The vertebrate glycoprotein hormones (GpHs), gonadotropins and thyrotropin, are heterodimers composed of a common α- and specific β-subunit. The recombinant heterodimer of two additional, structurally related proteins identified in vertebrate and protostome genomes, the glycoproteins-α2 (GPA2) and-β5 (GPB5), was shown to activate the thyrotropin receptor and was therefore named thyrostimulin. However, differences in tissue distribution and expression levels of these proteins suggested that they might act as nonassociated factors, prompting further investigation on these proteins. In this study we show that GPA2 and GPB5 appeared with the emergence of bilateria and were maintained in most groups. These genes are tightly associated at the genomic level, an association, however, lost in tetrapods. Our structural and genomic environment comparison reinforces the hypothesis of their phylogenetic relationships with GpH-α and -β. In contrast, the glycosylation status of GPA2 and GPB5 is highly variable further questioning heterodimer secretory efficiency and activity. As a first step toward understanding their function, we investigated the spatiotemporal expression of GPA2 and GPB5 genes at different developmental stages in a basal chordate, the amphioxus. Expression of GPB5 was essentially ubiquitous with an anteroposterior gradient in embryos. GPA2 embryonic and larvae expression was restricted to specific areas and, interestingly, partially overlapped that of a GpH receptor-related gene. In conclusion, we speculate that GPA2 and GPB5 have nondispensable and coordinated functions related to a novelty appeared with bilateria. These proteins would be active during embryonic development in a manner that does not require their heterodimerization.
In vertebrates, the more documented action of the gonadotropins FSH and LH is the control of steroid hormones secretion by the gonads, whereas TSH is mainly known as the primary regulator of T4 production by the thyroid (1). Gonadotropins, as well as TSH, are heterodimers of two noncovalently associated α- and β-subunits. Both subunits are glycosylated proteins belonging to the large family of cystine-knot proteins (2). These proteins display a unique cysteine bonding arrangement that confers a particularly flat and stable structure and helps association into either homo- or heterodimer. In vertebrates, the α-subunit is common to all glycoprotein hormones (GpH), whereas there are three different β-subunits, each one specific to FSH, LH, or TSH (1). The three GpHs are produced in two cell types located in the pars distalis of the pituitary, the thyrotropes and the gonadotropes, these later producing both LH and FSH. In primates and Equidaes, an additional GpH, the CG, is produced by trophoblast cells. In Equidaes, the β-subunits of LH and chronic gonadotropin are encoded by a single gene that is expressed in both tissues. In primates, the chronic gonadotropin β-subunit gene arose from a duplication of the LH β-subunit gene during primate evolution (3). This raises the number of β-subunit genes in vertebrates to four.
The GpH β-subunit genes must have resulted from duplications of an ancestral gene before the radiation of gnathostomes (4). However, clear orthologs of the GpH α- and β-subunit genes could not be identified in the genomes of urochordates (5), the admitted closest extant relatives of the vertebrates (6–8), and, to our knowledge, in any other nonvertebrate genomes. Interestingly, two proteins that share a reasonably high degree of identity with GpH α- and β-subunits have recently been discovered in both vertebrates and some protostomes (9–11). On the basis of their global structural similarity with GpH α- and β-subunits, they were considered as phylogenetically related and named glycoprotein-α2 (GPA2) and glycoprotein-β5 (GPB5), even though no molecular phylogenetic studies have been used to support these paralogy relationships (9).
GpHs act through the binding to specific receptors (12). Recombinant human GPA2/GPB5 heterodimer was shown to bind to the TSH receptor both in vitro and in vivo and stimulate T4 production in vivo (13, 14). These properties led Hsueh and collaborators (13) to give the name of thyrostimulin to the dimer. Recombinant fly GPA2/GPB5 dimer was also able to activate a GpH receptor-type G protein-coupled receptor in the fruit fly (10). The apparent similarity in the signaling system between the GPA2/GPB5 heterodimer and the vertebrate GpHs was used as a strengthening evidence for their parental relationships (10, 11).
