Four twist genes in zebrafish, four expression patterns

Abstract

Twist genes code for regulatory bHLH proteins essential for embryonic development and conserved across the metazoa. There are four genes that constitute the zebrafish twist family: twist1a, twist1b, twist2, orthologs of the mammalian twist1 and twist2 genes; and twist3—a gene from a new clade that does not exist in mammals. Presented here are their embryonic mRNA expression profiles. The study extends the known conservation of twist developmental patterns in tetrapods to the fish, e.g., expression in cephalic neural crest, sclerotome and lateral plate mesoderm. Some other expression domains are unique, like hypochord and dorsal aorta; some, like the notochord, may be ancestral patterns retained from protochordates; and the expression in invaginating/migrating cells may have been retained from the jellyfish. Perhaps this is one of the more ancient functions of twist—conserved from diploblasts to humans—to facilitate cell movement. Developmental Dynamics 236:2615–2626, 2007. © 2007 Wiley-Liss, Inc.

INTRODUCTION

The twist genes form a small family present across the metazoa, from the diploblast jellyfish to mammals. The embryonic lethality of twist null mutants of both vertebrates (mouse) and invertebrates (Drosophila) demonstrates that a functioning twist gene is essential for development and survival. In all species, twist genes are necessary for differentiation of mesodermal derivatives, and in vertebrates they are also required for the development of the head neural crest. All Twist proteins bear the group “A” basic Helix Loop Helix (bHLH) domain (Atchley and Fitch,1997), and thus are presumed to be nuclear transcription factors. Provided they are translated, their function is cell-autonomous and is reflected in their expression patterns.

Since all Twist proteins contain the highly conserved bHLH and carboxy domains, their biochemical functions are probably also conserved. However, both changing expression patterns and differences in the phenotypes of the twist mutants show that, at the level of developmental processes, their functions have diverged. The single Drosophila twist gene is first expressed in the invaginating ventral furrow cells and thus plays an early, essential role in gastrulation since null mutants die in the first hours of development, having failed to produce mesoderm (Simpson,1983; Nusslein-Volhard et al.,1984; Thisse et al.,1987). At later stages of Drosophila development, twist participates in further mesoderm differentiation and is very important in muscle formation. Among the vertebrates, the best characterized twist genes so far are from mouse and human. Each species carries two paralogs, twist1 and twist2, both of which become active at more advanced stages of embryogenesis than in the fly, long after gastrulation is complete—in the migrating and differentiating head neural crest, then in the sclerotomes and dermatomes of somites, in the lateral plate mesoderm and the developing limb buds (Wolf et al.,1991; Chen and Behringer,1995; Fuchtbauer,1995; Li et al.,1995; Gitelman,1997; O'Rourke et al.,2002). Neither twist1 nor twist2, however, participate in embryonic muscle formation (Gitelman,1997; Sosic et al.,2003). The two genes are very similar in their coding sequence, and their expression patterns overlap; yet the phenotypes of null mutants are very different. Twist1^−/− mice die at midgestation with open cephalic neural tube and cranial hemorrhages (Chen and Behringer,1995), while twist2^−/− mice die a few days postnatally with severe skin defects and abnormal energy metabolism (Sosic et al.,2003). In the adult, mammalian twist genes are expressed at basal levels in many tissues and overexpression contributes to tumor progression and metastasis in a variety of cancers (Yang et al.,2006). There are a few reports of expression of individual twist homologs in other tetrapods, including chick and frog (Hopwood et al.,1989; Stoetzel et al.,1998; Scaal et al.,2001; Tavares et al.,2001), and one report in a protochordate Amphioxus (Yasui et al.,1998) where the data suggest a role in notochord and somite differentiation. It must be noted, though, that some mRNA expression sites are not functional since twist is subject to posttranscriptional regulation (Gitelman,1997; Demontis et al.,2006).

Despite the strong conservation of the twist genes across the metazoa and their clearly different functions in vertebrates and invertebrates, there is almost no data concerning twist function or expression among the fishes, a basal vertebrate clade. Expression of one twist paralog has been characterized in Medakafish (Yasutake et al.,2004), where it was shown to function in vertebral column formation, and expression of a zebrafish homolog was occasionally used as a marker of axial mesoderm development (Halpern et al.,1995; Yan et al.,1995).

Here we present the embryonic expression patterns of all four genes that comprise the twist family in the zebrafish D. rerio. This is the first such characterization; it spans the gap in data between the amphioxus and tetrapods and permits a more detailed analysis of the evolution of twist gene function.

