Enhancer evolution in chordates: Lessons from functional analyses of cephalochordate cis-regulatory modules

Chordates comprise three major groups, cephalochordates (amphioxus), tunicates (urochordates), and vertebrates. Since cephalochordates were the early branching group, comparisons between amphioxus and other chordates help us to speculate about ancestral chordates. Here, I summarize accumulating data from functional studies analyzing amphioxus cis-regulatory modules (CRMs) in model systems of other chordate groups, such as mice, chickens, clawed frogs, fish, and ascidians. Conservatism and variability of CRM functions illustrate how gene regulatory networks have evolved in chordates. Amphioxus CRMs, which correspond to CRMs deeply conserved among animal phyla, govern reporter gene expression in conserved expression domains of the putative target gene in host animals. In addition, some CRMs located in similar genomic regions (intron, upstream, or downstream) also possess conserved activity, even though their sequences are divergent. These conservative CRM functions imply ancestral genomic structures and gene regulatory networks in chordates. However, interestingly, if expression patterns of amphioxus genes do not correspond to those of orthologs of experimental models, some amphioxus CRMs recapitulate expression patterns of amphioxus genes, but not those of endogenous genes, suggesting that these amphioxus CRMs are close to the ancestral states of chordate CRMs, while vertebrates/tunicates innovated new CRMs to reconstruct gene regulatory networks subsequent to the divergence of the cephalochordates. Alternatively, amphioxus CRMs may have secondarily lost ancestral CRM activity and evolved independently. These data help to solve fundamental questions of chordate evolution, such as neural crest cells, placodes, a forebrain/midbrain, and genome duplication. Experimental validation is crucial to verify CRM functions and evolution.     

Retinoids and nonvertebrate chordate development

Retinoic acid (RA) is required for the differentiation and morphogenesis of chordate-specific features, such as the antero-posterior regionalization of the dorsal hollow nerve cord and neural crest cells. RA receptors (RARs) have been reported exclusively in chordates, suggesting that the acquisition of the RAR gene was important for chordate evolution. A scenario is presented here for the establishment of an RAR-mediated developmental regulatory system during the course of chordate evolution. In the common chordate ancestor, RAR came to control the spatial expression pattern of Hox genes in the ectoderm and endoderm along the antero-posterior axis. In these germ layers, RA was required for the differentiation of epidermal sensory neurons and the morphogenesis of pharyngeal gill slits, respectively. As the diffuse epidermal nerve net in the chordate ancestor became centralized to form the dorsal nerve cord, the epidermal Hox expression pattern was carried into the central nervous system. Because the Hox code here came to specify neuronal identity along the antero-posterior axis, RA became inextricably linked to the antero-posterior patterning of the chordate central nervous system.     

Regulatory genes in the ancestral chordate genomes

Changes or innovations in gene regulatory networks for the developmental program in the ancestral chordate genome appear to be a major component in the evolutionary process in which tadpole-type larvae, a unique characteristic of chordates, arose. These alterations may include new genetic interactions as well as the acquisition of new regulatory genes. Previous analyses of the Ciona genome revealed that many genes may have emerged after the divergence of the tunicate and vertebrate lineages. In this paper, we examined this possibility by examining a second non-vertebrate chordate genome. We conclude from this analysis that the ancient chordate included almost the same repertory of regulatory genes, but less redundancy than extant vertebrates, and that approximately 10% of vertebrate regulatory genes were innovated after the emergence of vertebrates. Thus, refined regulatory networks arose during vertebrate evolution mainly as preexisting regulatory genes multiplied rather than by generating new regulatory genes. The inferred regulatory gene sets of the ancestral chordate would be an important foundation for understanding how tadpole-type larvae, a unique characteristic of chordates, evolved.     

The amphioxus rab GDP-dissociation inhibitor (GDI) gene is neural-specific: implications for the evolution of chordate rab GDI genes.