A physiological role of a GPA2/GPB5 heterodimer in thyroid regulation was, however, not supported by the almost normal thyroid phenotype observed in gpb5 knockout mice (14). Moreover, GPA2 appeared to be expressed in a wider variety of tissues than GPB5 in human and rat (9, 13, 15–17), and whereas the pituitary gland was the only tissue in which coexpression of GPA2 and GPB5 was observed in the same cells, there is a discrepancy about the cell type involved, corticotrophs in human (15) and unidentified but not corticotrophs in rat (13). In addition, the expression level of GPA2 was shown to be up to 100 times higher than that of GPB5 (15, 17). These differences in tissue distribution and expression levels strongly suggest that heterodimerization might not be the sole mode of action for these proteins but no alternative function and mechanism of action has been proposed yet. A better knowledge of GPA2 and GPB5 proteins therefore appeared necessary.
In this study, we searched for the presence of GPA2- and GPB5-related genes in the metazoa phyla, analyzed the genomic structure of these genes, and compared their chromosomal environment with those of the vertebrate GpH subunit genes. As a first step in the functional characterization of these genes, we investigated their tissue expression profiles, as well as that of a GpH receptor-related (GpHRR) gene, in a basal chordate, the amphioxus, during embryonic development. Amphioxus (also known as lancelet), the genome of which has recently been described (18), is a free-living basal chordate that retains a vertebrate-like anatomy in adults. We show here specific embryonic expression patterns of GPA2, GPB5, and GpHRR, suggesting their functional implication during development.
Materials and Methods
In silico screen for GPA2- and GPB5-related proteins
Identification of potential GPA2 and GPB5 homologs in distinct Eumetazoa phyla was achieved by screening genomes (assembled genomes whenever available or traces, i.e. untreated whole genome shotgun sequence runs) and expressed sequence tags (ESTs) on appropriate databases (Supplemental Table 1, published as supplemental data on the Endocrine Society’s Journals online web site at http://endoc.endsdojournals.org). Megablast or discontinuous megablast, two of the basic local alignment search tool (BLAST) programs (19) were used when the phylogenetic distance between the query sequence and target database was close enough. Otherwise, tBLASTn, which allows more flexibility, was used. The target sequences were considered as orthologous to GPA2 or GPB5 by using reverse blast (in which the target sequence is used as the query), when the highest hits were obtained with GPA2 or GPB5 sequences. Sequences were aligned manually with the use of the Se-AL software (http://tree.bio.ed.ac.uk/software/seal/).
Amphioxus total RNA isolation and cDNA synthesis
Adult specimens of the European amphioxus species Branchiostoma lanceolatum, were collected from coastal waters close to Banyuls-sur-Mer (France). Specimens were transferred to the laboratory facilities and kept in current sea water in tanks containing 2-cm-deep sea sand. Poly(A) RNA from the anterior third of 15 adult specimens was extracted using Dynabeads mRNA DIRECT kit (Dynal A.S., Oslo, Norway). cDNAs ready for 5′ and 3′ extensions were obtained using BD SMART rapid amplification of cDNA ends (RACE) cDNA amplification kit (BD Bioscience CLONTECH, Palo Alto, CA).
Amphioxus GPA2, GPB5, and GpHRR cDNA cloning
Sequences of gpb5 and two types of gpa2 (gpa2-I and gpa2-II) were found by BLAST searches on the lancelet B. floridae genome database. For GPB5, two primer sets (including nested primers) were designed for the separate amplification of partial sequences of the two exons from European amphioxus genomic DNA (Supplemental Table 2). Single, intron-spanning sets of primers were used for gpa2-I and gpa2-II. PCRs were performed as previously described (20) in 30-μl capillary tubes on 50 ng of genomic DNA using GoTaq DNA polymerase (Eurogentec, Saraing, Belgium) on a 1605 Rapid Cycler (Idaho Technology, Idaho Falls, ID) at temperatures 2–5 C under the lowest melting temperature of the primer pair (Supplemental Table 2). Amplified fragments were subcloned into pGEM-T easy vector (PCR-II kit; Promega, Madison, WI) and sequenced (MWG, Ebersberg, Germany).
Species-specific primers (Supplemental Table 2) were designed when necessary to generate GPB5, GPA2-I, and GPA2-II 5′ and 3′ ends using cDNAs ready for RACE PCR. No amplification was ever obtained for GPA2-II from these cDNAs. GPA2-I and GPB5 extensions were subcloned and sequenced. Full-length cDNA (FM863701 and FM863702) were obtained by fusion of the extensions followed by amplification using the nested universal primer from the RACE kit.
tBLASTn analysis of the lancelet genome using various vertebrate GpHRR or related invertebrate LGR1-type receptors as query allowed us to identify a single GpHRR gene. A set of primers (Supplemental Table 2) was designed to amplify an 850-bp-long fragment of the last exon encoding the transmembrane domain from European amphioxus genomic DNA. The amplified fragment was subcloned into pCRII-TOPO vector (Invitrogen, Carlsbad, CA) and sequenced (FM864151).
