HyBra1, a Brachyury homologue, acts during head formation in Hydra

There is a great deal of evidence that many of the genes regulating developmental processes are common among vertebrates and arthropods. Evidence is also beginning to accumulate that these same genes arose early in metazoan evolution as they occur in the Cnidaria, the first animals with a distinct body plan. For example, members of several families of transcription factors have been characterized. These include Hox genes (Schummer et al., 1992; Shenk et al., 1993a), other homeobox genes (Schummer et al., 1992; Grens et al., 1996; Gauchat et al., 1998; Mokady et al., 1998), bHLH genes (Grens et al., 1995) and winged helix genes (Martinez et al., 1997). In addition, homologues of genes in the Wnt cascade (Hobmayer et al., 1996) as well as a number of RTK genes (Steele et al., 1996; Stover and Steele, 1998) have been isolated and characterized. Some of these genes have functions similar to their homologues in more complex animals indicating an evolutionary conservation of function throughout the metazoa. For example, the Hox genes isolated from hydra appear to have a role in anterior-posterior patterning, while the emx homologue in hydractinia (Mokady et al., 1998) and the HNF-3β homologue in hydra (Martinez et al., 1997) have roles in the patterning of the head. In addition, Cnash, a hydra homologue of the achaete-scute class of bHLH genes, which has a role in specifying the cell fate of a class of cells as do these proneural genes in vertebrates and Drosophila (Grens et al., 1995).

However, there are also a number of these regulatory genes that are highly conserved on the sequence level among metazoans, but in which the function has not been obviously conserved. In some cases, the gene may have been co-opted for different functions throughout evolution. In others, the complexity of the organism and the multiple roles of the gene might obscure functions that are common. Because the cnidaria have a much simpler body plan with fewer components, a given regulatory gene most likely has fewer, and plausibly only a single, function. Hence, an examination of such genes in a cnidarian could provide clues as to the common aspects of their function among the animals where they occur.

One class of genes that play important roles in a number of developmental events in several species are the T-box genes. These genes are characterized by a highly conserved DNA-binding domain, the T-box (Kispert and Herrmann, 1993; Kispert et al., 1995; for reviews see Kispert, 1995; Smith, 1997; Papaioannou and Silver, 1998). The founding member, Brachyury, has been shown to play a critical role in mesoderm formation in vertebrates (for review see Smith, 1997). Homologues of this gene, and other T-box genes, have been cloned in organisms ranging from C. elegans to humans indicating the ancient origin of this gene family (Bollag et al., 1994; Agulnik et al., 1995). Although the T-box genes are highly conserved at the sequence level, there are clear differences in the function of these genes in terms of the tissue, the time during embryogenesis, and/or the developmental process that they are expressed in among the deuterostomes and protosotomes.

We have isolated and characterized HyBra1, a hydra homologue of Brachyury. Examination of its expression pattern in a variety of experimental circumstances indicates it plays a role in head formation in hydra, suggesting at first glance a function that is rather different from those observed in other developing embryos. However, a consideration of other genes with which Brachyury homologues are co-expressed suggests the evolutionary conservation of a molecular pathway that acts to invaginate or evaginate tissue during axis formation and formation of the germ layers.

Unless otherwise stated, the experiments were carried out with animals of the Basel strain of Hydra vulgaris, strain Zürich was used for LiCl treatment. Reg-16 is a regeneration-deficient strain of H. magnipapillata. For DAG treatments, the 105 strain (H. magnipapillata) was used (Achermann and Sugiyama, 1985). Embryos were obtained by mating vulgaris strain PA2. Both strains are derived from male and female strains described previously (Martin et al., 1997). Stock cultures of hydra were maintained as described previously (Martinez et al., 1997). Cultures were fed daily and starved 1 day before selection of animals for experiments.

