by adenylyl cyclase /cAMP pathway, but alternative signaling, such as phospholipase C/inositol phosphate (IP) and beta-arrestin pathways, are also requested for global activation of the response [82]. To our knowledge, investigations on teleost recombinant gonadotropin receptors mainly focused on the cAMP pathway, while a single study also addressed the IP pathway [73]. The large dissimilarity in cytosolic tail sequences between the duplicated LHR may reflect divergence in downstream intracellular signaling cascade and/or receptor internalization and desensitization process. Such differences may represent selective forces that have contributed to the conservation of the duplicated LHR. There are still limited data on the characterization of the signaling pathway activation and receptor trafficking in teleost fishes. Future investigations on the duplicated LH receptors are needed to characterize their respective activation specificities, and further infer the functional evolution of the LHR.
Comparative tissue distribution of the three eel gonadotropin receptors. Development of specific qPCR for eel fshr and two lhcgr allowed analyzing their tissue distribution in European silver eels.
Eel fshr, lhcgr1 and lhcgr2 transcripts are all expressed in the gonads. The three gonadotropin receptors were easily detectable in the immature ovary of female silver eels. The mean ovarian transcript level was used as a reference for comparison with the other tissues (Fig 5). The three receptors were also well detectable in immature testis of male silver eels, with a much higher expression of fshr in testis as compared to ovary. Previous qPCR data in Japanese eel also indicated the expression of fshr and lhcgr in both ovary and testis [27][28]30]. Early radioreceptor and autoradiographic studies in the European eel ovary, revealed binding sites for carp LH and hCG in both granulosa and external theca layers [83]. In Japanese eel, FSHR was immunolocalized in the testis, in both Sertoli cells surrounding spermatogonia cysts and interstitial Leydig cells [28].
In other female teleosts, in situ hybridization (ISH) revealed that fshr transcript is strongly expressed in granulosa cells, and weakly in theca cells, in chub mackerel, Scomber japonicus [84], Atlantic halibut, Hippoglossus hippoglossus [85] and Atlantic salmon [86]. Lhcgr1 transcript is also expressed in granulosa and weakly in theca cells in chub mackerel [84] and [85] Atlantic halibut. Similarly lhcgr2 transcript is highly expressed in granulosa cells and weakly in theca cells in Atlantic salmon [86]. In medaka, immunocytochemistry (ICC) revealed LHR1 protein labeling in granulosa and theca cells [72].
In male teleosts, fshr transcript is localized by ISH in Sertoli cells, and a weak expression is also detected in Leydig cells in African catfish [87]. ISH and ICC studies also showed FSH expression in both Leydig and Sertoli cells in Senegalese sole [21,88]. Lhcgr2 is expressed in Leydig cells in African catfish [87], and in both Sertoli cells and Leydig cells in zebrafish [89]. In Senegalese sole, ISH and ICC revealed lhcgr2 transcript and LHR2 in Leydig cells and free spermatids, supporting a direct role of LH on spermiogenesis [21,88,90].
In tetrapods, as in the other vertebrates, fshr and lhcgr are mainly expressed in the gonads [52]. The present study supports that fshr as well as duplicated lhcgr1 and lhcgr2 have all conserved a role in the control of gonadal function in the eel. Future studies, including ISH, should aim at further analyzing the cellular localization of the expression of the duplicated lhcgr in the eel, as well as in the other fish species, which have also conserved duplicate lhcgr.
Eel fshr, lhcgr1 and lhcgr2 are differentially expressed in the brain (Fig 5). The three gonadotropin receptors are also expressed in non-gonadal tissues, as analyzed by qPCR in the female European silver eel. Lhcgr2 showed a remarkable expression in the whole brain, higher than the expression measured in the ovary, while lhcgr1 transcript was at the limit of detection. Fshr was highly expressed in some brain regions, such as olfactory bulb and telencephalon (Fig 5). High brain expression of fshr was also reported in female Japanese eel at previtellogenic and late vitellogenic stages [30].
Fshr transcript is also found in the brain of males or females from other teleost species, such as Atlantic salmon [86], and Atlantic halibut [85]. Lhcgr2 transcript is expressed in the brain in various teleosts, including African catfish, Atlantic salmon and European sea bass [20,73,86,91]. Lhcgr1 is also expressed in the brain of the Korean rockfish, Sebastes schlegelii [92] and Atlantic halibut [85], while its expression is not detectable in the brain of the channel catfish [15].
Lhcgr expression has been reported in the brain of amphibians, birds and mammals including human [93][94][95]. In mammals, various functions for brain LHR have been suggested such as GnRH regulation [96], sensory modulation, or fetal neurogenesis [97]. In African clawed frog, Xenopus laevis, LH controls courtship songs via its cognate receptor expressed in forebrain vocal nuclei [95]. To our knowledge there is no report on FSHR in the brain of tetrapods. The fact that fshr and lhcgr are highly expressed in brain in the eel as in some other teleosts, opens new research avenues on the potential roles of gonadotropin signaling in the vertebrate brain. In the eel, the remarkable differential expression of the duplicated lhcgr in brain specifically confers to lhcgr2 major roles in brain functions.