Whole-mount in situ hybridization on embryos and larvae
A 402-bp-long fragment including the coding sequence of GPB5 was isolated and inserted into pBluescript-KS vector (Stratagene, La Jolla, CA). GPB5 RNA probe was obtained by T3 RNA polymerase using the digoxigenin RNA labeling kit (Roche Diagnostics GmbH, Mannheim, Germany). The control probe was obtained using T7 RNA polymerase. Similarly, antisense and control GPA2-I RNA probes were obtained from a 381-bp-long GPA2-I fragment. The 850 bp GpHRR fragment cloned into pCRII-TOPO vector was used for the synthesis of antisense (SP6 RNA polymerase) and control sense (T7 RNA polymerase) riboprobes.
Induction of spawning, in vitro fertilization and fixation of embryos and larvae were performed as described (21, 22). Specimens were kept at −20 C in 70% ethanol until processed. Whole-mount in situ hybridization of embryos and larvae was performed as described (23) but with the use of BM Purple (Roche) as chromogenic agent.
Results
GPA2 and GPB5 are present in major groups of bilaterian metazoan
In silico search for GPA2 and GPB5 resulted in the identification of complete or partial sequences of both GPA2 and GPB5 orthologs in all major groups of bilateria (Fig. 1, supplemental Fig. 1, and Tables 3 and 4). Bilaterians are composed of two major groups, the protostomes that subdivides into two subgroups, the lophotrochozoans (mollusks and annelids) and the ecdysozoans (including arthropods and nematodes) and the deuterostomes that contains the chordates. The search on genomes of sea anemone and hydra, both belonging to Cnidaria, a group that diverged before the radiation of bilateria, was, however, unsuccessful, even if it yielded other cystine-knot protein encoding genes (data not shown). The coding sequences of newly characterized GPA2 and GPB5 orthologs were given accession numbers BN001237 to BN001274 (Supplemental Tables 3 and 4). EST data allowed us to correct some of the predicted sequences. The nematode Caenorhabditis elegans GPB5 sequence (BN001265) was based on the EST EC018459. It differs from the gene model NC_003283.8 and includes an additional cysteine residue at its 5′ end allowing a normal cystine-knot-type structure to be adopted (supplemental Fig. 1B). The sequence of ESTs also allowed us to correct the predicted gene of the zebrafish gpb5 on chromosome 13. The predicted model for the Florida lancelet gpb5 sequence was truncated at its 5′ end compared with the sequence from this work (BN001268) that is supported by the European amphioxus cDNA (FM863701).
GPA2 and GPB5 genes in bilateria. Phylogeny of bilateria (8, 24 ). Boxes symbolize exons (open boxes for signal peptide) and lines, introns. The size of the boxes and lines are not proportional to the length of exons and introns. Vertical pins represent potential N-linked glycosylation sites. The second exon box in silkworm GPB5 is truncated because the occurrence of a second intron is suspected. Arrows indicate orientation (5′ to 3′) of the gene. Brackets indicate specific duplications. Double slash indicates interpolate genes. ND, Not determined. NF, not found.
Surprisingly, no gpa2-related sequences could be found in two birds (chicken and zebrafinch) genomic or EST databases. Also, neither gpa2 nor gpb5 could be found in honey bee and wasp genomic or EST databases.
In most species, single gpa2 and gpb5 genes were found. However, the silkworm and unrelated teleosts tetraodon and zebrafish have two gpb5 genes. The amphioxus has two gpa2 genes and the sea urchin has two sets of the two genes.
The gene structure is well conserved except in insects
The untranslated regions of gpa2 and gpb5 orthologs were not conserved enough for reliable interspecies alignments when specific EST data were not available, and, consequently, the number and position of intron sequences were compared only within the coding regions (Fig. 1 and supplemental Fig. 1). Our results show that the gene organization for gpa2 and gpb5 is well conserved except in insects. The gpa2 coding region contains three exons. The first intron splicing site is located four amino acids downstream of the first conserved cysteine residue, C1 (see supplemental Fig. 1A for cysteine numbering) and the second between the seventh and eighth amino acid downstream of the double cysteines C5C6. The second intron is missing in mosquitoes as well as in flour beetle, whereas fruit fly lacks the first one.