First-strand cDNA from whole animals was used for nested PCR with three fully degenerate primers encoding conserved amino acid sequences of the T-box. Bra-1 (5′AYGGNMGNMGNATGTTYCC3′) and Bra-3 (5′RAANSCYTTNGCRAANGG3′) were used as forward and reverse outer primers, while Bra-6 (5′TTYGGNGMNCAYTGGATG3′) together with Bra-3 used for nested PCR. PCR conditions were: 3 minutes at 94°C (1 cycle); 1 minute at 94°C, 1 minute at 42°C, 1 minute at 72°C (35 cycles); 10 minutes at 72°C. Fragments of the expected size were obtained and sequenced using standard procedures (Sambrook et al., 1989). Screening of a random-primed cDNA library (as described by Sambrook et al., 1989) derived from adult hydra (Sarras et al., 1994) with the random-primed PCR fragment yielded a single 926 bp clone, which was subsequently cycle sequenced (Amersham). The sequence of the clone, termed HyBra1 (GenBank accession number AF105065), consists of 129 bp of 5′ UTR, a putative start codon at position 130, the full T-box as shown in Fig. 1 (bp 251-797) and 129 bp of the activation domain.

The size of the transcript of the gene was determined by northern analysis using ~2 μg of poly(A)⁺ RNA isolated from 4000 hydra heads (Qiagen) probed with a random-primed 800 bp fragment of the HyBra1 gene. The hybridization conditions were 7% SDS, 0.5 M sodium phosphate buffer (pH 7.2), 5% sodium dextran at 60°C for 18 hours, The blot was washed with in 2× SSC/0.2% SDS for 15 minutes at room temperature, followed by 0.2× SSC/0.2% SDS at 60°C for 30 minutes. The presence of HyBra1 in embryos was determined using RT-PCR. Total RNA was isolated from 30-40 embryos at several embryonic stages (unfertilized egg, blastula, gastrula/postgastrula, cuticle formation stage, cuticle stage 2-10 days after fertilization), reverse transcribed with AMV or M-MLV as described by the manufacturer (Gibco), and 1 μl of the first-strand cDNA (corresponding to half an embryo) used as a template in 20 μl PCR. HyBra1-specific primers were used alone, or together with primers for the hydra EF-1α gene (GenBank accession number Z68181) as an internal control (forward primer: 5′GTTTGAAGCTGGTATTTC3′; reverse primer: 5′TGTTTACCAGTATCTTTG3′). PCR conditions were: 3 minutes 94°C (1 cycle), 1 minute 94°C, 1 minute 46°C, 1 minute 72°C (35 cycles), followed by 10 minutes 72°C.

T-boxes from GenBank database were aligned using ClustalW and a phylogenetic analysis was performed using the program Puzzle (Strimmer and von Haeseler, 1996), a maximum likelihood-based method applying quartet puzzling. Blosum 62 was used as a substitution model and, because of blocks of high conservation within the sequence, we assumed a rate heterogeneity along the sequence. The required parameter alpha of the gamma distribution was estimated by the program from the data set and found to be 0.6, justifying our initial assumption. 1000 replicate analyses were performed. Branch lengths correspond to the number of changes. Numbers at each node correspond to the percentage of replicates in which the node was supported, a measure of statistical support (Strimmer and von Haeseler, 1996). An analysis using the maximum parsimony program PAUP (Swofford, 1996) yielded basically the same tree topology (analysis not shown).

In situ hybridization analysis on whole mounts of animals or late-stage embryos was carried out as described previously (Grens et al., 1995, 1996) with some minor modifications. Inserts of HyBra1 plasmids were amplified by PCR using T3 and T7 primers and 500 ng of the PCR product was directly labeled with digoxygenin (DIG; Tautz and Pfeifle, 1989). Hybridization was carried out over 36 hours with approximately 0.025 ng/μl DIG-labeled RNA probe. Samples were incubated in a 1:2000 preabsorbed dilution of anti-DIG-antibody (Boehringer) in sheep serum overnight, washed in MAB (100 mM maleic acid, pH 7.5) and visualized with the alkaline phosphatase substrate BM Purple (Boehringer). The color reaction was stopped with 100% ethanol and animals were mounted in Euparal (Asco Laboratories).