The rab GDP-dissociation inhibitor (rab GDI) proteins are involved in the regulation of vesicle-mediated cellular transport. We isolated the amphioxus rab GDI gene, analyzed its expression during amphioxus development, and performed a phylogenetic analysis of the rab GDI family. In contrast to the two major rab GDI forms in mammals, the alpha and beta forms, there is only one rab GDI isoform in amphioxus. Our analysis indicates that the occurrence of the alpha and beta forms of rab GDI preceded the divergence of lineages leading to birds and mammals, and that the amphioxus rab GDI may have evolved directly from the common ancestor of both forms. While the mammalian rab GDI beta-genes are ubiquitously expressed, the rab GDI alpha genes are predominantly expressed in neural tissues. The expression analysis of the amphioxus rab GDI gene shows predominantly neural expression similar to that of the mammalian rab GDI alpha form, suggesting that the ancestral expression pattern of chordate rab GDI was neural. In addition, the chicken rab GDI beta-like gene also shows neural-specific expression, which indicates that the neural expression was retained in both early postduplication alpha and beta isoforms and that a novel function associated with ubiquitous expression may have evolved uniquely in mammals. These results reveal a likely scenario of functional divergence of the rab GDI genes after duplication of the ancestral gene. A similar pattern of evolution, in which one of the duplicated genes retained a role similar to that of the ancestral one while other genes were recruited into novel roles, was also observed in the analysis of chordate Otx and hedgehog genes. In the rab GDI, hedgehog, and Otx gene families, the gene retaining the ancestral role shows a lower rate of sequence evolution than its counterpart, which was recruited for a novel function.     

Brachyury (T) expression in embryos of a larvacean urochordate, Oikopleura dioica, and the ancestral role of T

The Brachyury, or T, gene is required for notochord development in animals occupying all three chordate subphyla and probably also had this role in the last common ancestor of the chordate lineages. In two chordate subphyla (vertebrates and cephalochordates), T is also expressed during gastrulation in involuting endodermal and mesodermal cells, and in vertebrates at least, this expression domain is required for proper development. In the basally diverging chordate subphylum Urochordata, animals in the class Ascidiacea do not employ T during gastrulation in endodermal or nonaxial mesodermal cells, and it has been suggested that nonnotochordal roles for T were acquired in the cephalochordate-vertebrate lineage after it split with Urochordata. To test this hypothesis, we cloned T from Oikopleura dioica, a member of the urochordate class Appendicularia (or Larvacea), which diverged basally in the subphylum. Investigation of the expression pattern in developing Oikopleura embryos showed early expression in presumptive notochord precursor cells, in the notochord, and in parts of the developing gut and cells of the endodermal strand. We conclude that the ancestral role of T likely included expression in the developing gut and became necessary in chordates for construction of the notochord.     

Ligand specificities of recombinant retinoic acid receptors RAR alpha and RAR beta.

Binding of retinoic acid (RA) to specific RA receptors alpha and beta (RAR alpha and RAR beta) was studied. Receptors were obtained in two ways: (1) full-length receptors were produced by transient expression of the respective human cDNAs in COS 1 cells; and (2) the ligand-binding domains of RAR alpha and RAR beta were produced in Escherichia coli. RA binding to the wild-type and truncated forms of the receptor was identical for both RAR alpha and RAR beta, indicating that the ligand-binding domains have retained the binding characteristics of the intact receptors. Furthermore, RA bound with the same affinity to both RAR alpha and RAR beta. Only retinoid analogues with an acidic end-group were able to actively bind to both receptors. On measuring the binding of various retinoids, we have found that the properties of the ligand-binding sites of RAR alpha and RAR beta were rather similar. Two retinoid analogues were capable of binding preferentially to either RAR alpha or RAR beta, suggesting that it may be possible to synthesize specific ligands for RAR alpha and RAR beta.     