The coding region of gpb5 contains only one intron in most species, inserted five amino acids downstream of the C3XGXC4 box (supplemental Fig. 1B). In the nematode C. elegans (and probably other nematodes) an additional intron is located within the signal peptide sequence as revealed by the EC018459 EST sequence. The fruit fly and southern house mosquito coding sequences were encoded by a single exon. The intron insertion site in yellow fever and malarian mosquitoes was shifted toward the 3′ end, between C8 and C9. A potential second splicing site (A/GT or G/GT) is located at this place in the two silkworm gpb5 sequences in addition to the first one. If not used, the sequence downstream of this potential splicing site would be very short compared with other arthropods. An additional intron is also likely present in the signal peptide sequence in green anole (BN001274) and in zebrafish gpb5-II (BN001273).
Gpa2 and gpb5 genes are located in close vicinity in most genomes
Gpa2 and gpb5 genes are located close to each other in all nonvertebrate species studied here (Supplemental Tables 3 and 4). These genes are usually within hundreds to thousands base pairs in the genome without interpolate genes, except in Caenorhabditis species in which several million base pairs separate them. The relative orientation of the two genes might be either head to head or tail to tail (Fig. 1). This syntenic organization is also observed in teleosts but not in frog, human, and chicken.
Conserved genomic environment between basal deuterostome gpa2/gpb5 and the vertebrate α- and β-subunits
Some of the genes found in the gpa2/gpb5 locus of basal deuterostomes are located in the genomic environment of the GpH α- or one of the β-subunit genes in vertebrates (Fig. 2 and Supplemental Table 5). Orthologs of sea urchin hnrnpL are found close to both lhβ on human chromosome 19 and the α-subunit gene on chromosome 6. An ortholog of archain, which is in the gpa2/gpb5 locus of sea squirt and amphioxus, is found on human chromosome 11, close to fshβ and gpa2. Vangl1 and bax orthologs, localized in the gpa2/gpb5 locus of the sea squirt, are close to human tshβ and lhβ, respectively.
Conserved synteny between gpa2 and gpb5 in basal deuterostomes and the human GpH α- and β-subunit genes. The chromosome/scaffold location is given for each taxa. Arrows indicate the orientation (5′ to 3′) of the gene. Double slash indicates interpolate genes. Precise chromosomal coordinates are given in Supplemental Table 5.
Some GPA2 and GPB5 proteins have no potential N-linked glycosylation sites
The overall protein structure of GPA2 and GPB5 is quite well conserved in bilateria (supplemental Fig. 1). All full-length GPA2 sequences contain 10 cysteine residues. Vertebrate GPA2 sequences are three amino acids longer in the region corresponding to loop 1 in cystine-knot proteins (see Fig. 3), whereas the nematode sequences are shorter in the loop 2. The third loop is also variable in length. This is the case for loops 1 and 3 in GPB5 sequences as well. GPB5s are more compact in deuterostomes than in protostomes. The length of loop 2 is, however, stable. Some of the GPB5 proteins are truncated and lack C9 and C10.
Schematic structural comparison between vertebrate GPA2, GPB5, and vertebrate GpH subunits. For GPA2 and GPB5, conserved cysteine residues are numbered as in supplemental Fig. 1 and disulfide bridges (dotted lines) are predicted after the hCG structure in the Protein Data Base (http://www.pdb.org/pdb/explore/explore.do?structureId=1hcn). Cysteine residues and bondings constitutive of the cystine-knot are in gray. Conserved amino acids between vertebrate GPA2 and GpH α-subunit as well as between vertebrate GPB5 and the putative ancestral GpH β-subunit (25 ) are in black. Positions of the conserved glycosylation sites are represented by Y-shaped structures. Conserved intron (I) positions are indicated (>). Orientation (NH2 to COOH) of the protein and loop numbers are as indicated for GpH α-subunit (see also supplemental Fig. 1).
The number and position of asparagine (N) residues having the potentiality to bear a glycosylated chain in GPA2 and GPB5 are highly variable (Fig. 1 and supplemental Fig. 2). Most vertebrate GPA2 have two potential N-linked glycosylation sites (NGSs) in a conserved position. The two teleost GPA2 have a third NGS and conversely, the lamprey might have only one (although the available sequence does not include the first exon yet). In Florida lancelet, GPA2 isoform-II does not show any NGS. This is also the case for a number of species like the sea urchin, sea hare, some insects and nematodes. Mosquitoes GPA2 have two NGSs. The same situation occurs for GPB5 with species that have none and others one or two sites. Vertebrate GPB5s usually have two NGSs, one of which being in a conserved position. No NGSs are found in fly and mosquitoes as well as some nematode GPB5.