Eel fshr, lhcgr1 and lhcgr2 are differentially expressed in other non-gonadal tissues. Both lhcgr1 and lhcgr2, but not fshr, transcripts were abundant in eye in the eel (Fig 5). Lhcgr2 transcript has also been reported in eye of the European sea bass and mummichog, Fundulus heteroclitus [91,98]. In human, lhcgr transcript and LHCGR protein have been identified in retinal photoreceptors, where LH receptor could be involved in local modulation of vision [99]. The expression of lhcgr1 and lhcgr2 in eye of the eel may reflect an ancestral role in this organ, which may have been conserved by both duplicated receptors.
Fshr and lhcgr1 were well expressed in the eel gills, while lower levels of lhcgr2 were found (Fig 5). Fshr transcript is also expressed in the gills of Senegalese sole [21], lhcgr1 in the gills of channel catfish [16] and Atlantic halibut [84], and lhcgr2 in the gills of Atlantic salmon, European sea bass and mummichog [20,91,98]. This suggests that gonadotropin receptors may participate in the complex multi-endocrine regulation of gill function in teleosts.
Fshr as well as both lhcgr1 and lhcgr2 were expressed at a low level in the eel adipose tissue. Lhcgr2 transcript was also reported in the fat tissue of European sea bass [91]. FSHR transcript and protein have been detected in the abdominal fat tissue of female chicken, where FSH stimulates lipid accumulation [100].
Eel fshr, lhcgr1 and lhcgr2 showed a very low expression, at the limit of detection level, in the other tissues investigated, such as pituitary, liver, intestine, muscle and thyroid. Lhcgr1 transcript is also undetectable by PCR in various non-gonadal tissues, including liver, intestine, heart and spleen, in Korean rockfish [92] and medaka [72]. In contrast, lhcgr1 transcript has been demonstrated in muscle, liver, stomach and heart in channel catfish [16] and in spleen in goby [101]. Lhcgr2 transcript has been observed also in liver, muscle and heart in African catfish [73], and in intestine and pituitary in European sea bass and mummichog [91,98]. This suggests variations in the non-gonadal target tissues and roles of LH, according to teleost species.
Evolutionary scenario of gonadotropin receptors. Genome search, phylogeny and synteny analyses allowed us to revise the number and evolutionary history of gonadotropin receptors in gnathostomes (Fig 6). A single copy of fshr gene is present in gnathostomes. A single lhcgr gene is also present in sarcopterygians, including in the basal representative species, the coelacanth, and it is positioned in tandem with fshr. This tandem, also observed in a holostean actinopterygian, the gar (for fshr-lhcgr1), is likely the result of a local duplication of an ancestral gonadotropin receptor gene in early gnathostomes (Fig 6).
In the elephant shark, fshr and lhcgr were not found positioned in tandem but each gene was on a distinct scaffold (S3 Table). This questions the origin of the fshr-lhcgr tandem. This suggests either that the tandem arose only in osteichthyan lineage after the emergence of the chondrichthyans, or that the tandem position was not maintained in the extant chondrichthyans. Further analysis using other chondrichthyan genomes would allow a better resolution of the glycoprotein hormone receptor evolution history.
Among tetrapods, genome search in sauropsids revealed multiple and independent lhcgr gene mutation or loss events in the squamate, crocodilian, and chelonian lineages. This suggests that gonadotropin action would be mediated only by fshr in these species. In contrast, birds have conserved lhcgr and fshr, positioned in tandem, as in the other tetrapods (Fig 3).
Our investigation revealed that duplicated lhcgr (named lhcgr1 and lhcgr2) are present in some actinopterygians, including the eel and some other teleosts, as well as in a holostean species, the gar. The coexistence of duplicated lhcgr in various actinopterygians, including extant teleosts, rules out a former hypothesis, that the duality of teleost receptors would result interallelic gene conversion and not from gene duplication [21]. We raised the hypothesis that these duplicated lhcgr genes may be the result of a second local gene duplication event, that would have occurred early in the actinopterygian lineage, before the split between holosteans and teleosts (Fig 6).