Treatment of hydra (105 strain, H. magnipappillata) with DAG (0.1 mM dioctanoyl-sn-glycerol) was carried out as described by Müller (1989). The working solution was prepared in 3 ml hydra medium, sonicated for 30 seconds and cooled on ice for 5 minutes. On the first day, animals were treated for 30 minutes in a 6 cm Petri dish and, on subsequent days, they were exposed to DAG for 2 hours. After each treatment, they were washed 5 times in hydra medium. Animals were fed 3× per week during the treatment.

Animals were bisected either directly beneath the tentacle ring or in the middle of the body column, and allowed to regenerate. Periodically thereafter, samples were fixed in fresh 4% paraformaldehyde in hydra medium and subsequently analyzed for expression of HyBra1 with in situ hybridization. In parallel samples, the regenerating tip was isolated and grafted laterally into the 2-region (as defined by Bode and Bode, 1984) of a host animal. Host and donor tissue were strung onto a 0.25 mm diameter fish line and held in contact with one another with flanking pieces of parafilm for about 2 hours in order to heal. Secondary head formation was analysed 2 days after grafting. A secondary head was defined by the presence of a hypostomal dome surrounded by at least two tentacles, while partial heads were defined by a single distal tentacle.

Using nested PCR and first-strand cDNA derived from whole hydra, fragments of two genes, each with a high level of similarity to the T-box of Brachyury were isolated. One of them, termed HyBra1 was used to screen a hydra cDNA library (Sarras et al., 1994) and yielded a single clone 926 bp in length that included some of the 5′UTR, the entire T-box and approximately 100 bp on the 3′ end of the T-box. Northern analysis using stringent conditions indicated the presence of a single transcript of approximately 2.3 kb (data not shown). Sequence comparison with other Brachyury homologues indicates that HyBra1 shows a a very high degree of conservation in the conserved T-box (about 75-80% amino acid identity; Fig. 1). In general, most domains that are conserved among these homologues are also conserved in HyBra1. This includes those residues that have been shown to be implicated in dimerization and direct contact with the palindromic binding site by crystallographic studies on Mouse Brachyury (Müller and Herrmann, 1997).

Phylogenetic analyses indicate that T-box genes fall into one well-defined subfamily, the Brachyury (T) subfamily, and 2-6 other subfamilies that are quite distinct from the Brachyury subfamily. A maximum likelihood analysis shows that HyBra1 clearly clusters with the T/Brachyury-like genes (Fig. 2) as indicated by the level of statistical support (95%). The branching order within the T-subfamily, however, cannot be resolved with certainty. Hence, we conclude that HyBra1 is a true orthologue of Brachyury.

Because of the tissue dynamics of an adult hydra, the developmental processes that were active in setting up the body plan remain continuously active to maintain the form and regional distribution of cell types in the adult (for review see Bode and Bode, 1984). In situ hybridization analysis in adult hydra revealed that HyBra1 is strongly expressed throughout the hypostome, the apical half of the head (Fig. 3A). Expression drops sharply in the tentacle zone, the lower half of the head (Fig. 3B), and can not be detected below the tentacles.

Of the two tissue layers, the ectoderm and endoderm, HyBra1 is always strongly expressed in the endoderm of the hypostome, and occasionally some faint staining in the ectoderm of the hypostome was observed. The endoderm of the hypostome is folded into wedges that protrude into the gastric cavity. Correspondingly, the gastric cavity is reduced to a canal with branches that separate neighboring wedges or protrusions (Fig. 3D). In the hypostome, the endoderm contains two cell types, endodermal epithelial cells and mucous cells. The endodermal epithelial cells extend from the base of the tissue layer next to the basement membrane to the edge of the tissue layer facing the gastric cavity. The mucous cells are located among the epithelial cells on the outer edge of each protrusion facing the gastric cavity (see Fig. 3D). The unstained areas are the location of the mucous cells, thus, expression of HyBra1 is confined to the epithelial cells of the endoderm (Fig. 3C).