Several structural features are identical in vertebrate GPA2 or GPB5 and vertebrate GpH subunits
The similarity in amino acid sequences between vertebrate GPB5 and the putative ancestral vertebrate GpH β-subunit [i.e. before the serial duplications leading to the three GpH lineages (see Ref. 25)] is rather high, particularly around loop 3 (Fig. 3 and supplemental Fig. 1B). They are about the same length and share the intron splicing site. The major difference resides in the carboxy terminal end, which is extended in GpH β-subunits, providing them with an additional disulfide bonding. Vertebrate GpHα and GPA2 are more dissimilar. The length of loops 1 and 3 is spectacularly different (Fig. 3 and supplemental Fig. 1A). The two intron-splicing sites in GPA2 are precisely located at the basis of the two loops, 1 and 3. This latter site is in the same position in GpH α-subunits. α-subunits lack the disulfide bonding that links loop 1 to loop 3. In contrast, they have two additional cysteine residues, allowing a specific bridge between the aminoterminal end and the C3XGCC4 box. It is to be noted that in C. intestinalis and sea squirt, GPA2 also contains the CxGCC-type box as in the GpH α-subunits. Loop 2 segments could reliably be aligned, sharing several amino acid positions, and notably the conserved NGS (supplemental Fig. 1A).
A single GpHRR gene in amphioxus
A single GpHRR gene (fgenesh2_pg.scaffold_ 231000008) was shown to be present in the Florida lancelet genome. Its direct genomic environment contains two genes, stonin and tfIIa that have homologs in close vicinity of the GpH receptor genes in vertebrates (Supplemental Table 6).
GpHRR gene, GPA2-I, and GPB5 expression in amphioxus embryo
Expression of GPA2-I was first detected in midneurula-stage embryos (Fig. 4, A and B). It was restricted to the endoderm from the middle to its posterior end. Expression of GPA2-I in the endoderm in late neurula, before mouth opening, was more extended, with its anterior limit reaching pharyngeal endoderm (Fig. 4C). GPA2-I was also expressed at this stage in the cerebral vesicle, in the central part relative to both axes (Fig. 4, C–E). In the pharynx, expression was detected in the right lateral endoderm (Fig. 4, C, D, and F), which corresponds to the club-shaped gland anlagen. Some staining was also detected, depending on precise developmental stages, in the presumptive area of the Hatschek’s pit (Fig. 4D). At larval stage, the GPA2-I gene was strongly expressed in the cerebral vesicle, posterior to the frontal eye. The labeled club-shaped gland was clearly identified at this stage (Fig. 4G). In the trunk and tail, expression was detected in the gut as well as the mesoderm but at a lower level (Fig. 4H).
Amphioxus embryonic expression pattern of GPA2 (A–H), GPB5 (I–M), and GpHRR genes (N–V). Anterior to the left, dorsal to the top (for side views) are shown. Scale bars, 50 μm. Pictures are side views, except where specified. GPA2. A, Midneurula showing expression in the middle and posterior part of the endoderm. B, Dorsal view of the specimen shown in A. C, Premouth stage embryo showing expression in the central and posterior parts of the endoderm as well as the cerebral vesicle (arrowhead) and a part of the right lateral pharyngeal endoderm (double arrowhead). D, Head of a premouth stage embryo showing the expression in the cerebral vesicle (arrowhead), in the lateral pharyngeal endoderm (double arrowhead). The anterior endoderm in which the Hatschek’s pit will later develop (arrow) is also labeled. E, Dorsal view of the specimen shown in D focused on the cerebral vesicle labeling (arrowheads). F, Ventral view, focused on the lateral endoderm. The labeling is restricted to the right part of the pharyngeal endoderm (arrowhead). G, Head of a larva showing expression in the cerebral vesicle (arrow) posterior to the frontal eye (double arrowhead). The club-shaped gland also shows a specific labeling (arrowhead). H, Posterior part of the specimen shown in G showing a high labeling in the gut as well as a lower labeling in the mesoderm. GPB5. I, Midneurula showing ubiquitous expression except in the most posterior part. J, Dorsal view of the specimen shown in I. K, Premouth stage embryo showing ubiquitous expression; an anteroposterior gradient is clearly observed. L, Larva showing expression in the most posterior part of the gut. M, Head of the larva shown in L with a high expression in the club-shaped gland (arrow). GpHRR: N, Midneurula showing expression in the middle and posterior part of the endoderm. O, Dorsal view of the specimen shown in N. P, Early premouth-stage embryo showing expression in a restricted area of the anterior part of the cerebral vesicle (arrowhead). Q, Dorsal view of a specimen at a same stage as in P showing expression in two symmetrical groups of cells in the cerebral vesicle. R, Embryo at the premouth stage, older than the specimen shown in P, showing expression in the cerebral vesicle and middle and posterior endoderm. S, Enlargement of the head of the specimen shown in R with expression in the anterior and posterior parts of the cerebral vesicle (arrowhead and double arrowhead, respectively). T, Dorsal view of the specimen shown in R with labeling in the anterior (arrowhead) and posterior part of the cerebral vesicle. U, Larva showing expression in the gut starting posterior to the pigmented spot (arrowhead). The labeling can be observed until the anus (double arrowhead). The anterior part of the head is out of focus. V, Enlargement of the head of the specimen shown in U showing expression in the cerebral vesicle (arrowhead) posterior to the frontal eye (double arrowhead).