These duplicated lhcgr, which are present in a non-teleost actinopterygian, the gar, are clearly not the result of the teleost-specific whole-genome duplication (3R). Furthermore, phylogeny and synteny analyses show that there was no impact of 3R on the number of The presence of gonadotropin receptors (fshr, lhcgr, lhcgr1, lhcgr2) in representative extant species of gnathostome lineages is shown on the right. In some species, the presence or absence of some gonadotropin receptors could not be assessed, due to the lack of genomic data, and these cases are indicated by a question mark. Two gonadotropin receptors, fshr and lhcgr are present in gnathostomes including chondrichthyans and osteichthyans. They were positioned in tandem in the early osteichthyes. The fshr-lhcgr tandem was maintained after the split between sarcopterygians and actinopterygians. Gene loss of lhcgr in various sauropsid groups led to the presence of only fshr in some extant sauropsids. A duplicated lhcgr was generated by a local gene duplication event in the actinopterygian lineage before the emergence of teleosts. Teleost whole-genome duplication (3R) generated duplicated fshr, lhcgr1 and lhcgr2. Multiple and selective gene losses after the 3R led to the maintenance of only a single copy of fshr, lhcgr1 and lhcgr2 and to the physical separation of the tandem fshr-lhcgr. Additional lhr losses occurred independently through the teleost radiation, i.e. lhcgr1 was lost in cyprinidae and lhcgr2 in some percomorphs. Recent tetraploidization (4R) event has generated additional copies of lhcgr1 in salmonids. doi:10.1371/journal.pone.0135184.g006 Duplicated LH Receptors in Actinopterygians gonadotropin receptor genes in extant teleosts, due to multiple gene losses after the 3R. Our investigation in the eel, a representative species of a basal group of teleosts (elopomorph) suggests that these losses of 3R-duplicated gonadotropin receptor genes would have occurred soon after the 3R. Thus, only one copy of the 3R-duplicated fshr, lhcgr1 and lhcgr2 was retained, while the other one was lost early during teleost radiation (Fig 6), likely via the rediploidization process that follows polyploidization event. In agreement with previous studies [21,53], one consequence of these 3R-duplicate gonadotropin gene losses in teleosts, is the separation of the fshr-lhcgr tandem in all extant teleosts, with potential implications in the regulatory mechanisms of their expression.
Additional loss events of lhcgr1 or lhcgr2 have occurred independently throughout teleost radiation, so that some extant teleosts have conserved only a single lhcgr (lhcgr1 or lhcgr2), according to species (Fig 6). The loss of lhcgr1 in cyprinids and the preferential loss of lhcgr2 that occurred repeatedly through the percomorph lineage, suggest that the presence of two lhcgr may not be fully stabilized in teleosts. This diversity in lhcgr duplicate gene conservation or losses among teleosts, also raises the question of the possible presence of duplicate receptors in teleost species for which only one lhcgr has been characterized, while genomic data are still limited.
The recent tetraploidization events (4R) that occurred independently in cyprinids and salmonids, are reflected by some additional numbers of lhcgr genes, such as in Atlantic salmon (two lhcgr1 genes) (Fig 6). In contrast, loss of the other putative 4R-duplicated gonadotropin receptors in these species, illustrates the progressive gene fractionation occurring in polyploid genome [55].
Comparisons between gonadotropin receptor amino acid sequences revealed some divergences between teleost LHR1 and LHR2 paralogs, mainly concerning the ECD and ICD domains. These differences, which may affect receptor-binding properties, signaling pathways and recycling, might represent selective forces that have contributed to the conservation of duplicate LHR. We also compared the tissue distribution of the expression of fshr and duplicated lhcgr transcripts, using the eel as a model. All three gonadotropin receptors are expressed in the ovary and testis, and are thus involved in the mediation of the control of male and female reproductive function. In non-gonadal tissues, differential expression patterns were revealed. For instance, high expression of fshr and lhcgr2, but very low lhcgr1 expression was observed in the brain. Different gene regulatory environments could have driven the duplicated lhcgr to distinct extra-gonadal functions. This may act as evolutionary forces in the conservation of the duplicated lhcgr. The present discovery of duplicated lhcgr in actinopterygians, including various teleost species of biological, ecological and aquaculture relevance, opens new research avenues in basic and applied reproductive endocrinology, as well as in evolutionary biology. It will promote future investigation on the comparative function and regulation of gonadotropin receptors, in order to decipher the physiological significance of the conservation of duplicated LHR receptors in various species. Sequences of LHR and FSHR of representative vertebrate species, including chondrichthyan sarcopterygians and actinopterygians, were aligned using ClustalW and manually adjusted. Cysteine residues are in bold red. Cysteine residues potentially involved in a disulfide bond are highlighted in blue. Orange residues represent conserved residues (> 85%) among FSHR and LHR as determined from a general alignment including additional sequences (56 FSHR and 69 LHR sequences, S1 Table). Signal peptides are indicated in italics. Leucine-rich repeats (LRR) of the extracellular domain are delimited by a horizontal black line, and beta-strands of the LRR are highlighted in light blue. Conserved transmembrane alpha-helices (TH) of the transmembrane domain are highlighted in grey. Putative N-glycosylation sites are highlighted in green, potential tyrosine sulfation sites in the hinge region are in red, potential phosphorylation sites in pink and palmytoilation site in green. Specific LHR1 amino acid signatures are highlighted in orange while specific LHR2 signatures in yellow. Abbreviations: Cb, cysteine box; el, extracellular loop; il, intracellular loop. (TIF) S1

Author Contributions
Conceived and designed the experiments: GM SD. Performed the experiments: GM. Analyzed the data: GM SD. Contributed reagents/materials/analysis tools: GM SD. Wrote the paper: GM SD.