Introduction

Emergence of the tadpole-like body plan in an ancestral chordate is still a big mystery in the evolutionary biology. To know what happened in the genome of the ancestral chordate for acquiring the chordate body plan, regulatory genes in the tunicate genome that encode transcription factors and signaling ligands have been comprehensively identified (Hino et al. 2003; Satou et al. 2003a, b; Wada et al. 2003; Yagi et al. 2003; Yamada et al. 2003) because the tunicates and vertebrates are a sister group within chordates (Delsuc et al. 2006). The regulatory gene set identified in these previous studies may comprise that possessed by an ancestral chordate, most part of which is shared with vertebrates. Vertebrate genomes contain additional regulatory genes not found in the tunicate genome, such as Tbx4/5, Gbx, and wnt11 (Dehal et al. 2002). Some are a result of gene loss in the tunicate genome, as protostomes and vertebrates share clear orthologs (e.g., Wnt4). Others appear to have rapidly evolved in Ciona, making their phylogenetic positions obscure. The remaining may be genes that were innovated in the vertebrate lineage. Such classification is apparently essential for understanding of the origin of chordates. Recent decoding of the amphioxus genome gave us an opportunity to solve this problem because the cephalochordates are thought to be the most basal group within chordates. In this study, we compare the major regulatory gene families of chordate genomes and focus on genes that were either not present or regarded as ‘orphan’ genes in the Ciona genome.

Materials and methods

Regulatory genes in the genome of Branchiostoma floridae (Putnam et al. 2008) were identified with the TBLASTN program (Altschul et al. 1990) using the Ciona genes and the human genes as queries. The best-fit gene models that were decided on the basis of homology and expressed sequence tag evidences were employed if several candidate gene models were predicted for a single locus. If no gene models were available, the corresponding genome sequences were manually translated. These models were subjected to reciprocal BLAST analyses against the human and Ciona proteomes (threshold E value = 1E−05). Those that displayed reciprocal best-hit relationships both between the amphioxus and human proteins and between the amphioxus and Ciona proteins were considered most likely orthologs of these human and Ciona proteins, as the phylogenetic relationships between regulatory proteins of human and Ciona have been extensively examined (Hino et al. 2003; Satou et al. 2003a, b; Wada et al. 2003; Yagi et al. 2003; Yamada et al. 2003). The proteome sets are IPI ver3.24 for human and JGI ver1.0 for Ciona. Several subfamilies consist of the single members, and members of each subfamily contain a unique domain(s). In these cases, their orthologies are obvious by alignments, and therefore, we did not perform further molecular phylogenetic analyses. Id/Emc (group-D bHLH), EBF/COE (group-F bHLH), and some homeobox subfamilies are included in this category (ESM Fig. S6). Six orphan homeodomain proteins (70306/102512, 86186/124544, 99353, 102805, 106421, and 107468/124758) were highly diverged, and comparable sequences were not found in any known genomes. No gene models for a putative Branchiostoma ortholog of HMG2L included an HMG-box probably because of a sequence gap, and therefore, this protein was not included in the molecular phylogenetic tree. However, overall high similarities of this Branchiostoma protein with Ciona and human HMG2L protein suggested that this Branchiostoma protein is a bona fide ortholog of HMG2L, and therefore, it is tentatively included in ESM Table S1. For all of the other protein subfamilies, molecular phylogenetic analyses were performed. In molecular phylogenetic analysis, fly proteins was basically included for comparison, and, if necessary, proteins of other animals were included. Especially for genes whose clear counterparts were not found in the representative animal genomes, we sampled as many proteins as possible from the public protein database and we considered them as candidate orthologs. The T-Coffee program was employed for generating alignments (Notredame et al. 2000). The alignments of conserved functional domains (e.g., homeodomains for homeobox proteins and bHLH domains for bHLH proteins) were used for constructing phylogenetic trees. Maximum likelihood trees were constructed with the PHYML program (Guindon and Gascuel 2003) with the WAG amino acid substitution matrix (Whelan and Goldman 2001), the proportion of invariant sites calculated from the alignment, and four rate categories with a gamma distribution parameter estimated from the data. Trees were tested with 100 bootstrap pseudoreplicates. Bayesian trees were constructed by the MRBAYES program (ver 3.1.2; Ronquist and Huelsenbeck 2003). Two independent runs were conducted (each with four chains) until the average standard deviation of split frequencies becomes less than 0.01 or until the two runs converge onto the stationary distribution. In cases that no clear orthologs of mammalian genes were found in the Ciona or Branchiostoma genome, we similarly examined whether their orthologous genes are found in the protostomes. In cases that no clear orthologs of Ciona and Branchiostoma genes were found in the human genome, we also similarly examined whether their orthologous genes are found in the genomes of mouse, Xenoups tropicalis and zebrafish. Molecular phylogenetic trees constructed in the present study are shown in ESM Figs. S1–S5. Multiple Branchiostoma gene models often encode very similar polypeptides. In most cases, it was hard to distinguish whether each of these genes represents haplotypes or whether these genes are due to assembly artifacts or are actually encoded in different loci (Putnam et al. 2008). Therefore, the number of the gene models shown in Table 2 is not equal to the total gene number of the regulatory genes.