Expression of HyBra1 in the adult head suggests it would also have a role in the development of the head. Since head development occurs during budding, hydra’s form of asexual reproduction, during head regeneration, as well as during embyrogenesis, this possibility can be examined in three different circumstances.

During asexual reproduction, the first sign of a developing bud is the formation of a circular ectodermal placode in the lower third of the body column. The placode evaginates and then elongates into a cylindrical protrusion, which subsequently forms a head at the distal end and a foot at the proximal end. HyBra1 expression is first observed as a small distinct spot comprising 10-15 endodermal epithelial cells in the region where a bud will form, but before any morphological sign of bud formation is visible (Fig. 4A). By bud stage 1 (placode stage; bud stages as described by Otto and Campbell (1977)), when the ectoderm thickens locally, the underlying endoderm expresses HyBra1 in a slightly larger, perfectly round spot (Fig. 4B). As the placode protrudes and elongates into a cylinder, HyBra1 expression of the gene expands to encompass the area of the future hypostome, but does not extend further in a proximal direction (Fig. 4C-F).

Thus, HyBra1 expression remains strictly confined to the distal tip forming a cap of expression throughout all stages of bud evagination. As the gene is expressed very early during budding, the hypostome appears to be specified at a very early stage. We also note that expression during budding always appears to be much stronger than in the hypostome of the adult suggesting that the requirement for the gene during head formation is higher than for the maintenance of the head in the adult.

When hydra are bisected, the apical end of the lower half regenerates a head in 2-3 days. We monitored HyBra1 expression during head regeneration in animals bisected at two axial locations. When bisected directly beneath the tentacles, HyBra1 expression was first detected 3 hours later (Figs 5A,B, 6A), and rapidly increased in intensity during the next 1-2 hours (Fig. 5C) reaching a maximum by 6 hours (Figs 5D, 6A) since no appreciable increase was detected at 12 hours (data not shown). By 36-48 hours, tentacle buds emerged precisely at the basal margin of the expression domain (Fig. 5E) indicating that HyBra1 expression is confined to the developing hypostome. Thereafter, expression tended to decrease reaching the somewhat lower levels typical of an intact animal by 72 hours.

When animals were bisected in the middle of the body column, HyBra1 expression was first detectable at the apical tip by 6-9 hours, reaching a maximum expression around 12-18 hours (see Fig. 6A). Since the rate of head formation is graded down the body column, the slower appearance of HyBra1 is consistent with the slower rate of head formation when animals are bisected at this level (Wilby and Webster, 1970; MacWilliams, 1983b; Technau and Holstein, 1995).

Reg-16, a mutant strain of Hydra magnipapillata, which is defective in head regeneration (Achermann and Sugiyama, 1985), provides another approach for examining the role of this gene in head formation. When reg-16 is bisected, most animals do not regenerate a head, while some form a partial head and a few regenerate a complete head. Expression of HyBra1 was followed for 6 days in animals bisected beneath the tentacles. In contrast to the wild type, where strong HyBra1 expression is observed by 6 hours, none of the reg-16 animals expressed HyBra1 before 48 hours. By 6 days, most of the animals had not formed head structures and no HyBra1 expression was detected (Fig. 7A). A few regenerates (about 5%) formed a single medial tentacle at the apical tip. In these animals, HyBra1 expression occurred either at a very low level (Fig. 7B) or was not detected. A small fraction (10-20%) of the regenerates developed a complete head (i.e. a hypostomal dome plus at least two tentacles) and, in every one of these, HyBra1 was expressed (Fig. 7C).

These results obtained with reg-16 coupled with those obtained for normal regenerates and budding indicate that the expression of HyBra1 is tightly correlated with development of the head and specifically with hypostome development.