Evolution of developmental roles of Pax2/5/8 paralogs after independent duplication in urochordate and vertebrate lineages

Background

Gene duplication provides opportunities for lineage diversification and evolution of developmental novelties. Duplicated genes generally either disappear by accumulation of mutations (nonfunctionalization), or are preserved either by the origin of positively selected functions in one or both duplicates (neofunctionalization), or by the partitioning of original gene subfunctions between the duplicates (subfunctionalization). The Pax2/5/8 family of important developmental regulators has undergone parallel expansion among chordate groups. After the divergence of urochordate and vertebrate lineages, two rounds of independent gene duplications resulted in the Pax2, Pax5, and Pax8 genes of most vertebrates (the sister group of the urochordates), and an additional duplication provided the pax2a and pax2b duplicates in teleost fish. Separate from the vertebrate genome expansions, a duplication also created two Pax2/5/8 genes in the common ancestor of ascidian and larvacean urochordates.

Results

To better understand mechanisms underlying the evolution of duplicated genes, we investigated, in the larvacean urochordate Oikopleura dioica, the embryonic gene expression patterns of Pax2/5/8 paralogs. We compared the larvacean and ascidian expression patterns to infer modular subfunctions present in the single pre-duplication Pax2/5/8 gene of stem urochordates, and we compared vertebrate and urochordate expression to infer the suite of Pax2/5/8 gene subfunctions in the common ancestor of olfactores (vertebrates + urochordates). Expression pattern differences of larvacean and ascidian Pax2/5/8 orthologs in the endostyle, pharynx and hindgut suggest that some ancestral gene functions have been partitioned differently to the duplicates in the two urochordate lineages. Novel expression in the larvacean heart may have resulted from the neofunctionalization of a Pax2/5/8 gene in the urochordates. Expression of larvacean Pax2/5/8 in the endostyle, in sites of epithelial remodeling, and in sensory tissues evokes like functions of Pax2, Pax5 and Pax8 in vertebrate embryos, and may indicate ancient origins for these functions in the chordate common ancestor.

Conclusion

Comparative analysis of expression patterns of chordate Pax2/5/8 duplicates, rooted on the single-copy Pax2/5/8 gene of amphioxus, whose lineage diverged basally among chordates, provides new insights into the evolution and development of the heart, thyroid, pharynx, stomodeum and placodes in chordates; supports the controversial conclusion that the atrial siphon of ascidians and the otic placode in vertebrates are homologous; and backs the notion that Pax2/5/8 functioned in ancestral chordates to engineer epithelial fusions and perforations, including gill slit openings.     

Unexpectedly large number of conserved noncoding regions within the ancestral chordate Hox cluster

The single amphioxus Hox cluster contains 15 genes and may well resemble the ancestral chordate Hox cluster. We have sequenced the Hox genomic complement of the European amphioxus Branchiostoma lanceolatum and compared it to the American species, Branchiostoma floridae, by phylogenetic footprinting to gain insights into the evolution of Hox gene regulation in chordates. We found that Hox intergenic regions are largely conserved between the two amphioxus species, especially in the case of genes located at the 3' of the cluster, a trend previously observed in vertebrates. We further compared the amphioxus Hox cluster with the human HoxA, HoxB, HoxC, and HoxD clusters, finding several conserved noncoding regions, both in intergenic and intronic regions. This suggests that the regulation of Hox genes is highly conserved across chordates, consistent with the similar Hox expression patterns in vertebrates and amphioxus.     

Origins of anteroposterior patterning and Hox gene regulation during chordate evolution

All chordates share a basic body plan and many common features of early development. Anteroposterior (AP) regions of the vertebrate neural tube are specified by a combinatorial pattern of Hox gene expression that is conserved in urochordates and cephalochordates. Another primitive feature of Hox gene regulation in all chordates is a sensitivity to retinoic acid during embryogenesis, and recent developmental genetic studies have demonstrated the essential role for retinoid signalling in vertebrates. Two AP regions develop within the chordate neural tube during gastrulation: an anterior 'forebrain-midbrain' region specified by Otx genes and a posterior 'hindbrain-spinal cord' region specified by Hox genes. A third, intermediate region corresponding to the midbrain or midbrain-hindbrain boundary develops at around the same time in vertebrates, and comparative data suggest that this was also present in the chordate ancestor. Within the anterior part of the Hox-expressing domain, however, vertebrates appear to have evolved unique roles for segmentation genes, such as Krox-20, in patterning the hindbrain. Genetic approaches in mammals and zebrafish, coupled with molecular phylogenetic studies in ascidians, amphioxus and lampreys, promise to reveal how the complex mechanisms that specify the vertebrate body plan may have arisen from a relatively simple set of ancestral developmental components.     