Results and discussion

Regulatory genes encoding transcription factors and signaling molecules in major families were comprehensively surveyed on the amphioxus genome. The genes identified were classified based on results of molecular phylogenetic analyses (ESM Table S1). Comparisons of major transcription factor families and signaling ligands/receptors encoded in the Ciona and amphioxus genomes are summarized in Tables 1 and 2.

Table 1 Numbers of subfamilies of transcription factor genes encoded in the genomes of human, Ciona, and Branchiostoma
Full size table
Table 2 Regulatory genes potentially acquired or lost during the evolution of chordates
Full size table

Only 26 gene families of the 311 examined were putatively vertebrate-specific (or partly mammalian or amniote-specific), as both of the amphioxus and Ciona genomes did not have the counterparts. These putative vertebrate-specific genes include GDF2/BMP10, DDIT3 (bZIP transcription factor), SRY (SoxA), and FGF19 (Table 2). While most of these regulatory genes are represented in amphibians and fish, some genes, such as SOHLH1/SOHLH2 (bHLH transcription factor) and Fkhl18 (Fox transcription factor), lack an apparent ortholog in the genomes of amphibians and fish. These ‘missing’ genes are candidates for regulatory elements that emerged in the vertebrate evolutionary process.

There are a few genes found in the human and Ciona genomes but not in the amphioxus genome. These include ACSCL3, MESP1, MNT, BACH1, VDR, EFNA1, and genes in three FGF subfamilies. However, because there are orthologs for ACSCL3, MESP1, MNT, and VDR in protostomes, these genes appear to have been lost in the amphioxus lineage. In contrast, other genes, such as BACH1 (bZIP transcription factor) and EFNA (ephrin-A), are candidates that the common ancestors of Ciona and vertebrates acquired after the divergence of amphioxus and tunicates/vertebrates. In addition to BACH1 and EFNA, we could not find any clear orthologs for three Fgf subfamilies in the amphioxus lineage. However, there are at least four ‘orphan’ Fgf genes in the Branchiostoma genome. It is possible that some of them belong to these subfamilies, but such orthologous relationships are hidden because of their rapid evolution in the amphioxus lineage. Of course, it is also possible that these Fgfs are ‘orphans’ and the amphioxus genome lacks orthologs for these FGFs.

Two types of ephrins are present in vertebrates; Ephrin-A proteins are GPI-anchored to the cell membrane and Ephrin-B proteins are transmembrane proteins (Kullander and Klein 2002). In protostomes, the genome of Drosophila melanogaster does not contain any GPI-anchored ephrins, while the genome of Caenorhabditis elegans includes four potential GPI-anchored ephrins (Kullander and Klein 2002). However, molecular phylogenetic studies indicate that these protostome ephrins are not closely related to vertebrate ephrin-A or ephrin-B (Satou et al. 2003b). Thus, vertebrate ephrin-A and ephrin-B emerged after the split of protostomes and deuterostomes. The Ciona genome contains clear orthologs both of ephrin-A and ephrin-B (Satou et al. 2003b), while the amphioxus genome contains clear orthologs of ephrin-B, but not for ephrin-A. Therefore, it is likely that the last common ancestor of Ciona and vertebrates included an ephrin-A gene, but the common ancestor of all chordates did not. There are two types of genes encoding ephrin receptors (Eph), each of which binds to ephrin-A or ephrin-B. Molecular phylogenetic trees suggest that the amphioxus Ephs are more similar to those of nematode, insects, and sea urchins. This observation is consistent with lack of clear orthologs of ephrin-A in the amphioxus genome. However, it is possible that these genes were not included in the current assembly of the amphioxus genome or that these genes were lost in the amphioxus lineage.