The early expression of HyBra1 during budding and head regeneration suggests that it might be involved in the process of head determination. One approach is to compare the time of the appearance of HyBra1 expression in a regenerating tip with the time that tip was determined to form a head. A standard grafting procedure (e.g. MacWilliams, 1983a,b) involving the formation of a secondary head was used to measure when this determination had taken place. As shown in Fig. 8A, animals were bisected beneath the tentacles, allowed to regenerate and periodically the regenerating tip was transplanted to the upper body column of another hydra (see Materials and Methods for details). 2 days after grafting, the fraction of transplants forming a secondary head was measured. As shown in Fig. 8B, the fraction forming a secondary head rose from a value of 15% at T=0 to 75% by T=5-6 hours with a T50 of 4-5 hours. Since all of the regenerates expressed HyBra1 by 4-5 hours (Fig. 6A), and only half were committed to head formation by that time, expression of the gene is consistent with a role in the head determination process.

However, though HyBra1 is expressed before head determination, this expression does not imply the tissue is irreversibly determined to form a head. Four results support this view. (1) Only 40-50% of the 4-5 hours regenerating tips formed heads upon transplantation after 2 days even though all regenerating tips expressed the gene. (2) In a second experiment (see Fig. 9A), apical tips regenerating for 4-5 hours were transplanted to a host, and periodically samples examined for HyBra1 expression. As shown in Fig. 9B, the fraction expressing the gene drops from 100% to 50-60% in 2 days which is similar to the fraction that will form heads. Hence, those transplants that do not form heads lose HyBra1 expression. (3) In a variation of the same experiment, we transplanted apical tips immediately after decapitation to a host. These grafts did not express HyBra1 at the time of transplantation (Fig. 9B). Surprisingly, 6 hours after grafting almost all (96%) of the grafts strongly expressed the gene at their distal ends in a manner characteristic of regenerating tips. During the next 2 days, however, most of the transplants (60%) ceased expressing the gene, which correlates reasonably well with the fraction (15%) of these transplants that were expected to eventually form heads (see Fig. 8B). (4) We also examined HyBra1 expression during foot regeneration at the basal end of the upper half of animals bisected in the middle of the body column. Surprisingly, we found that HyBra1 is transiently expressed reaching a maximum by 9 hours following bisection and then slowly disappearing thereafter (Fig. 6B). As with head regeneration, expression was confined to the regenerating tip.

The head activation level of the tissue, i.e. its potency to form a head, is graded along the body axis of the adult animal and has a maximum in the head (MacWilliams, 1983b). The expression of HyBra1 in the adult head as well as during head regeneration could be a response to a high level of this gradient. The overall level of the gradient along the body column can be raised by treatment with diacylglycerol (DAG) (Müller, 1989), or lowered with LiCl (Hassel and Berking, 1990). Hydra were treated with both compounds to determine if the range of HyBra1 expression was altered.

Treatment of animals of the 105 strain of Hydra magnipapillata with DAG for up to 14 days resulted in the formation of ectopic tentacles and heads, which has been shown to coincide with a rise in the head activation gradient (Müller, 1990). Fig. 10A shows a DAG-treated animal that formed an ectopic head and several ectopic tentacles. Three observations can be made. (1) Secondary heads characterized by a hypostomal dome and at least 2 tentacles surrounding it, always strongly express HyBra1 in the developing hypostome. (Fig. 10B). (2) Formation of single tentacles was not accompanied by the expression of HyBra1, suggesting that its expression is tightly correlated with hypostome, but not tentacle, formation (Fig. 10B). (3) The range of expression of HyBra1 in the adult head of a DAG-treated animal remained unaltered, that is, it did not extend basally into the body column (see Fig. 10A), as would be expected if the expression of this gene were readily influenced by an altered head activation gradient in the body column.

