Introduction

The forkhead domain, also called the winged helix domain, is present in a large number of proteins that constitute a family of transcription factors that activate important pathways of embryogenesis (Weigel et al., 1989; Mahlapuu et al., 2001; Lehmann et al., 2003) and cell differentiation in eukaryotes (Brissette et al., 1996; Dottori et al., 2001; Nakae et al., 2003). This domain, determined by the Pfam database (Finn et al., 2010) as including 96 amino acids, was named after the description of the crystal structure of the hepatocyte nuclear factor-3γ (HNF-3γ)/FOXA family, which is involved in pancreas and liver tissue development (Clark et al., 1993). Many proteins of the forkhead family are considered to be tissue-specific regulators of development (Bravieri et al., 1997). Examples include craniopharyngeal development—FOXE1 (Lehmann et al., 2003), cell growth and insulin responsiveness—FOXO1 (Gross et al., 2008), hair formation and keratinocyte differentiation—FOXN1 (Nehls et al., 1994) and ovarian formation and function—FOXL2 (Cocquet et al., 2002; Baron et al., 2004; Uhlenhaut and Treier, 2006; Veitia, 2010; Jaubert et al., 2011). The role of each protein member of the domain family was discovered primarily by studies of congenital defects associated with mutations, often observed within the forkhead domain (Benayoun et al., 2011).

The FOXL2 (forkhead box L2) gene is an important member of this extensive family and is primarily responsible for ovarian development and maturation (Uhlenhaut et al., 2009). The majority of studies involving FOXL2 describe functional differences of mutations associated with blepharophimosis–ptosis–epicanthus inversus syndrome (BPES), which is characterized by eyelid malformations (BPES type II) or, in some cases, premature ovarian failure (BPES type I) (Zlotogora et al., 1983; Crisponi et al., 2001; De Baere et al., 2001). An exclusive feature of the protein, observed only in mammals, is a tract of 14 alanines, which has been suggested to be a region with strong functional constraints (Cocquet et al., 2003). For instance, expansions of alanines represent ∼30% of the causes of BPES type II (De Baere et al., 2003).

Although there are vast amounts of data available from clinical and expression analysis of Foxl2, only a few studies, restricted to a reduced set of vertebrate species from teleosts to mammals, are focused on the evolution of the gene. To date, the most interesting observation in the evolutionary analysis is the presence of paralogs in the fish group, the origin of which was suggested to be in accordance with the fish-specific whole genome duplication (Christoffels et al., 2004; Jaillon et al., 2004). An extra copy of the Foxl2 gene has been reported in Oncorhynchus mykiss (rainbow trout), Takifugu rubripes (fugu), Tetraodon nigroviridis (pufferfish) (Baron et al., 2004), Danio rerio (zebrafish), Gasterosteus aculeatus (stickleback), Oryzias latipes (medaka) (Jiang et al., 2011) and Salmo salar (Atlantic salmon) (von Schalburg et al., 2011).

Motivated by the previous evolutionary context of Foxl2, the present work was conducted to improve the robustness of Foxl2 evolutionary history, thus including a large number of sequences and also searching for new copies of Foxl2 in vertebrates. Evolutionary descriptions of new Foxl2 single and duplicated copies were obtained for different groups of vertebrates, including the chondrichthyan Callorhinchus milii (elephant shark), the coelacanth Latimeria chalumnae (Comoros coelacanth), the bird Taeniopygia guttata (zebra finch) and the marsupial Monodelphis domestica (opossum). Moreover, single copies of Foxl2 from the neotropical cichlid species Cichla monoculus (peacock bass) and the chondrichthyan species Rhizoprionodon lalandei (Brazilian sharpnose shark) and Callorhinchus callorynchus (Plownose chimaera) were sequenced. Furthermore, the syntenic region analyses of both Foxl2 copies in different vertebrate species were discussed together with the phylogenetic results to support the present conclusions about the duplication event of Foxl2.