GPB5 expression was first detected at midneurula stage quite ubiquitously. Only the most posterior part of the embryo was not labeled (Fig. 4, I and J). In late neurula, before mouth opening, an anteroposterior gradient of expression was observed with the posterior end more labeled than the anterior one (Fig. 4K). In larva, expression was still ubiquitous but a higher signal was detected in the most posterior part of the gut (Fig. 4L) and in the club-shaped gland (Fig. 4M).
Expression of the GpHRR gene started at midneurula in the posterior half of the endoderm (Fig. 4, N and O). At this stage expression was rather similar to that observed for GPA2-I. At late neurula, before mouth opening, the receptor gene was also expressed in the cerebral vesicle (Fig. 4, P–T). It started with few anterior cells (Fig. 4, P–Q) and later another area became labeled more caudally (Fig. 4, R–T). At the larval stage, a faint expression was still observed in the cerebral vesicle, posterior to the frontal eye (Fig. 4V). The receptor gene was also expressed discontinuously in the gut, from the central part to the anus (Fig. 4U).
Discussion
Strengthened evidences for GPA2/GPB5 orthologous relationships with GpH subunits
The present extensive in silico search for gpa2 and gpb5 genes in metazoan genomes shows that homologs of both proteins are present in all main groups of bilateria, including protostomian lophotrochozoans and ecdysozoans and deuterostomes. In contrast, the absence of both these proteins in two representative species of Cnidaria reported here strongly suggests that gpa2 and gpb5 homologs appeared with the emergence of bilateria. These proteins must thus fulfill important functions possibly related to a novelty that appeared with the bilaterian organization. Nevertheless, a secondary loss in Cnidaria cannot be excluded. Such a secondary loss of gpa2 and gpb5 homologs was indeed observed in the two hymenopterans, honey bee and wasp. Noticeably, some other developmentally important genes are also absent from the honey bee genome (26). As for these genes, the functions of gpa2 and gpb5 could then be dispensable or fulfilled by other proteins in these species. Gpa2 was also undetected in chicken and zebrafinch, but it has been shown that some genes known to exist in chicken from EST or cDNA data are missing in the available reconstructed genome (27).
Our comparison of vertebrate GPB5 sequences with a putative ancestral GpH β-subunit (25) clearly showed that, in addition to the cystine-knot forming cysteines, they share a reasonably high number of amino acid positions as well as the intron-splicing location. They mainly differ in that GpH β-subunits have an extended carboxy-terminal end allowing formation of a seat belt structure that wraps around the α-subunit and strengthens the dimer configuration (28). Several amino acid positions are also very close or identical between the α-subunit and vertebrate GPA2, notably in the region flanking loop 2, including a glycosylation site. It is also of particular interest in view of its phylogenetic position (see Fig. 1) that one of the peculiarities of the sea squirt GPA2 sequence, the CXGCC-type box, is shared with the vertebrate α-subunits. The α-subunits mainly differ from GPA2 by the shortened length of loops 1 and 3. A modification of the intron acceptor sites might be responsible for this shortening as splicing sites are precisely located at the basis of these loops in GPA2.