RESULTS

The Four Genes of the Zebrafish twist Family: Phylogeny and Terminology

Four different twist paralogs were identified in zebrafish as a result of searches for twist-related sequences in the EST and genomic databases, screening of an embryonic expression library, sequencing and comparison with the known twist genes. They most likely represent the entire complement of the zebrafish twist genes (Gitelman,2007). Alignment of the deduced twist protein sequences was straightforward in the highly conserved bHLH domain and the carboxy region (Fig. 1); in the more divergent amino region, it was facilitated by the presence of three conserved islands, two of which were recognizable as a bi-partite nuclear localization signal. A phylogenetic analysis placed all the vertebrate twist genes, including those of zebrafish, into three clades (Gitelman,2007). Each clade contains both teleost and tetrapod genes (Fig. 2), indicating that there were three ancestral vertebrate genes prior to the teleost–tetrapod split.

Figure 1. Alignment of the zebrafish Twist proteins. Alignments were generated as described in the Experimental Procedures section. The highly conserved bHLH domain is marked with a bar. The species symbols are: (Mm) Mus musculus, mouse; (Gg) Gallus gallus, chicken; (Dr) Danio rerio, Zebrafish; (Sp) Strongylocentrotus purpuratus, sea urchin. The alphanumeric suffixes are based on the vertebrate phylogenetic analysis (Gitelman,2007). The full gene names are Gg3 = twist3 of chick, etc.

Figure 2. Phylogenetic relationship between the zebrafish and the tetrapod twist homologs. The aligned Twist proteins sequences (Fig. 1) were used to construct a Neighbor Joining phylogenetic tree. All sequences, except gap-runs, were used to produce a minimum distance tree using the NJ technique. Shown are 1,000 bootstrap reacquisitions of the data. Numbers show the bootstrap values–percentage frequency of the consensus branching patterns (see Experimental Procedures section). The scale bar shows branch length corresponding to the indicated number of sequence changes. Gene and species names are as listed in Figure 1. Grey fields labeled I, II, and III indicate the three vertebrate twist clades.

Genes of the third group do not exist in the mammals but were found in other vertebrate species. The vertebrate genes were named based on orthology to the mammalian twist1 and twist2 genes and according to the rules of the vertebrate gene-naming convention. This involved changing some reported names (Table 1). For example, “zebrafish twist” (Halpern et al.,1995, Yan et al.,1995) was found to be most similar to the mouse twist2/dermo-1 and, therefore, is now called twist2. The sequence “zebrafish twist2” (GenBank, AF205258) was not orthologous to either twist2 or twist1 of mouse, but belongs to clade3, and therefore was re-named twist3. There are two co-orthologs to the mammalian twist1: twist1b (previously twist1, GenBank) and twist1a, a newly found zebrafish gene.

Table 1. Variable Number of Twist Genes in Vertebrates

Species	Gene	Current name (phylogenetically correct)	Previous name(s)	Accession no.
a Wolf et al. (1991). b Li et al. (1995). c Tavares et al. (2001). d Scaal et al. (2001). e Gitelman (2007). f GenBank entry. g Halpem et al. (1995), Yan et al. (1995).
Mouse	Mm 1	twist1	mtwist/twist1a	NM_011658
	Mm 2	twist2	twist2/dermo1b	NM_007855
Chick	Gg 1	twist1	G-twistc	Y08261
	Gg 2	twist2	Dermo-1d	CD765266
	Gg 3	twist3	Nonee	BK006265
Zebrafish	Dr 1a	twist1a	Nonee	EF620930
	Dr 1b	twist1b	twist-1/twist-relatedf	AF205259
	Dr 2	twist2	Zebrafish twistg	EF620931
	Dr 3	twist3	twist2f	AF205258
Sea urchin	Sp	twist	Nonee	BK006287

Differential Expression of the Zebrafish twist Genes

The distribution of twist1a, twist1b, twist2, and twist3 transcripts was determined in zebrafish embryos from zygote to 36 hr post-fertilization (hpf). Wholemount RNA in situ hybridizations were performed with both short gene-specific probes from outside the conserved regions, and with full-length transcript probes; the latter generated specific signals that were identical to the short-probe staining, but were more robust. Although the twist genes all contain highly conserved bHLH and carboxy domains, there was no evidence of cross-hybridization between the probes and sense-strand probes did not produce any staining.

Overall, the four zebrafish twist genes showed similarities but also significant differences regarding the pattern as well as the timing and level of their expression. None was expressed maternally or at the mid-blastula transition. Transcripts were first detectable at the shield stage. All four twist genes were expressed in various mesodermal lineages. All four marked either cephalic neural crest (CNC) or its target structures. Their individual spatio-temporal expression patterns are described below.