Contrary to the previous report on the origin of paralog copies of Foxl2, this work shows an ancient origin of the paralog copies in ancestors of gnathostomes, leading to a suggestion for a new nomenclature for those genes, namely, Foxl2A (the most studied gene form, generally known as Foxl2) and Foxl2B (for the diverged duplicated form) in vertebrates.

Materials and methods

The methodology adopted in the present work is summarized in Figure 1.

Figure 1
figure 1

Fluxogram of the methodology adopted in the present work. The general steps performed for the evolutionary analysis of Foxl2 gene.

Full size image

Sequence acquisition and alignment procedures

Forkhead domain sequences

Protein sequences of the forkhead family were collected from a large number of eukaryotes including fungi, plants and animals. A complete list of Uniprot identifiers (converted to GenBank accession numbers in the Uniprot browser, http://www.uniprot.org), available in the Pfam database (Finn et al., 2010) for the forkhead family, was used to retrieve the corresponding nucleotide and amino-acid sequences utilizing the GBreader software (Razente HL, Braz ASK, Scott LPB (unpublished)). This procedure allowed for the acquisition of 2464 sequences of all available forkhead family proteins in GenBank (Supplementary Material 1).

To avoid redundancy due to highly similar sequences, and thus ensure the use of only a set of representative data, a clustering methodology was performed using the CD-HIT software (Li and Godzik, 2006). This program clustered proteins with identity ⩾90% (the threshold used in the present work) and retrieved only one representative protein from each cluster. This threshold value was applied because the forkhead family exhibits a high conservation level in the domain region and, thus, lower values would indicate the loss of important representative data. The previous analysis resulted in a set of sequences (1023 amino-acid sequences relative to 370 eukaryotic species) that were aligned with the HMMER v. 3 software (Finn et al., 2011) using a Hidden Markov Model (HMM) profile of the forkhead domain. The final alignment was edited in GeneDoc (Nicholas et al., 1997) software to select the conserved regions of the domain and avoid poorly aligned portions (final alignment length: 111 sites; see Supplementary Material 2).

Foxl2 sequences

Because our focus is on Foxl2, besides the sequences presented in the aforementioned data set, other Foxl2 sequences were obtained from the Elephant Shark Genome Project (http://esharkgenome.imcb.a-star.edu.sg/), Ensembl (Flicek et al., 2011), BouillaBase (http://BouillaBase.org) and JGI (http://genome.jgi.doe.gov/) databases (access from December 2011 to March 2012). Only full-length sequences annotated as FOXL2 were retrieved in the first search; however, deep searches were also conducted in all metazoan genomes available in these databases to detect additional, unreported duplicated copies. A tblastn search was performed using Foxl2 sequences from species closely related to the target species as queries. Because the resulting hits were centered only in the forkhead domain, the retrieval was expanded across the hit to obtain the complete Foxl2 coding sequence. These sequences were submitted to an ORF (open reading frame) search and protein translation using Geneious 4.8.5 software (Drummond et al., 2009). Moreover, other Foxl2 sequences were retrieved by a blastp search directly from the GenBank database, available at the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov/). A complete list of Foxl2-retrieved sequences and details of the database searches are shown in Supplementary Material 3.