The genomic environment of the gpa2/gpb5 locus is not well conserved between basal deuterostome species. Archain was the only gene present in this locus both in sea squirt and amphioxus. Nevertheless, we show that several genes present in the locus of one or the other basal deuterostome species are also in the loose vicinity of GpH subunit genes in vertebrates.
Taken together, these data suggest that GpH α- and β-subunit genes arose from a duplication of a genomic region comprising the gpa2/gpb5 locus (Fig. 5) and probably archain, hnrnpl, bax, cep164, and vangl1 genes. We identified gpa2 and gpb5 in lamprey genomic data. The presence of a gonadotropin hormone in lampreys has been already suspected for several decades (29–31) and its β-subunit recently identified (32). The original duplication of the locus thus probably occurred some time between the divergence of urochordates (the group including the sea squirt) and the emergence of lampreys. Gpa2 and gpb5 probably remained in close proximity until the divergence of teleosts. The gpa2/gpb5 locus was then broken during tetrapod radiation.
Hypothetical model for the origin and evolution of the GpH subunits. GpH α- and β-subunits originated from a duplication of the gpa2/gpb5 locus sometime between the radiation of urochordates and the emergence of early vertebrates. The GpH β-subunit locus was then specifically submitted to serial duplications to give rise to three different lineages before the radiation of gnathostomes.
Do GPA2 and GPB5 exclusively function through heterodimerization?
Our observations indicate that gpa2 and gpb5 are kept in tight contact at the genomic level in most bilateria. Such a proximity argues in favor of a temporally or spatially coordinated regulation of the two genes (33). Because the locus integrity is conserved in teleosts, the regulatory constraints are still probably in place in basal vertebrates, even though the original duplication of the gpa2/gpb5 locus had already occurred. In other words, the emergence of GpHs and the functions that they could have taken over up to the teleost lineage did not have loosening effects on the constraints. These constraints could be related to a heterodimerization phenomenon (i.e. the necessity for the two counterparts to be expressed in the same cell and at the same time), but vertebrate GpH β-subunits that play their role as heterodimer are not kept in tight contact at the genomic level with the α-subunit, indicating that genomic proximity is not compulsory for heterodimeric partner genes.
In contrast to what is observed for the GpH subunits, we show that the number and position of potential N-linked glycosylated chains are highly variable in GPA2 and GPB5 proteins. In most vertebrates, both GPA2 and GPB5 can potentially be glycosylated. This was indeed the case for GPA2 and GPB5 purified from rat pituitary (13). In most hexapods, either GPA2 or GPB5 has no NGS. Remarkably, in sea hare or tick, none of the proteins displays a potential NGS. Obviously some of these proteins might have O-linked instead of N-linked glycosylation chains. As for GpHs, the oligosaccharide chains were shown to be important for the secretion and activity of human recombinant GPA2/GPB5 heterodimer (34). When either GPA2 or GPB5 was produced as unglycosylated protein, the dimer was not efficiently secreted by HEK293T cells, and its TSH receptor activation was decreased. If it was a general standard, GPA2/GPB5 heterodimer would hardly be secreted and active in species in which one or both proteins are not glycosylated. In addition, GPB5s lack the C-terminal extension of the GpH β-subunits that consolidates the dimeric structure. These structural characteristics, together with the differences observed in the tissue distribution and expression level in mammals as mentioned earlier, do not favor a mode of action of these proteins as heterodimer.
GPA2 and GPB5 have substantial structural similarities with not only the glycoprotein hormone subunits but also the eight-membered ring cystine knot-containing bone morphogenetic protein antagonists (35). They also are glycoproteins with a cystine-knot structure at their carboxy-terminal end. Representatives of several members of this family of proteins are also present in protostomes, indicating that they originated early in bilaterian evolution (35). Most of them have been characterized for their role in embryonic development or tissue remodeling. They usually act through binding to a bone morphogenetic protein family member, thus antagonizing their action (35). Expression of GPA2 and GPB5 has already been mentioned in fly embryo and larvae (10), suggesting that they might also be involved in fly development. It is thus tempting to hypothesize that, as for the CCAAT/enhancer binding protein-β/consensus activating protein-1/nuclear receptor family of proteins, GPA2 and GPB5 might act by antagonizing other structurally close developmental factors.