Twist1a

Twist1a expression (Fig. 3) was first seen as a very weak signal in the area of the organizer at the shield stage (Fig. 3A,A′), but it was transient and almost immediately disappeared (Fig. 3B,B′). More persistent expression started at the bud stage: first at the level of caudal hindbrain and then at the lateral borders of the developing hindbrain neural keel, the location of the earliest CNC (Fig. 3C,C′). With rounding of the neural keel and drawing of the neural crest domains nearer each other, the signal clearly remained in the dorsal part of the neural rod (Fig. 3D′,D-1), indicating that twist1a marked premigratory crest cells. By the 10-s stage, the premigratory crest signal became stronger and advanced rostrally toward the midbrain and forebrain, where it appeared in a semicircle enfolding the optic vesicles (Fig. 3E,E′). The signal was weaker in the midbrain (Fig. 3E). As segmentation progressed and the yolk started to elongate, twist1a expression in all brain areas subsided significantly (Fig. 3F,G). Sections through the brain areas (Fig. 3D-1 and D-2) showed that this decrease in expression coincided with the start of emigration of the CNC cells (Fig. 3D-2), whose twist1a signal was much weaker than that of the premigratory cells (Fig. 3D-1). No signal was observed in the areas lateral to the neural keel, indicating that the migrating CNC cells do not express twist1a at detectable levels (compare Fig. 3E with Fig. 4E). However, by about 36hpf, twist1a signal became intense again, now in the CNC derivatives (Fig. 3H), specifically in the pharyngeal and branchial arches and in some of the vasculature of the forebrain (Fig. 3H,H′,H″). It should be noted that when CNC cells reach their target organs, they mix with the cells of the resident mesoderm and from expression studies alone it is not possible to ascertain whether it is the CNC descendents, the resident cells, or both that express the original crest markers.

Figure 3. Embryonic mRNA expression pattern of the zebrafish twist1a gene. All images labeled with the same letter represent the same embryo in different views. All images are oriented with rostral at the top; in all lateral views anterior is to the left. Embryos shown are at the following stages: A, ∼60% epiboly; B, ∼75% epiboly; C, 2–3 somites; D, 6–8 somites; E, 10–12 somites; F, 15 somites; G, 24 somites; H,36 hpf. A–E: Dorsal views; A′–E′: lateral views, except B′, which shows the animal pole view. F,G: Head region, dorsal view; H: ventro-lateral view; F′–H′: body, dorsal view; F″–H″: lateral view; G′-1, G″-1: higher magnification of the dorsal and lateral views of the trunk region and the migrating sclerotome; D-I: DIC image of 6–8 somites embryo section through the hindbrain. ba, branchial and pharyngeal arches; cfb, chb, cmb, neural crest at the level of the fore-, hind-, and midbrain; d, dorsal somite; fb, forebrain area; hb, hindbrain area; mb, midbrain area; mnc, migratory neural crest; nc, neural crest; or, organizer region; pfb, pectoral fin bud primordia; scl, sclerotome.

Figure 4. Embryonic mRNA expression pattern of the zebrafish twist1b gene. All images labeled with the same letter represent the same embryo in different views, except A1, which is slightly older than the embryos in A and A′. All images are oriented with rostral at the top; in all lateral views anterior is to the left. Embryos shown are at the following stages: A, late shield; A1, ∼8hpf; B, 1–2 somites; C, 5–6 somites; D, 10 somites; E, 12 somites; F, 15 somites; G, 24 somites; H, 30 hpf; I, 36 hpf. A, A1: Dorsal view; A′: animal pole view; B–I: dorsal veiw of the head region; B′–I′: lateral view of the body; B″–I″: dorsal view of the body; H″: tilted profile view to better show the branchial arches. ap, animal pole area; ba, branchial and pharyngeal arches; cfb, chb, cmb, neural crest of the fore-, hind-, and midbrain areas; fb, forebrain; fbv, forebrain blood vessels; hb, hindbrain; lpm, lateral plate mesoderm; lnp, lateral neural plate; mb, midbrain; mt, myotome; nd, nephric duct; olf, olfactory placode; op, optic vesicle; ot, otic vesicle; pfb, pectoral fin bud primordia; s, somite; tb, tail bud area; arrow, signal in the area of the future ventral gut.