Experimental procedures were also employed to retrieve Foxl2 sequences for three organisms, the chondrichthyan species R. lalandei and C. callorynchus, and the neotropical cichlid C. monoculus. The animals were collected from Brazilian rivers and marine areas according to Brazilian laws for environmental protection (wild collection permit, SISBIO 12337-1, 15729-1). The experimental research on the animals was conducted according to the international guidelines of Sao Paulo State University (Protocol no. 34/08—CEEA/IBB/UNESP). The tissue samples were available at the Laboratory of Integrative Genomics of Sao Paulo State University, and the genomic DNA was extracted from muscle and liver tissues using the phenol–chloroform method (Sambrook and Russel, 2001). The Foxl2 sequences were amplified by PCR using degenerate primers (forward 5′-GTNGCNYTNATHGCNATGGC-3′ and reverse 5′-CCARTANSWRCARTGCATCAT-3′) constructed with the Primer3Plus software (Untergasser et al., 2007). For the construction of these primers, several Foxl2 protein sequences of vertebrates (Supplementary Material 4) were retrieved based on a blastp search in the Expasy Proteomic Server (http://ca.expasy.org/) using a Foxl2 sequence from Oreochromis niloticus (accession number Q6JA05) as the query. These sequences were aligned using ClustalW (Thompson et al., 1994), and the primers were constructed for the most conserved regions. The PCR cycling sequence was as follows: 1: 95 °C for 2 min; 2: 95 °C for 1 min; 3: 55 °C for 30 s; 4: 72 °C for 1 min; 5: steps 2–4 for 30 cycles; and 6: 72 °C for 5 min. The amplicons were ∼250 bp, and the full gene sequences were obtained by the genome walking technique using Genome Walker kit (Clontech Laboratories, Mountain View, CA, USA) according to the manufacturer’s protocol. All amplicons were cloned in p-GEM-T plasmid vectors (Promega Corporation, Madison, WI, USA), and the corresponding clones were sequenced with an ABI Prism 3100 DNA sequencer (Perkin-Elmer, Waltham, MA, USA) using ABI Prism Big Dye Terminator Cycle Sequencing Ready Reaction kits (Perkin-Elmer). Both strands from different clones were sequenced at least three times to ensure the sequencing quality. Finally, each of these sequences had their amino-acid sequences predicted by an ORF search performed using the Geneious 4.8.5 program (Drummond et al., 2009). These sequences were also included in the data set for forkhead phylogeny.

The full Foxl2 amino-acid sequences present in the forkhead data set used in the domain phylogeny (Supplementary Material 5), and their clustered full sequences, were used for the subsequent Foxl2 phylogeny reconstruction. The amino-acid sequences were aligned using the Muscle algorithm (Edgar, 2004), and obviously misplaced aligned regions were corrected manually. Poor-quality regions observed in the N-terminal portion were excluded from the analysis using the Seaview software (Supplementary Material 6). At the end, the reference data set for Foxl2 phylogeny was composed of 64 amino-acid sequences relevant to 50 metazoan species, and the alignment length included 392 sites.

Forkhead and Foxl2 phylogenetic analysis

The choice of the best-fit model of evolution was performed with ProtTest3 (Darriba et al., 2011) for the forkhead and Foxl2 sequences using the Akaike information criterion (Akaike, 1974) for the best model selection.

The phylogenetic reconstruction was determined by maximum likelihood and Bayesian methods implemented in the Phyml v3.0.1 (Guindon and Gascuel, 2003) and Beast v1.7.0 (Drummond and Rambaut, 2007) software, respectively. The Bayesian analyses were conducted with Beast software allocated in CIPRES Science Gateway (Miller et al., 2010). For maximum likelihood analysis, the approximate likelihood ratio test (Shimodaira-Hasegawa-like) reliability test (Anisimova and Gascuel, 2006) was adopted, and the values were supported by posterior probabilities obtained by Bayesian analysis. For Bayesian method generations, the burn-in was determined in Tracer (Rambaut and Drummond, 2007) through log likelihood scores, and data were summarized in TreeAnnotator (Drummond and Rambaut, 2007) after trees that were out of the convergence area had been discarded. The visualization and the final tree edition were performed using FigTree v1.3.1 (Drummond and Rambaut, 2007) and the software package Mesquite (Maddison and Maddison, 2001). In all phylogenetic analyses, the proportion of invariable sites and γ-distributed rate variation across sites were estimated, and the substitution rate categories were set in four categories. The random local clock model for Bayesian analysis was chosen because it allows for rate variation among lineages, and the model can run a series of local molecular clocks as described in Drummond and Suchard (2010) (see parameters in Tables 1 and 2).