Treatment with 1 mM LiCl for 7 days leads to the formation of ectopic feet along the body column (Hassel and Berking, 1990) as well as a lowering of the head activation gradient (Maggiore and H. R. B., unpublished results). Animals of the Zürich strain of Hydra vulgaris were treated for 7 days and analyzed for HyBra1 expression. No changes were observed (data not shown). Hence, the absence of changes in the expression of HyBra1 along the body column of the adult with either treatment indicates that expression of this gene is not governed by the head activation gradient.

Finally, we examined the expression pattern of HyBra1 during embryogenesis. Embryogenesis in hydra is quite simple (e.g. Martin et al., 1997). In brief, the fertilized egg undergoes 6-7 cleavage divisions resulting in a blastula consisting of a single layer of cells. Then, gastrulation occurs by ingression of single cells from all parts of the blastula into the blastocoel thereby forming a solid stereo gastrula. Subsequently, the cells of the outer layer, the presumptive ectoderm, secrete a thick cuticle and the embryo enters a relatively quiescent stage. 2-3 weeks after fertilization, the mass of epithelial cells in the center of the embryo organize into an epithelial layer, the endoderm. Within a day thereafter, the cuticle cracks at the apical end, and the hatchling emerges and elongates from a spherical form into the cylindrical shape of the adult animal forming a head and foot during and shortly after elongation.

The expression pattern of HyBra1 during this process was examined in two ways. In one, embryos at several stages were collected, RNA isolated and RT-PCR carried out. As shown in Fig. 11, HyBra1 was detectable in adult animals, but not during early or middle stages of embryogenesis. Other experiments indicated that it is also not expressed during the late cuticle stages (data not shown).

To gain more precise information on the timing and location of expression of the gene, in situ hybridization analysis was carried out on whole mounts of embryos. Consistent with the RT-PCR results HyBra1 expression was not detected during the early stages of embryogenesis (early cleavage stages, blastula, gastrula) before cuticle formation (data not shown), whereas control in situ hybridizations with EF-1α showed ubiquitous expression in these stages (data not shown). During the final stages, the endoderm forms, the cuticle becomes porous and, subsequently, the cuticle cracks open at the anterior end of the embryo (Martin et al., 1997). Just prior to hatching, a small patch of HyBra1 expression was found in the endoderm at the presumptive apical pole (Fig. 12A). This onset of expression occurred most likely shortly after the formation of the endodermal layer. As the embryo expanded and elongated during hatching, the expression of the gene increased in intensity as well as in area covering the apical end (Fig. 12B,C). 5-10 hours later the tentacles emerge indicating that HyBra1 expression was confined to the developing hypostome (Fig. 12D).

Thus, the expression pattern of HyBra1 during embryogenesis is similar to that found during budding. In both modes of reproduction, the gene is expressed shortly before the onset of both head formation, and continues to be expresssed in the formation of the hypostome.

The positional cloning of T or Brachyury as the gene responsible for the T mutation in mice (Herrmann et al., 1990) marked the isolation of the founding member of a novel class of transcription factors which are all implicated in important developmental processes. This class of genes is characterized by a conserved DNA-binding domain, the T-box, which spans 180-200 amino acids and is usually located at the N terminus of the gene (Kispert and Herrmann, 1993). A number of other T-box-containing genes, which fall into several subfamilies, have been isolated from metazoans ranging from C. elegans to humans indicating the ancient origin of the gene family (Agulnik et al., 1995; Papaioannou and Silver, 1998). However, to date no T-box genes have been identified in diploblasts. Thus, HyBra1 is the first member of this gene family identified in a basal metazoan lacking the mesoderm.

Detailed molecular studies on Brachyury established that the T-box is a DNA-binding domain, which recognizes a palindromic 20 nucleotide sequence in vitro (Kispert and Herrmann, 1993). In a recent crystallographic study, Müller and Herrmann (1997) suggested that Brachyury binds as a homodimer, based on their identification of the critical residues implicated in the dimerization of the Brachyury protein production as well as those involved in the binding of the DNA. Interestingly, HyBra1 shares all of these critical residues suggesting that it might have similar dimerization and DNA-binding properties. These diagnostic residues as well as our phylogenetic analysis of the gene family demonstrates that HyBra1 is a clear orthologue of Brachyury.