In mesoderm, twist1a was expressed most prominently in paraxial mesoderm starting at 13–14ss. The earliest expression was seen in the sclerotome and dorsal somites (Fig. 3F′,F″,G″). By 15ss, twist1a-expressing sclerotome cells of the most mature somites started to emigrate. Their movement around the notochord and neural tube occurred in a subsiding rostro-caudal gradient (e.g., Fig. 3G″) through the 24 somite stage. By about 36hpf, as post-migratory sclerotome cells have settled at the site of the future vertebrae, only a very weak sclerotome signal could be detected in the tail area (Fig. 3H″). At no stage of the paraxial mesoderm development was twist1a signal detectable in the presegmental plate. In the lateral mesoderm, a weak twist1a signal was detectable by 26hpf in the developing pectoral fin bud field (not shown) and by 36 hpf twist1a transcripts were clearly concentrated in the dorsal fin bud (Fig. 3H′).

Twist1b

Twist1b expression was the broadest of all four zebrafish twist genes. Its transcripts were first detected at the late shield stage, ∼6hpf, as a homogeneous signal in the areas lateral to the embryonic axis (Fig. 4A,A′). As gastrulation proceeded, this signal spread rapidly toward both the animal and the vegetal poles (Fig. 4A1), but was still absent in the axial region at 8hpf. This early pattern was transient, however, as it was no longer detectable by the beginning of segmentation at about 10hpf, when both the anterior neural plate and several mesodermal tissues started to express twist1b (Fig. 4B,B′).

In the anterior neural plate at this stage, a weak but distinct signal emerged (Fig. B) that delineated its borders (Fig. 4B′), the site of the cephalic neural crest. But unlike twist1a, it was not seen in the dorsal-most margins (compare Fig. 4B′C′, with Fig. 3D′). Therefore, twist1b either does not mark the pre-migratory neural crest or marks it weakly. The cephalic signal persisted through early segmentation (Fig. 4D,D′). At about 12 ss (Fig. 4 E,E′), its expression became much stronger and expanded together with migrating crest cells at all brain levels. As crest cells proliferated and moved toward their target tissues, there was an increase in both the number of twist1b-expressing cells and in the per cell signal intensity. The signal remained strong in CNC target tissues at least until 36hpf (Fig. 4I,I′). Specifically, the area of the olfactory placode (receiving the forebrain crest), tissues surmounting the eyes (fore- and midbrain crest), the pharyngeal arches (mid- and hindbrain crest), as well as all areas of the hindbrain crest, including the margins of the otic vesicles, all showed intense twist1b expression (Fig. 4E–H, H″, I, I′). In the forebrain area, the signal at the level of the developing eye, first detected at 1–2ss (Fig. 4B), extended rostrally to include the area of the olfactory placodes and became quite intense by the 12-somite stage (Fig. 4E), as the population of migrating CNC increased. At the beginning of the pharyngula period, it diminished (Fig. 4H), becoming, by 36hpf, localized to blood vessels in the developing brain (e.g., Fig. 4I,I′ and 4I-1). More caudally, the pharyngeal and the branchial arches retained the strong twist1b expression throughout the mid-pharyngula stage.

In the mesoderm, twist1b was expressed in the paraxial, intermediate, and lateral domains. Following the early widespread expression in the gastrulating embryo (Fig. 4A-1), a weak specific signal appeared in the first two newly formed somites at 10.5 hpf (Fig. 4B″). It became robust by 10 ss in the myotomes of all somites (Fig. 4D′,D″). By about 12 ss (Fig. 4E′,E″), there occurred a significant change. With a transition from a cuboidal to a chevron shape, the more mature somites gradually lost twist1b signal forming a caudal-to-rostral gradient of expression (Fig. 4E′–G′), with the strongest myotomal signal found in the 4–5 youngest somites. In the more differentiated somites, twist1b transcripts remained at the apex of the chevron, in the area of the muscle pioneer cells, and in the transverse myosepta (Fig. 4G′). By 36hpf (Fig. 4I′), practically no somitic signal could be detected any longer. A new site of strong, but transient, twist1b expression emerged at about the 24 ss in the area of the tail bud (Fig. 4G′); it persisted until 30hpf (Fig. 4H′), but disappeared entirely by 36 hpf, except for a faint signal in the area of the caudal notochord (Fig. 4I′). No twist1b signal could be detected in the presegmental plate at any of the examined stages.

In the intermediate mesoderm, the forming pronephric ducts expressed twist1b (Fig. 4F″) from about the 15-ss stage and until 30hpf (Fig. 4G′,H′). In addition, at approximately the 15-ss stage, just before yolk tube elongation, a distinct twist1b signal appeared near the area of the future ventral gut (Fig. 4F′, arrow).

In the areas of lateral mesoderm, prominent twist1b expression became apparent at about 10hpf (Fig. 4B″); it persisted with mesoderm differentiation and converged medially, outlining the narrowing mesodermal field (Fig. 4D″). By 24 ss, the developing pectoral fin buds, to which the lateral mesoderm cells contribute substantially, showed a distinct twist1b signal, which increased with further development (Fig. 4H″,I,I′).