Table 1 Maximum likelihood reconstruction tree parameters
Full size table
Table 2 Bayesian method reconstruction tree parameters
Full size table

Evaluation of the Foxl2 syntenic region

In an attempt to provide additional support for the orthology for most of the newly described Foxl2 sequences, especially the coelacanth L. chalumnae, the avian T. guttata and the marsupial M. domestica copies, and to understand the composition and organization of the genes in the proximity of Foxl2, the syntenic regions were analyzed in D. rerio, G. aculeatus, G. morhua, O. latipes, O. niloticus, T. nigroviridis, T. rubripes, L. chalumnae, T. guttata and M. domestica. Unfortunately, it was not possible to obtain the syntenic regions of Foxl2 for C. milii because the genomic data in the blast hits are still not organized in scaffolds and the reads are short (http://esharkgenome.imcb.a-star.edu.sg/). Furthermore, for comparison, some ortholog sequences from representative species such as Xenopus tropicalis (frog) and Homo sapiens were included in the analysis of the syntenic regions.

Most of the synteny analysis was conducted using Ensembl with the genes identified using its genome browser. Because there were only partial transcript predictions for O. niloticus and L. chalumnae, without gene identification, each partial transcript sequence in the genomic region of Foxl2A and Foxl2B was retrieved and then queried by a blastp search against the NCBI database, with the identities and the E-values annotated from the highest similar hits from blast results (Supplementary Material 7). For each partial transcript, the retrieval was expanded to 50 kb of the flanking regions. The sequences acquired were inserted into the online program Softberry FGENESH (http://www.softberry.com/) using T. rubripes as the reference genome to recover the whole gene transcriptional region and confirm the identification of genes in these syntenic regions (Supplementary Material 7). However, for the gene Pik3cd (phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit δ isoform), there was no transcript annotation in the L. chalumnae Foxl2B syntenic region in Ensembl. Because this gene appeared in our analysis as an important marker for Foxl2B, a tblastn search was conducted in Ensembl to identify the localization of Pik3cd in L. chalumnae using a sequence of the Pik3cd protein from D. rerio (accession number AAH54896) as the query. Because the resulting hit from the blast search was positioned to be syntenic with Foxl2B, 50 kb of the flanking region was retrieved, and gene prediction was performed using Softberry FGENESH to exclude a possible pseudogene and to confirm the correct identification of Pik3cd in L. chalumnae. For the final predicted gene, its corresponding protein sequence (Supplementary Material 7) was used as the query for a blastp search in NCBI to check the protein prediction reliability and integrity.

Exhaustive search for the Foxl2 duplicated copy

To assure that Foxl2 duplicated copies were not missed during the blast search, we proceeded to perform an exhaustive search in the Ensembl genomes, isolating the flanking regions in which Foxl2B was located and conducting a gene name search of important Foxl2B vicinity markers identified in our work, including Pik3cd, Tmem201 (transmembrane protein 201), Clstn1 (calsyntenin-1), Ctnnbip1 (β-catenin-interacting protein 1) and Lzic (leucine zipper and Ctnnbip1 domain-containing protein). We restricted our searches to annotated genomes that exhibited at least one of those markers. Once a syntenic region was identified, the nucleotide sequences of that region were retrieved. These sequences ranged from 15 to 1000 kb, depending upon the size of the available region. A gene prediction was performed in Softberry FGENESH (http://www.softberry.com/) using T. rubripes, X. tropicalis and H. sapiens as reference genomes. For all predicted genes, their corresponding amino-acid sequences were used as queries for a blastp search in NCBI to identify results with similarity to the forkhead protein family. When any similarity was observed, the corresponding sequence was inserted in the domain phylogeny to verify the homology with Foxl2.

Nomenclature reference to systematic relationships

All taxonomic groups and systematic relationships mentioned in this work were based on the iTOL (Interactive Tree of Life) (Letunic and Bork, 2011).