GPA2, GPB5, and the GpHRR are involved in amphioxus embryonic development
BLAST analysis allowed us to identify a single GpHRR gene in amphioxus. At the genomic level, it is closely linked to two genes that are also present within the GpH receptor genes loci in vertebrates. In the absence of any genuine glycoprotein hormone in amphioxus, it was thus logical to infer that a GPA2/GPB5 heterodimer would be its natural ligand. In any case it was interesting to compare the GpHRR gene expression profile with that of GPA2 and GPB5.
Gpa2-I, gpb5, and gphrr are all expressed from midneurula stage in amphioxus. This is the first time that expression of these proteins is described during embryonic development. Interestingly, expression of amphioxus gphrr appears rather similar to that of gpa2-I. From late neurula to larval stages, they are both expressed in the endoderm and cerebral vesicle anlagen. Gpa2-I, but not gphrr, is also expressed in the club-shaped gland, a mucus-producing structure that derives from the right side of the endoderm and that is not conserved after metamorphosis (36). Before mouth opening, gpa2-I expression was observed in some but not all embryos in the presumptive area of the Hatschek’s pit, a structure that has long been considered as a primitive pituitary gland homolog (37, 38). This is suggestive of a very short timing of expression. Also, the expression of gphrr is very dynamic in the cerebral vesicle at these stages. It starts with few cells at late neurula stage, extending more caudally in the cerebral vesicle and then fades at larval stage. In contrast, gpa2-I expression appears stronger and more extended at the larval stage than before mouth opening. The territories of expression of gpa2-I and gphrr in the endoderm also slightly differ in that the receptor expression extends caudally from the central part, whereas that of gpa2-I starts just posterior to the pharynx. In contrast to gpa2-I, gpb5 expression is almost ubiquitous from late neurula to the stage before mouth opening, with an increasing anteroposterior gradient pattern. Gpb5 tends to be more specifically expressed at later stages, suggesting that it can then play a different function.
What these functions might be still needs to be deciphered, but the expression patterns of gpa2-I and gpb5 are so different that it seems unlikely that in the amphioxus, at these stages, GPA2 and GPB5 require a compulsory heterodimeric association to play a role. As mentioned above, a stimulating hypothesis is that GPA2 or GPB5 might act by antagonizing some other developmental factors. Conversely, the similarity of expression between GPA2-I and GpHRR suggests that they might be functionally linked. Whether GPA2 may interact with the receptor as monomer, homodimer, or heterodimer, possibly with a different partner, remains to be elucidated. In this respect, a recent study reported the possibility for the vertebrate GpH α-subunit to homodimerize in cells that do not express a GpH β-subunit (39). Different roles for the glycoprotein hormone α-subunit alone have previously been described in mammals, even if the mechanism of action has not been elucidated yet. Of particular interest is that these actions concern cell differentiation, in the pituitary (40) or endometrium (41–43). In amphioxus, the dynamic expression patterns of gpa2-I and gphrr, particularly in the cerebral vesicle, club-shaped gland, or presumptive area of the Hatschek’s pit, strongly suggest that they might be involved in the development of these structures.
Acknowledgments
We thank Jean-Louis Plouhinec and Sylvie Mazan for initial GPA2 and GPB5 in silico search and identification of amphioxus genomic sequences from traces. Many thanks also to Claire Cahoreau, Sylvie Dufour, Yves Combarnous, and Raymond Counis for helpful discussions.
This work was supported by Agence National pour la Recherche and Centre National de la Recherche Scientifique grants (to H.E.). The postdoctoral position of S.B. was supported by the Association pour la Recherche sur le Cancer.
Present address for S.D.S. and B.Q.: Universite Paris Diderot-Paris 7, Equipe d'Accueil Conventionnée 7059 Centre National de la Recherche Scientifique, Physiologie de l’Axe Gonadotrope, CC 7007, 5 Rue Thomas Mann, F-75205 Paris cedex 13, France.
Present address for C.B.: Centre National de la Recherche Scientifique Unité Mixte de Recherche 7138, Université Pierre et Marie Curie, Paris 06, F-75252 Paris cedex 5, France.
Disclosure Summary: The authors have nothing to disclose.
First Published Online March 5, 2009
For editorial see page 3446
Abbreviations
- BLAST,
Basic local alignment search tool;
- EST,
expressed sequence tag;
- GPA2,
glycoprotein-α2;
- GPB5,
glycoprotein-β5;
- GpH,
glycoprotein hormone;
- GpHRR,
GpH receptor-related;
- NGS,
N-linked glycosylation site;
- RACE,
rapid amplification of cDNA ends.