Twist2

The twist2 signal was first faintly detectable at the shield stage in a small group of cells in the area of the organizer (not shown; very similar to Fig. 3A), but became both broader and stronger by about 50–60% epiboly (Fig. 5A,A′). As the gastrulation proceeded, it persisted at high levels during the extension of the axial mesoderm rostrally and formation of the notochord (Fig. 5A–F, A′–F′), but was absent elsewhere in the embryo. It reached its peak in the chordamesoderm at about 10.5 hpf (1–2 ss) (Fig. 5D), then diminished quickly, especially in its anterior part (Fig. 5E,F) and, from about 10-ss onward, remained as a thin line of expression (Fig. 5F,H). Post in situ cross-sections of 20-hpf embryos (Fig. 5H-1) as well as sagittal sections of 24-hpf embryos, clearly showed that this narrowed expression domain was due to staining in the hypochord (Fig. 5I-1). At 36hpf, twist2 was expressed in the dorsal aorta (Fig. 5J,J′,J-1), whose genesis is closely associated with the hypochord (Cleaver et al.,2000). The floor plate and the posterior cardinal vein did not express twist2 (Fig. 5J-1). During this time interval, a strong signal remained in the cells of the tail bud (Fig. 5G–I and H-2). It persisted throughout the segmentation period, but started to diminish at about 24hpf (Fig. 5I) and disappeared altogether in the mid-pharyngula period (36hpf) (Fig. 5J).

Figure 5. Embryonic mRNA expression pattern of the zebrafish twist2 gene. All images labeled with the same letter represent the same embryo in different views. All images are oriented with rostral at the top; in all lateral views anterior is to the left. Embryos shown are at the following stages. A: ∼60% epiboly; B: ∼75% epiboly; C: ∼90% epiboly (10hpf); D: 1–2 somites; E: 3–4 somites; F: 10–11 somites; G: 14 somites; H: 24 somites; I: 24 hpf; J: 36hpf. A–F and J: Dorsal views; A′–F′, G–J: lateral views. N.B. Not shown is the start of twist2 expression as a weak signal at the earlier shield stage. ba, branchial and pharyngeal arches; da, dorsal aorta; fb, forebrain; hyp, hypochord; hb, hindbrain; not, notochord; or, organizer region; pcv, posterior cardinal vein; scl,sclerotome; tb, tail bud.

In the paraxial mesoderm, during mid-segmentation, at about 14 ss (Fig. 5G), there was an abrupt initiation of twist2 expression in the sclerotome, similar to the pattern of twist1a. The emigration of twist2-expressing cells (Fig. 5H and 5H-1) followed the rostro-caudal gradient of somite maturation. Slightly later, the twist2-expressing cells also started to emigrate from posterior sclerotomes. By about 30hpf, the entire length of the notochord-neural tube assembly was enveloped by twist2-expressing cells. The signal gradually declined in the post-migratory sclerotome cells (Fig. 5I). At none of the stages of the paraxial mesoderm development, was the twist2 signal detected in the adaxial cells or in the presegmental plate. It was also absent from the lateral plate mesoderm.

In the areas of the developing brain, twist2 expression was first detected at the level of hindbrain at about 14 ss. At first extremely faint, the signal became clearer by 24hpf (Fig. 5I). By the mid-pharyngula stage (Fig. 5J), while still diffuse and weak, twist2 signal could also be seen in the midbrain and forebrain areas, where twist2 was expressed in the developing blood vessels of the pharyngeal and branchial arches, and in the forebrain vasculature (Fig. 5J,J′).

Twist3

Twist3 displayed the simplest pattern and the latest onset of expression of all the zebrafish twist genes. The signal first appeared at 13 ss (Fig. 6A) in the dorsal forebrain (Fig. 6A,A″) and at the lateral borders of the cephalic neural plate (Fig. 6A). At this stage, twist3 was also detected in the lateral plate mesoderm (Fig. 6A), most strongly in the pectoral fin field (Fig. 6A′,A″). The signal in the fin buds became very strong by about 24hpf (Fig. 6C′,C″) and subsequently localized mostly to mesenchymal cells of the dorsal bud in the area of the progress zone (Fig. 6D′,D″,D-1). At about 24hpf, there also appeared a distinct twist3 signal at the tip of the tail (Fig. 6C). It further expanded to mark the area of the tail fin primordium. At this stage, twist3 was also weakly expressed in more mature somites at the peak of the chevron, where muscle pioneer cells are found, and in the transverse myosepta (Fig. 6C).