Results and discussion

Foxl2 gene search and sequencing

During the genome searches of annotated sequences, blast and exhaustive searches for Foxl2, diverged duplicated copies were found in the chondrichthyan C. milii, in the teleosts A. burtoni, G. morhua, M. zebra and O. niloticus, in the coelacanth L. chalumnae, in the bird T. guttata and in the mammal M. domestica (one copy ortholog to Foxl2A and the other to Foxl2B, as will be discussed). On the other hand, the Foxl2 sequences obtained by genome walking exhibited just one copy of the gene in the chondrichthyans R. lalandei (accession number JX012129) and C. callorynchus (accession number JX012130), and in the teleost C. monoculus (accession number JX012131). However, the degenerate primers used probably anneal only to Foxl2A sequences, precluding the detection of paralog copies. The same outcome may have occurred in the other teleost species in which the Foxl2 sequences were obtained by PCR procedures in previous studies.

All the sequences obtained for Foxl2 from genome database searches and experimental procedures, which were subsequently used for phylogenetic analysis, exhibited a single exon (as expected for the Foxl2 gene) and a translated protein with variable sizes (Supplementary Material 3). Furthermore, the present gene searches revealed the presence and absence of Foxl2 copies.

Forkhead phylogeny

The domain phylogeny combined sequences from all forkhead family groups (A–R) (Figure 2a). As our goal with this approach was to demonstrate that all Foxl2 analyzed sequences were monophyletic (especially the ones predicted in the present study), additional analyses on the domain phylogeny have not been implemented. Besides, there are well-established studies that already address the evolution of the forkhead family (Kaestner et al., 2000; Mazet et al., 2003; Katoh, 2004; Tu et al., 2006; Hannenhalli and Kaestner, 2009).

Figure 2
figure 2

Phylogenetic relationship of the forkhead domain and Foxl2 sequences. The letters in the forkhead phylogeny (a) represent each different member of the domain family (FoxA to FoxR). The branch colors indicate the consensus statistical support from PhyML and BEAST analysis: blue (<70%) and red (⩾70%). The scale bars indicate the average number of amino acid substitutions per site. In the Foxl2 phylogeny (b), the first and second branch values represent the approximate likelihood ratio test (aLRT; SH-like) and the posterior probability from the PhyML and BEAST programs, respectively. The clades of each Foxl2 form are indicated by right side bars. The highlighted clades represent chondrichthyans (blue), teleosts (green), coelacanth (yellow) and tetrapods (red). The asterisks represent the procedure used to obtain each sequence in the phylogeny: accession number from GenBank (*), genome search (**) and genome walking (***). The scale bars indicate the average number of amino-acid substitutions per site.

Full size image

The Foxl2 branch was clearly identified from the rest of the forkhead proteins, and all the Foxl2 sequences obtained were placed in a monophyletic group, thus characterizing them as members of the Foxl2 subfamily (Figure 2a and Supplementary Material 5). Besides, the domain phylogeny showed the following: (1) the sequence of the sponge Suberites domuncula seems to be the proper outgroup for Foxl2 phylogeny and (2) there is evidence of orthology among the extra Foxl2 copies. For instance, all the new duplicated copies obtained in this study (see the topic ‘Foxl2 gene search and sequencing’), as well as the paralogs reported in other papers, were placed within the Foxl2 branch, including the diverged duplicated copies in the tetrapods T. guttata and M. domestica within a clade with the duplicated copies in T. nigroviridis and T. rubripes. Moreover, the sequence of the agnathan P. marinus that could not be inserted in the posterior Foxl2-specific phylogeny exhibited Foxl2A orthology in the domain approach (Supplementary Material 5).

A new view of Foxl2 evolution

The Foxl2 phylogeny based only on the amino-acid sequences of the forkhead domain (data not shown) did not exhibit a total resolute tree, probably because of the high conservation observed in this region, and was thus characterized by a low phylogenetic signal. In this case, based only on an almost full-length amino-acid sequence alignment, a tree with well-supported branches was generated (Figure 2b).