Evolution of new characters after whole genome duplications: Insights from amphioxus

Additional copies of genes resulting from two whole genome duplications at the base of the vertebrates have been suggested as enabling the evolution of vertebrate-specific structures such as neural crest, a midbrain/hindbrain organizer and neurogenic placodes. These structures, however, did not evolve entirely de novo, but arose from tissues already present in an ancestral chordate. This review discusses the evolutionary history of co-option of old genes for new roles in vertebrate development as well as the relative contributions of changes in cis-regulation and in protein structure. Particular examples are the FoxD, FGF8/17/18 and Pax2/5/8 genes. Comparisons with invertebrate chordates (amphioxus and tunicates) paint a complex picture with co-option of genes into new structures occurring both after and before the whole genome duplications. In addition, while cis-regulatory changes are likely of primary importance in evolution of vertebrate-specific structures, changes in protein structure including alternative splicing are non-trivial.     

The emergence and loss of gonadotropin-releasing hormone in protostomes: orthology, phylogeny, structure, and function

Gonadotropin-releasing hormone (GNRH) is a neuropeptide critical for reproductive activation and maintenance in vertebrates. The recent elucidation of molluscan GNRH-like sequences led to several important questions regarding the evolution of the GNRH family. For instance, are molluscan and chordate GNRHs true orthologs? Has GNRH been retained in most protostomian lineages? What was the function of the ancestral GNRH? The goal of this review is to provide a critical analysis of GNRH evolution based on data available from the known forms of protostomian GNRH. Judging from the orthology between chordate and protostomian GNRH receptors, conservation of several structural motifs on the GNRH peptide, and exon/intron arrangement conserved between protostomian and chordate GNRH genomic sequences, we conclude that chordate and protostomian GNRHs likely share a common ancestor. Based on our analysis of phylogenetic distribution, we also hypothesize that GNRH may have been lost in the ecdysozoan lineage but preserved in lophotrochozoans. Lastly, we propose that the ancestral function of GNRH is to serve as a general neural regulator, and its considerable specialization in reproduction seen in chordates is a consequence of neofunctionalization following gene duplication.     

The search for non-chordate retinoic acid signaling: lessons from chordates

Signaling by retinoic acid (RA) is an important pathway in the development and homeostasis of vertebrate and invertebrate chordates, with a critical role in mesoderm patterning. Classical studies on the distribution of nuclear receptors of animals suggested that the family of RA receptors (RARs/NR1B) was restricted to chordates, while the family of RA X receptors (RXR/NR2B) was distributed from cnidarians to chordates. However, the accumulation of data from genome projects and studies in non-model species is questioning this traditional view. Here we discuss the evidence for non-chordate RA signaling systems in the light of recent advances in our understanding of carotene (pro-Vitamin A) metabolism and of the identification of potential RARs and members of the NR1 family in echinoderms and lophotrochozoan trematodes, respectively. We conclude, as have others before (Bertrand et al., 2004. Mol Biol Evol 21(10):1923-1937), that signaling by RA is more likely an ancestral feature of bilaterians than a chordate innovation.     

Retinoic acid receptor isoform RAR gamma 1: an antagonist of the transactivation of the RAR beta RARE in epithelial cell lines and normal human keratinocytes.