In adult hydra, the sequence of events in the development of a head during budding or regeneration, that is the formation of the hypostome and tentacle zone, has been followed using markers that are hypostome-specific (Technau and Holstein, 1995; Hermans-Borgmeyer et al., 1996; Martinez et al., 1997), tentacle-specific (Bode et al., 1988; Weinziger et al., 1994), or mark the entire head (Javois et al., 1986). However, none of these markers appears during the early stages of head formation. Budhead, a hydra homologue of the fork head/HNF-3β members of the winged helix family of transcription factors, is expressed in a broad area during the initial stage (stage 0/1), of bud formation, and only much later is expression reduced to the presumptive hypostome (Martinez et al., 1997). However, during head regeneration, the gene is expressed considerably later than HyBra1. It appears weakly and broadly in the upper half by 4-8 hours following decapitation and is not confined to the apical tip until 12 hours (Martinez et al., 1997).

In sharp contrast, HyBra1 is strongly expressed during the initial or very early stages of head formation during both budding and regeneration. Before the first morphological sign of bud formation [placode formation], a small circular spot with a strong level of expression appears. During head regeneration, expression of the gene begins by 3 hours following bisection, or shortly after wound healing is complete, and reaches a maximal level by 4-6 hours. In both cases, expression of the gene is confined to the region that will form the head. Hence, the expression of HyBra1 provides evidence that the gene plays a role early in head formation.

This conclusion is strengthened by two other results. (i) In the adult, the tissues are continually undergoing patterning to maintain the form of the animal (see Bode and Bode, 1984 for review). Correlated with this ongoing process, HyBra1 expression is continuous in the head, and only in the head of the adult. (ii) Expression of the gene is directly correlated with the level of head regeneration in the head regeneration-deficient mutant, reg-16 (Achermann and Sugiyama, 1985). Should no head form, the gene is never expressed. If regeneration results in a partial head, a low level of HyBra1 is detected while, if a complete head forms, then the normal level is found in the head.

Expression of HyBra1 is confined to the hypostome, the apical half of the head. This is true in the fully formed heads of adults as well as in the developing heads during bud formation and regeneration. In developing heads, expression of the gene is in all the tissue apical to the level where the tentacles emerge, hence, in the developing hypostome. This was confirmed by the DAG results. In animals that formed ectopic tentacles and ectopic heads, HyBra1 expression was always found in the developing hypostome of an ectopic head, but never observed in the vicinity of individual ectopic tentacles. Hence, the gene plays a role in the formation of a specific part of the head, the hypostome.

Since HyBra1 is expressed very early during head formation, it may have a role in the head determination process. Consistent with this is a strong level of HyBra1 expression in regenerating tips before these tips are committed to head formation. However, the expression of HyBra1 is reversible. It vanishes in a regenerating tip in which the process of head formation has been initiated, but is subsequently repressed by transplanting it to a location in the body column not conducive for head formation. Hence, this reversible expression indicates that HyBra1 expression may be part of the head determination process, but is not the critical step leading to irreversible determination.

A large body of transplantation and regeneration experiments have defined three patterning processes governing head formation in hydra predicted earlier on theoretical grounds (Gierer and Meinhardt, 1972; MacWilliams, 1983a,b; for review see Bode and Bode, 1984). One is the source density gradient (or head activation gradient; MacWilliams, 1983b), which is maximal in the head and graded down the body column. Head activation is defined by the ability of a piece of tissue to induce the formation of a secondary head when grafted into the body column of another animal. Hence, this gradient, which is a stable tissue property (MacWilliams, 1983b), is the basis of the regeneration polarity of the body column. As part of the tissue dynamics of an adult hydra, tissue of the body column is continuously displaced apically from the body column into the head and basally into the foot (Campbell, 1967; Otto and Campbell, 1977). Hence, the head activation gradient is also dynamic, and is maintained by a signal[s] produced in the head and continuously transmitted down the body column.