There are a few genes that are found in the amphioxus and protostome genomes but not in the vertebrate and tunicate genomes. These include the homeobox genes rough, reversed polarity, NK7.1, pox-neuro, defective proventriculus, and a nuclear receptor, NR5B. The nuclear receptor family NR5 comprises two subfamilies, NR5A and NR5B. Genes belonging to the NR5A subfamily (NR5A1 and NR5A2 in human) are present in genomes from protostomes to vertebrates as well as in the Ciona and amphioxus genomes. In contrast, NR5B genes have only been observed in protostome genomes, while an ortholog of this gene is present in the amphioxus genome (Fig. 1). Therefore, it is likely that the ancestral chordate genome included this gene but it was lost after the split of amphioxus and tunicates/vertebrates in the tunicate/vertebrate lineage. We could not identify gene families that were present in the Ciona and protostome genomes but not in the vertebrate and amphioxus genomes.

Fig. 1
figure 1

A molecular phylogenetic tree of nuclear receptors (NR5 and NR6 subfamilies) constructed by the maximum likelihood method. Four numbers of each of three major nodes indicate bootstrap values by the maximum likelihood, neighbor-joining, maximum parsimony methods, and posterior probabilities by the Bayesian method. All of four methods confirm that the amphioxus genome has a gene encoding a protein belonging to the NR5B subfamily

Full size image

An ortholog of an ‘orphan’ Fox gene present in the Ciona genome, Ci-orphan-Fox5, is present in the Branchiostoma genome. Molecular phylogenetic analyses revealed that this gene has a potential counterpart in insect (D. melanogaster) and fish (Fugu and Medaka) genomes but not in higher vertebrates (ESM Figs. S1 and S2). Although the function of this gene is unknown, its distribution may provide a clue for understanding vertebrate evolution. Similarly, Nkx-C was first identified in the Ciona genome and is present in the fly and amphioxus genome but not in vertebrate genomes. Therefore, these genes emerged before the split of deuterostomes and protostomes, but were lost in the vertebrate lineage. In comparison, three genes, namely Nkx-A, ADMP, and Twist-like are not found in protostome genomes. While ADMP is present in fugu and Xenopus genomes, Nkx-A and Twist-like are found only in the Ciona and amphioxus genomes. The Twist-like genes in these species are orthologs, but this relationship does not extend to Twist proteins of vertebrates and protostomes (ESM Fig. S5), although Ciona Twist-like has a similar function to Twist (Imai et al. 2003) and both of the Ciona and Branchiostoma genomes lack a clear Twist ortholog.

In spite of extensive molecular phylogenetic analyses, orphan regulatory genes remain both in the tunicate genome and in the amphioxus genome. Some of these genes may have been rapidly evolved in these species and thus obscured their similarity to regulatory genes in other organisms. Alternatively, the ancestral chordate may have possessed these orphan genes, but they were lost during subsequent evolution. In some cases, syntenic relationships provide strong evidence for orthologous relationships that are otherwise impossible to resolve by the molecular phylogeny. For example, the orphan TGFβ gene, called as TGFβ-NA1(OrphanTGFβ-1), is encoded in the first intron of another TGFβ gene (Ci-BMP2/4, an ortholog of vertebrate BMP2 and BMP4) in the Ciona genome (Fig. 2). This gene organization may be ancestral situation within deuterostomes. Previous neighbor-joining molecular phylogenetic analysis failed to assign Ciona TGFβ-NA1 to a specific subfamily (Hino et al. 2003), and the present maximum likelihood and Bayesian trees weakly indicated that this orphan TGFβ is grouped with sea urchin Univin, zebrafish DVR-1, frog VG1, and human GDF1 and GDF3 (ESM Figs. S3 and S4). In the present survey on the amphioxus genome, we discovered a TGFβ gene (tentatively named GDF1/3-like2) encoded in the first intron of BMP2/4 gene in this species (Fig. 2). Similar genetic organization is present in the sea urchin and zebrafish genomes. Univin, a BMP2/4 related TGFβ, is found to be a 5′-neighbor of BMP2/4 in the sea urchin genome (Fig. 2; Sodergren et al. 2006). DVR-1 is a 5′-neighbor of BMP2a in the zebrafish genome (Fig. 2). However, it is possible that there are additional exons of BMP2 that have not yet been identified because of insufficient experimental data, and univin and DVR-1 are also encoded in the first intron of BMP2/4. Additionally, the gene 5′ to the amphioxus GDF1/3-like2/BMP2/4 is an ortholog of human FBXW7. The Ciona ortholog of FBXW7 is also the 5′ neighbor of these two TGFβ genes. The gene 3′ to these TGFβ genes in the amphioxus genome encodes a receptor tyrosine kinase. The best candidate in the sea urchin genome for the ortholog of this amphioxus RTK gene was 5′ neighbor of univin. This genetic organization supports the conclusion that these genes, sea urchin univin, Ciona TGFβ-NA1, Branchiostoma GDF1/3-like2, and zebrafish DVR-1, are orthologous for each other.