The first thing that is evident in the phylogeny of Foxl2 is that it is possible to split the tree into three main clades (evidenced by a black bar in the right side of Figure 2b). Based on this result, together with other findings of the present work (that is, results from syntenic region analysis), a new nomenclature for the Foxl2 gene in vertebrates is proposed. In summary, one clade includes the Foxl2 gene of nonvertebrate taxa; another includes the two distinct copies of Foxl2 from vertebrates, which we propose have evolved from the same duplication event. One copy is known in tetrapods as Foxl2 and is considered the most thoroughly studied form of the gene (here named Foxl2A), and the other copy includes the duplicated diverged form (here named Foxl2B).

The Foxl2A sequences obtained from genomes and experimental procedures, together with other sequences already known, were grouped within the Foxl2A clade of vertebrates, thus confirming their orthology (Figure 2b). Interestingly, Foxl2B copies were placed in a monophyletic group, including the newly described ones in this study (C. milii, A. burtoni, G. morhua, M. zebra, O. niloticus, L. chalumnae, T. guttata and M. domestica; Figure 2b). The Foxl2B sequences were represented as a separate clade from the Foxl2A sequences. The main point of this finding is the fact that the C. milli, L. chalumnae, T. guttata and M. domestica duplicated copies were positioned within the Foxl2B clade, which refutes the hypothesis, proposed by Jiang et al. (2011), that the paralog Foxl2 copies are a product of the duplication event exclusive of teleosts. Thus, we hypothesized that the Foxl2B copies arose from an old duplication event; however, it was necessary to perform an analysis in the syntenic region of the predicted paralog copies and compare them with the teleost copies to reinforce the orthology of the duplicated sequences.

In some cases, orthology among sequences can be inferred from synteny analysis because small syntenic blocks of the genome are frequently maintained and inherited throughout distant lineages, such as the H. sapiens and D. rerio genomes (Postlethwait et al., 2000). The present synteny analysis clearly showed that the flanking gene composition for Foxl2 copies was similar in the analyzed species. For a clearer presentation, only the common syntenic genes among the analyzed species are shown in Figure 3.

Figure 3
figure 3

Schematic representation of the syntenic regions of Foxl2 copies. Foxl2A (a) and Foxl2B (b) are positioned as the reference genes. The transcriptional orientation of the genes is depicted by the direction of the arrows. The observed genes represent only the common markers for the analyzed species and not the complete syntenic block composition. Cab39, calcium-binding protein 39; Clstn1, calsyntenin-1; Clstn2, calsyntenin-2; Copb2, coatomer protein complex subunit β2; Ctnnbip1, β-catenin-interacting protein 1; Foxl2A, isoform A of the forkhead box L2; Foxl2B, isoform B of the forkhead box L2; Hes6, hairy and enhancer of split 6; Itm2c, integral membrane protein 2C; Klhl17, Kelch-like protein 17; Lzic, leucine zipper and Ctnnbip1 domain-containing protein; Mrps22, mitochondrial 28S ribosomal protein S22; Nmnat1, nicotinamide nucleotide adenylyltransferase 1; Nmnat3, nicotinamide nucleotide adenylyltransferase 3; Noc2l, nucleolar complex protein 2 homolog; Per2, period circadian protein homolog 2; Pik3cb, β-isoform of the phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit; Pik3cd, δ-isoform of the phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit; Plekhn1, pleckstrin homology domain-containing family N member 1; Rbp2, retinol-binding protein 2; Rbp7, retinol-binding protein 7; Samd11, sterile α-motif domain-containing protein 11; Slc25a33, solute carrier family 25 member 33; Tmem201, transmembrane protein 201; Ube4b, ubiquitin conjugation factor E4 B.