Figure 6. Embryonic mRNA expression pattern of the zebrafish twist3 gene. All images labeled with the same letter represent the same embryo in different views. All images are oriented with rostral at the top; in all lateral views anterior is to the left. Embryos shown are at the following stages. A: 13 somites; B: 20 somites; C: 24 hpf; D: 36 hpf. A–D: Lateral view; A′–D′: dorsal view of the body; A″–D″: dorsal view of the head region. D″: Tilted profile for better view of the structures. All lateral images are oriented with rostral to the left. ba, branchial and pharyngeal arches; cfb, chb, cmb, neural crest of the fore-, hind-, and midbrain areas; fbv, forebrain blood vessels; hb, hindbrain; lnp, lateral neural plate; lpm, lateral plate mesoderm; mb, midbrain; pfb, pectoral fin bud primordia; tfp, area of the tail fin primordium.

DISCUSSION

Developmental expression histories of conserved genes can illuminate the evolution of gene functions. Twist genes are particularly informative for such comparative studies: they form a small, distinct, highly conserved family of essential transcription factors whose functions in directing developmental programs have diverged during evolution. Phylogenetic analysis pointed to two early rounds of twist duplication prior to the teleost–tetrapod split that led to three ancestral vertebrate twist genes (Gitelman,2007; Fig. 2). Further duplication(s) in the teleost lineage led to two zebrafish co-orthologs of mammalian twist1: twist1a and twist1b. Twist2 is the single ortholog of the mammalian twist2, and twist3 is a representative of a new class of twist genes, present in fish, frogs, and some amniotes (chicken), but absent in mammals. The expression of twist3 reported here is the first study of a member of this ancient but fast-evolving class of twist genes. It shows the most restricted patterns in both space and time, with the latest onset of expression among the paralogs. With further studies of twist functions, it may become possible to infer why the twist3 paralog, which has been retained in other tetrapods, e.g., chick and frog, was lost from mammals.

The fact that both zebrafish and tetrapod twist genes are expressed in head crest, suggests that this is an ancestral vertebrate twist function, which preceded the teleost-tetrapod split. Having the expression patterns of all four twist zebrafish genes permits insight into how this family evolved in these two groups of vertebrates. Some expression differences in the CNC probably reflect changes in their specific developmental roles in head development, which is not surprising given the significant variations in morphology of the head and neck areas in these species. We think that on the molecular level, some of these differences are due to divergence of the cis regulatory elements. Specifically, the more restricted expression patterns of the zebrafish orthologs are due to subfunctionalization (Force et al.,1999) that has partitioned more complex regulatory regions into reduced, paralog-specific expression domains.

It also means that the superficially simple expression of the murine twist genes in the head crest belies a complex regulation. In the mouse, twist1 is expressed relatively uniformly in both migrating and post-migratory head crest in a temporal gradient corresponding to the gradient of cell emigration (Wolf et al.,1991; Chen and Behringer,1995; Fuchtbauer,1995; Gitelman,1997; Ishii et al.,2003). twist2/dermo-1 expression marks only the postmigratory head crest cells in their target tissues, where it overlaps with twist1 (Li et al.,1995). In zebrafish, all four genes are expressed in various combinations in the different spatial and temporal subsets of the head crest. Here are a few examples. Twist1a's strongest expression is in the pre-migratory CNC twist (Fig. 3D′, D-1) where the twist1b signal is low (Fig. 4B–D). On the other hand, in the migrating CNC cells, twist1a is barely detectable (Fig. 3E,F), while twist1b is expressed highly (Fig. 4E,F).

The only other fish twist gene whose expression has been characterized so far is (Yasutake et al.,2004) the medaka ortholog of zebrafish twist1a (Gitelman,2007). The distribution of twist1a transcripts is very similar in both fish, with one notable exception. Medaka twist1a is clearly expressed in migrating CNC, whereas zebrafish twist1a is not. This is interesting with respect to the evolution of the fish and of twist function. There are five fish species in which there is sufficient data to generate a complete twist family phylogeny: zebrafish, medaka, stickleback, fugu, and green spotted pufferfish (Gitelman,2007). All of these, except zebrafish, belong to the Percomorphs (Miya et al.,2003), the crown fish that form the “bush at the top of the [teleost] tree” (Nelson,1989). Combining the known expression and phylogenetic data suggests that ancestral twist1 CNC functions have been partitioned between the two co-orthologs in zebrafish, but retained by twist1a in the Percomorphs (Fig. 7). While twist1a orthologs of all these fish and the twist1b ortholog of zebrafish are closely related to tetrapod twist1, the Percomorph twist1b genes form a divergent and fast-evolving clade (Gitelman,2007; Fig. 7). Since the sequences of developmental genes that have lost ancestral functions and acquired new ones diverge more rapidly (Williams and Holland,1998; Sedlacek et al.,1999), and since all vertebrate twists genes examined are expressed in CNC derivatives, we hypothesize that Percomorph twist1b is involved in building the interarcual cartilage (Travers,1981), a CNC-derived structure that is a defining feature of the Percomorphs.