Fig. 2
figure 2

A possible origin of GDF1/GDF3, or VG1, and Lefty. GDF1/GDF3 and the predicted orthologs are depicted in red. BMP2/BMP4 and its orthologs are depicted in black. Cyan and blue arrows represent genes encoding an F-box/WD-repeated protein and a receptor tyrosine kinase, respectively. The synteny of these genes was partially conserved in the tunicate and sea urchin genome

Full size image

This genetic organization also implies that the founder gene of the subfamily including univin, Ciona TGFβ-NA1, Branchiostoma GDF1/3-like2, and zebrafish DVR-1 arose by gene duplication of BMP2/4 at the emergence of deuterostome. This hypothesis is supported by the molecular phylogenetic trees that imply a close relation between univin/TGFβ-NA1/GDF1/3-like2/DVR-1 and BMP2/BMP4 (ESM Figs. S3 and S4). Zebrafish DVR-1 is an ortholog of frog VG1/Derrier and human GDF1/GDF3, which are not located near BMP2 or BMP4 in these species (Fig. 2). The founder gene arisen by gene duplication of BMP2/4 might be translocated in the most vertebrate genomes.

An additional ‘orphan’ TGFβ was also identified in the amphioxus genome (tentatively named as GDF1/3-like1). This gene product is most similar to GDF1/3-like2 and secondarily to BMP2/4 (ESM Figs. S3 and S4). This gene is located next to Lefty, a divergent member of the TGFβ superfamily encoding an antagonist for another TGFβ, Nodal (Fig. 2). Because Lefty and Nodal are found only in the deuterostome genomes, these genes likely evolved from the other TGFβ at the emergence of the ancestral deuterostome. Previous studies demonstrate that acquisition and evolution of the Nodal system is one of the key events of deuterostome evolution (Duboc et al. 2004). Although molecular phylogenetic analyses failed to identify the origins of Lefty, it is very unlikely that these two related genes, GDF1/3-like1 and Lefty, are located next to each other by chance. It is also unlikely that some evolutionary constraint affects this gene pair exclusively in this species. Therefore, we propose that Lefty arose by two duplications of the ancestral GDF1/3 or by a duplication of BMP2/4-GDF1/3 gene pairs.

The counterparts for most of orphan Ciona genes are not found in the amphioxus genome (ESM Table S1). For example, there is a gene called Notrlc in the Ciona genome. This gene is similar to Hand, but a bona fide Hand ortholog is encoded at a distinct locus. The absence of the ortholog of the Ciona orphan genes in the Branchiostoma genome suggests that most of these genes evolved or arose after the split of the tunicate and vertebrate lineages, although it cannot be denied that the ancestral chordate had some part of these genes and Branchiostoma and vertebrates lost them independently.

Many regulatory genes not present in the Ciona genome were identified in the amphioxus genome. On the other hand, 46 genes in 26 regulatory gene families were not discovered in the Ciona and Branchiostoma genomes, although it is still possible that a fraction of these ‘missing’ genes are included in unsequenced genomic regions. In summary, the ancestral chordate appears to have had a repertory of regulatory genes similar to that of modern vertebrates, but it is possible that approximately 10% of regulatory genes in the vertebrate genome were generated after the emergence of vertebrates. Together with the emergence of these regulatory genes, the emergence of paralogous genes for each subfamily in vertebrate genome (2.3 on average) have undoubtedly made important contributions to the evolution of modern vertebrates. As the Ciona genome has lost many regulatory genes, the regulatory gene set for the developmental program required to form tadpole type larva may be smaller than the gene set found in Ciona and Branchiostoma genomes.