The diversity of isoforms of retinoic acid (RA) receptors (RARs) and of DNA sequences of retinoic acid-responsive elements (RAREs) suggests the existence of selectivities in the RAR/RARE recognition or in the subsequent gene modulation. Such selectivities might be particularly important for RAREs involved in positive feedback, eg. the RAR beta RARE. In the present work we found that in several epithelial cell lines, reporter constructs containing the RAR beta RARE linked to the HSV-tk promoter were transactivated in the presence of RA by endogenous RARs and co-transfected RAR alpha 1 and RAR beta 2 isoforms, but not by RAR gamam 1. On the contrary, this latter isoform behaved towards the RAR beta RARE as an inhibitor of the transactivation produced by endogenous RARs and by cotransfected RAR alpha 1 and RAR beta 2. RAR gamma 1 also behaved as an antagonist of the transactivation produced by cotransfected RXR alpha. The natural RAR beta gene promoter or RAR beta RARE tk constructs were not activated by the endogenous receptors of normal human keratinocytes (NHK), which are known to contain predominantly RAR gamma 1. It was, however, possible to activate to a certain extent RAR beta RARE-reporter constructs in NHK by co-transfecting RAR alpha 1, RAR beta 2 or RXR alpha. The antagonist behavior of RAR gamma 1 towards the RAR beta RARE may explain why in certain cell types such as keratinocytes, RAR beta is neither expressed nor induced by RA.     

Retinoic acid resistance of the variant embryonal carcinoma cell line RAC65 is caused by expression of a truncated RARα

P19 embryonal carcinoma (EC) cells differentiate when treated with retinoic acid (RA). The P19 EC-derived mutant cell line RAC65 is resistant to the differentiation-inducing activity of RA. We show that these cells express a truncated retinoic acid receptor alpha(mRAR alpha-RAC65), probably due to the integration of a transposon-like element in the RAR alpha gene. This receptor lacks 71 C-terminal amino acids and terminates in the ligand-binding domain. In CAT assays in RAC65 cells, mRAR alpha-RAC65 fails to trans-activate the RAR beta promoter, which contains a RA-response element. In wild-type P19 EC cells mRAR alpha-RAC65 functions as a dominant-negative repressor of RA-induced RAR beta activation. Gel retardation assays demonstrate that mRAR alpha-RAC65 is still able to bind to the RA-response element of the RAR beta promoter, indicating that competition with functional RARs for the same binding site leads to the observed dominant-negative effect. In addition, in two RAC65 clones in which wild-type hRAR alpha was stably transfected RA-sensitivity was restored and in one RAR beta expression could be induced by RA. Taken together, these data show that the primary cause of RA-resistance of RAC65 cells is the expression of a defective RAR alpha, which prevents the trans-activation of RA-responsive genes and results in a loss of the ability to differentiate.     

Essential role of Bmp signaling and its positive feedback loop in the early cell fate evolution of chordates

In chordates, early separation of cell fate domains occurs prior to the final specification of ectoderm to neural and non-neural as well as mesoderm to dorsal and ventral during development. Maintaining such division with the establishment of an exact border between the domains is required for the formation of highly differentiated structures such as neural tube and notochord. We hypothesized that the key condition for efficient cell fate separation in a chordate embryo is the presence of a positive feedback loop for Bmp signaling within the gene regulatory network (GRN), underlying early axial patterning. Here, we therefore investigated the role of Bmp signaling in axial cell fate determination in amphioxus, the basal chordate possessing a centralized nervous system. Pharmacological inhibition of Bmp signaling induces dorsalization of amphioxus embryos and expansion of neural plate markers, which is consistent with an ancestral role of Bmp signaling in chordate axial patterning and neural plate formation. Furthermore, we provided evidence for the presence of the positive feedback loop within the Bmp signaling network of amphioxus. Using mRNA microinjections we found that, in contrast to vertebrate Vent genes, which promote the expression of Bmp4, amphioxus Vent1 is likely not responsible for activation of cephalochordate ortholog Bmp2/4. Cis-regulatory analysis of amphioxus Bmp2/4, Admp and Chordin promoters in medaka embryos revealed remarkable conservation of the gene regulatory information between vertebrates and basal chordates. Our data suggest that emergence of a positive feedback loop within the Bmp signaling network may represent a key molecular event in the evolutionary history of the chordate cell fate determination.