Full size image

In the Foxl2A synteny analysis (Figure 3a), a common general composition as well as gene orientation was observed in all species analyzed with emphasis on Mrps22 (mitochondrial 28S ribosomal protein S22) and Pik3cb (β isoform of the phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit) that were observed near the Foxl2A copies (except Mrps22 for O. niloticus and T. nigroviridis; however, this result can be related to the limited scaffold size in the case of O. niloticus). The Foxl2A syntenic region for the fish Oryzias latipes was not analyzed here because syntenic information for this region is currently unavailable in Ensembl.

A conservation pattern was also displayed in the syntenic region of the Foxl2B copy (Figure 3b). In this case, Pik3cd and Tmem201 were both close to Foxl2B in all analyzed species, except for G. morhua, which does not show the copy of Tmem201; however, this result can also be related to the limited scaffold size. It was possible to observe some rearrangements in the orientation and repositioning of some genes. For instance, the Tmem201 gene was located upstream to Foxl2B in D. rerio, L. chalumnae, T. guttata and M. domestica, whereas it was downstream in all other analyzed species, suggesting the occurrence of rearrangements in this genomic region. Another feature observed in the Foxl2B cluster of L. chalumnae, T. guttata and M. domestica was the inversion rearrangement involving the Clstn1 and Ctnnbip1 genes.

The syntenic regions of all described Foxl2B copies, including the new ones in this study, exhibited high composition conservation, which supports the hypothesis of the orthology of these sequences, also shown by the phylogeny of Foxl2. Based on this scenario of orthology among all Foxl2B genes, the hypotheses of the duplicated copies being generated in an exclusive duplication event of teleosts (Jiang et al., 2011) can be discarded. The duplication event must have occurred in an ancestor of gnathostomes, for the Foxl2B copy to appear in chondrichthyans, teleosts, coelacanth and tetrapods (Figure 4). Posteriorly, the Foxl2A and Foxl2B copies would have suffered an additional duplication event exclusive of teleosts. In this context, the generated copies were possibly lost, as the presence of new duplicated copies, besides Foxl2A and Foxl2B, was not reported in teleosts. Indeed, the presence of a possible Foxl2A pseudogene described in S. salar, with short identity with the forkhead domain, and exhibiting a divergent N and C terminal regions when compared with Foxl2A and Foxl2B (von Schalburg et al., 2011), may represent a remnant of the Foxl2 copies generated in the exclusive whole genome duplication occurred in teleosts.

Figure 4
figure 4

Summary of the evolutionary scenario for Foxl2 copies plotted in the classical phylogeny of vertebrates. Taxa are labeled as carrying (highlighted in gray) or missing the Foxl2B copy. The asterisks (*) represent the species in which an exhaustive search for the Foxl2B copy was performed.

Full size image

A significant observation was that the Pik3cb and Pik3cd genes in the Foxl2A and Foxl2B syntenic regions, respectively, have been reported to have evolved from a duplication that gave rise to the two forms (Brown and Auger, 2011).The same study verified that both Pi3k copies remained in teleosts and tetrapods; therefore, the duplication event also occurred in an ancient ancestor in vertebrates. Taken together with our present analysis, this scenario indicates that Foxl2A and Foxl2B were not products of a single gene duplication, and considering the timing of the duplication as placed somewhere in the early lineages of vertebrates, we may speculate about this event as an outcome from one of the rounds of whole genome duplication occurred in the ancestor of vertebrates (1R) and gnathostomes (2R) (Donoghue and Purnell, 2005; Meyer and Van de Peer, 2005; Kasahara, 2007). Searches of other genes observed in the syntenic region of Foxl2, such as Clstn1 and Clstn2 (Calsyntenin-2); Rbp2 (Retinol-binding protein 2) and Rbp7 (Retinol-binding protein 7); and Nmnat1 (Nicotinamide nucleotide adenylyltransferase 1) and Nmnat3 (Nicotinamide nucleotide adenylyltransferase 3), did not reveal any related evidence associating them with a duplication event. In this context, it would be very informative to perform a phylogenetic analysis of those genes to understand their evolutionary history and their relationship to the Foxl2 and Pi3k duplication event.