Figure 7. Partitioning of expression among vertebrate Twist1 orthologs. A phylogeny of the vertebrate twist1 genes (after Gitelman,2007) shows that following the teleost–tetrapod split, duplication of the ancestral twist1 in the teleost lineage produced two co-orthologs twist1a and twist1b. Branch lengths are proportional to sequence divergence as indicated by the scale bar. Hairpin-shaped curves represent the domain of the pre-migratory cranial neural crest (CNC) in the developing brain, while the circles represent migrating CNC. twist expression is indicated by filled curves and circles, while lack or low expression is represented by open curves and circles (after Wolf et al.,1991, Fuchtbauer,1995, Gitelman,1997, for mouse twist1; Hopwood et al.,1989, for Xenopus twist1; and Yasutake,2004, for Medaka twist1a). Parsimony suggests that the ancestor to both teleosts and tetrapods expressed twist1 in both premigratory and migrating neural crest. Species are as follows: Mm, Mus musculus, mouse; Xl, Xenopus laevis,- South African clawed frog; Ol, Oriyzias latipes, Medakafish; Ga, Gasterosteus aculeatus, stickleback; Tn, Tetraodon nigroviridis, pufferfish; Dr, Danio rerio, zebrafish. Genes of the Percomorphs are shaded.

Mesoderm differentiation also showed concurrent conservation and divergence of twist expression domains compared to the tetrapods. For example, the mammalian twist1 gene is expressed in the sclerotome, whereas of its two zebrafish co-orthologs, twist1a has retained this expression domain, but twist1b has lost it. Likewise, in the mouse, twist is active both early, in the somatic lateral plate mesoderm (LPM), and later in the limb buds, which arise from LPM. In the zebrafish, all twist genes except twist2, show a signal in the pectoral fin bud, the homolog of the forelimb bud of tetrapods (Figs. 3H′, 4I, 6D′), but only twist1b and twist3 are expressed in the pre-bud LPM (Figs. 4B″, 6A). Such differential conservation is indicative of subfunctionalization events. On the other hand, the single twist gene of the protochordate Amphioxus and all vertebrate twist2 genes, in both teleosts and tetrapods, are expressed in the notochord or its derivatives. More generally, therefore, this axial twist expression probably represents an ancestral function within the chordates.

There are also several mesodermal expression sites that were identified in zebrafish, but not in tetrapods; for example, twist1b in the nephric duct and young myotomes (Fig. 4F′,G′,F″), twist2 in the dorsal aorta (Fig. 5J,J′,J-1), and twist3 in the area of the developing tail fin (Fig. 6C,D,D′). These may, therefore, reflect neofunctionalization events. However, it is also possible that they are ancestral functions retained in some species and lost in others. In the case of the nephric duct, while no equivalent expression was described in the mouse embryo, we have shown twist1 expressed in several nephropathologies in adults (Kida et al.,2007; unpublished data). The myotome expression of the zebrafish twist1b is curious because tetrapods do not express twist in embryonic muscle precursors, but outside the vertebrates twist is expressed in the somitic muscle progenitors in the Amphioxus, in the muscle precursor cells of Drosophila, and even in muscle-forming cells of the diploblast jellyfish (Currie and Bate,1991; Yasui et al.,1998; Spring et al.,2000). We, therefore, suggest that this expression is an ancient metazoan feature that was retained in the fish embryos but lost in the tetrapod embryos.

Nevertheless, it would be premature to conclude that the vertebrate twist family is involved in embryonic myogenesis. Our preliminary results with anti-twist morpholinos suggest that the post-transcriptional regulation of twist genes first seen in mouse embryos (Gitelman,1997), also takes place in the zebrafish, and that there is no twist function in the fish myotomes. Therefore, in some locations, twist mRNA expression is probably non-functional but represents atavistic retention or ancient patterns.

The process of gastrulation is highly conserved throughout the animal kingdom, such that even in diploblasts, e.g., the cnidarian Podocoryne carnea, there are early cells during medusa development that invaginate between ectoderm and endoderm and form a third cell layer (Spring et al.,2000). Jellyfish twist is expressed in these cells, Drosophila twist is expressed in the invaginating cells of the ventral furrow, zebrafish twist2 is detected in some of the first axial cells of the intermediate layer in the organizer region (Fig. 5A and earlier). And all vertebrate twist genes are involved in the so-called “second gastrulation,” the highly conserved vertebrate process of neural crest migration. Perhaps the most ancient function of twist, and the most conserved one, from diploblasts to human, is to facilitate cell movement.