An additional important observation was that no duplicated copy of Foxl2 was detected in any of the 23 eutherian genomes searched, even in the exhaustive search in 18 of those species, thus suggesting that the Foxl2B copy was lost in this group (Figure 4 and Supplementary Material 8). Foxl2B was also not detected in the amphibian species X. tropicalis, in the sauropsidian Anolis carolinensis (anole lizard), Gallus gallus (chicken) and Meleagris gallopavo (turkey), in the monotreme Ornithorhynchus anatinus (platypus) or in the methaterian species Sarcophilus harrisii. These results indicate an evolutionary process characterized by events of independent losses of Foxl2 copies during the evolutionary history of gnathostomes. However, except for eutherians, few species were analyzed, and because most of these genomes are still in an assembly process, this scenario of independent losses will only be clearly understood and confirmed with future studies.

Foxl2B activity

The scenario described in the current work for paralog copies also raises an important issue about the role of Foxl2B, especially with regard to sexual differentiation. In an expression study involving the fish O. mykiss (Baron et al., 2004), Foxl2B was shown to be expressed later than Foxl2A in ovaries. The authors suggested that this difference indicated a neofunctionalization process for Foxl2B. In another study with the fish S. salar (von Schalburg et al., 2011), Foxl2B exhibited a similar but also very distinctive pattern of expression compared with Foxl2A. This difference was observed especially in males, in which Foxl2B was expressed at higher levels in testes and extragonadal tissues when compared with Foxl2A. In females, however, Foxl2B was restricted to the gills, skin and ovary. Although the authors did not draw any conclusions about the roles of Foxl2B, the different patterns of expression in different types of tissues could indicate a subfunctionalization or a neofunctionalization process for the duplicated copies. However, only analyzing the expression pattern of the single ancestral gene, we can clearly understand the outcome of the duplicated copies of Foxl2.

Another interesting observation was based on the analysis of the polyalanine tract that is exclusive to mammals. In M. domestica, the Foxl2A protein showed 16 alanines (Supplementary Material 6), unlike the strict value of 14 alanines suggested by Cocquet et al. (2003) for all mammals. In addition, the Foxl2B protein in M. domestica did not exhibit the polyalanine tract, which may indicate a change in the protein function compared with the Foxl2A protein. However, how Foxl2B adapted its function during the evolution of gnathostomes remains unknown, pending additional functional studies.

Conclusion

The broad study with the domain approach and the incorporation of new Foxl2 duplicated sequences in the inferred phylogeny provided the basis for a novel discussion of evolutionary studies of Foxl2. The current study identified duplicated forms of Foxl2 in different vertebrate groups, which suggests the need for a revision of the nomenclature for Foxl2A and Foxl2B. Notably, the established evolutionary analysis strongly supported the conclusion that the predicted Foxl2B copies in chondrichthyans, coelacanth and tetrapods are indeed orthologous to all Foxl2B copies of teleosts. Furthermore, the duplication event that gave rise to the paralog copies was not unique to teleosts and was a common event that occurred in an ancestor gnathostome. Our results also suggest that this duplication was not exclusive to Foxl2. From the description of new Foxl2B copies in more fish species, obtained by genome searches, it is reasonable to believe that there are more teleost species, not yet reported, that carry a Foxl2B copy. In this context, it remains unknown whether most species in this group carry this additional copy, whereas others have lost it, or if all teleost species have the duplicated version of Foxl2. The distribution of Foxl2B was shown to be diversified in tetrapods, with a strong indication of loss of this copy in eutherians. In addition to this newly reported scenario, more information is required for understanding the role of the Foxl2B copy in vertebrates, especially in relation to sexual differentiation.

Data archiving

Sequence data have been submitted to GenBank: accession numbers JX012129, JX012130 and JX012131.