EXPERIMENTAL PROCEDURES

Identification of Twist Sequences

Zebrafish twist1a cDNA was originally obtained by screening a 48-hr embryonic expression library (Gitelman,2007) with zebrafish twist2 cDNA, which was kindly provided by Drs. Yan and Postlethwait (Halpern et al.,1995; Yan et al.,1995). For RNA in situ hybridizations, full-length twist1b, twist1a, and twist3 cDNAs were used that were purchased from Open Biosystems, Inc., Huntsville, AL.

twist genes from other species were identified by BLAST searches (Altschul et al.,1990,1997) of putative and actual translation products from accessible databases, with the conserved bHLH-containing carboxy region. Sequence data, whether identified as cDNAs, genomic DNA contigs, trace sequences, ESTs, or our own sequence data, were collected, compared, and assembled using AssemblyLIGN (Accelrys, San Diego, CA).

Alignment and Phylogenetic Analysis

Amino acid sequences of different twist genes were deduced from their coding regions in MacClade, v4.08 (Maddison and Maddison,2002) and aligned using ClustalX (Jeanmougin et al.,1998) Minor adjustments were, therefore, made in MacClade manually, typically by sliding one or a few amino acids from one side of a gap to another. The phylogenetic tree was constructed using PAUP, v4.0b10 (Swofford,2003).

Whole-Mount RNA In Situ Analysis, Sections, and Imaging

Zebrafish embryos were collected, raised at 28.5°C, and staged according to Kimmel et al. (1995), dechorionated with forceps, and fixed overnight in 4% paraformaldehyde in phosphate buffered saline at 4°C. Fixed embryos were transferred into 100% methanol, rinsed 4 times in methanol and stored at −20°C. Probes for RNA in situ analysis were generated from plasmids each carrying one of the full-length zebrafish twist cDNAs, twist1a, twist1b, twist2, twist3, or short, unique regions from these cDNAs. The plasmids were linearized and dig-labeled transcripts were produced using the Roche (Basel, Switzerland) products and protocols. The transcripts were then subjected to alkaline hydrolysis to obtain probes of about 100 nt in length. A standard zebrafish RNA in situ procedure was followed (e.g., Thisse et al.,1993; and C. B. Moens) with the following modifications. ProteinaseK treatments were 10 μg/ml in phosphate buffered saline + 0.1% Tween-20 for 0, 1, 2, 3, 4, and 6 min for embryos at stages (hpf): 0–5, 5–12, 12–16, 16–19, 19–24, >24, respectively. The post-proteinaseK glycine step was omitted; the embryos were transferred directly into 4% paraformaldehyde and re-fixed for 20 min. The pre/hybridization salt concentration was reduced to 2× SSC and hybridization temperature was set to 65°C. Antibody blocking and immunostaining were done in maleic acid buffer, 2% Boehringer Blocking Reagent (MA-BBR, Roche) made according to the manufacturer's protocol with the addition of 20% heat-inactivated fetal calf serum. Post-immunostaining washes were done in the MA-BBR buffer with the addition of 5% fetal calf serum. NBT/BCIP staining was done according to the manufacturer's protocol (Roche). Color development was left to proceed at room temperature, in the dark, for about 16 hr. For storage, the embryos were rinsed and kept in 4% paraformaldehyde at 4°C. For sectioning, whole mount processed embryos were embedded in OCT (Tissuetek), oriented, and frozen in the cryostat microtome. Sections (18 mm) were collected on Aminosilane-coated slides, which were then air dried at RT, o.n., mounted in 90% glycerol, and photographed under DIC of Leica DMR microscope with Spot camera. Whole embryos were photographed using a Leica MZFl III stereoscope and Nikon D1X digital camera. Images were processed using PhotoshopCS (Adobe).

Identification of the Expression Sites In the Zebrafish Embryos

To identify the multiple varied expression sites for all four twist genes and at different developmental stages and to verify them, a variety of resources were used. Many hundreds of entries in The Gene Expression Data collection of the ZFIN and the Zebrafish Database at the NIH together with a couple of hundred published articles where zebrafish gene expression is described have served as invaluable resources for understanding the expression patterns.

Acknowledgements

We are grateful to Drs. Yi-Lin Yan and John Postlethwait for the zebrafish twist cDNA (now twist2), and Dr. Ajay Chitnis for the original twist1 (now twist1b). This work was supported by an Israel/USA Binational Science Foundation Grant to Inna Gitelman.