Although it is plausible that HyBra1 expression is a response to a high level of head activation in the hypostome region, this is unlikely. Raising or lowering the overall level of the head activation gradient with DAG or LiCl, respectively, had no effect on the range or location of HyBra1 expression. Were expression of the gene responsive to the gradient level, DAG treatment should have led to an extension in a basal range of its expression into the tentacle zone and possibly into the body column.

While the head activation gradient provides a specific level of a potential to form a head at a given axial level along the body column, a second patterning process termed ‘unstable head activation’ (MacWilliams, 1983b), or ‘head formation process’ (HFP) is the actual process of head formation. It has several properties that coincide with the pattern of HyBra1 expression. (i) As is HyBra1 expression, the HFP is confined to the apical regenerating tip of the lower half of a bisected animal that will form a head. (ii) The half-life of HFP as measured by transplantation experiments is about 12 hours (MacWilliams, 1983a,b), which is also the half-life for the loss of HyBra1 expression in a regenerating tip that is prevented from forming a head. (iii) The variable level of HyBra1 expression, which corresponds to the variable extent (complete, partial or none) of head regeneration in decapitated reg-16 animals, also correlates well the level of HFP in heads, partial heads and upper body columns (=no head formation) (Rubin and Bode, 1982). (iv) Finally, the time required for a regenerating tip to become determined to form a head increases with increasing distance from the head, which implies a decreasing level of head activation (MacWilliams, 1983a,b; Technau and Holstein, 1995). Similarly, the timing of the initial appearance of HyBra1 increases with increasing distance from the head. Thus, the level of expression of the gene is intimately linked with the level of the HFP.

Head inhibition, the third process governing head formation, is continuously produced in the head and transmitted basally to prevent tissue of the body column, which is always capable of forming a head, from doing so (MacWilliams, 1983a). Three results indicate that HyBra1 expression is inversely correlated with the level of head inhibition. (i) Removal of the head of normal animals leads to a rapid drop in the level of head inhibition and a rise in HyBra1 expression. (ii) The head inhibition level remains high following decapitation of a reg-16 adult and, concomitantly, the gene is not expressed. (iii) When a regenerating tip is grafted into a region of high head inhibition, expression of HyBra1 is down regulated. Hence, expression of this gene is actively repressed by head inhibition. In sum, these considerations indicate that, of the three head patterning processes, HyBra1 expression is intimately linked with two of them, the HFP and head inhibition. It provides molecular evidence for the distinct effects and properties of the local, rapidly changing head activation and the slowly changing source density gradient as predicted by a reaction-diffusion theory (Gierer and Meinhardt, 1972). Finally, the transient expression of HyBra1 during foot regeneration provides more evidence for a complication in the patterning processes. Since a head invariably regenerates at an apical end, it is commonly assumed that the HFP alone is initiated at this end resulting in the formation of a head. However, the transient expression of HyBra1 in a regenerating foot suggests that both head and foot formation processes are initiated at the basal end. Other evidence is also consistent with this idea. Cnox-2, the hydra homologue of the Deformed gene, which is expressed in epithelial cells of both layers throughout the animal except in the head (Shenk et al., 1993b), transiently vanishes during foot regeneration. Hence, both head and foot formation appear to be initiated at the basal end, but the head formation process is eventually suppressed, plausibly by head inhibition from the existing head.

During embryogenesis, HyBra1 is initially expressed very late, shortly before hatching, in a small area that will form the hypostome. To our knowledge, this is the first gene isolated from a cnidarian, for which the expression pattern during embryogenesis has been determined. The pattern provides insight into two aspects of reproduction in hydra. One has to do with the timing of the initiation of head formation in hydra. In hydra, the sperm entry point, which is established before fertilization is also the location of the anterior end of the developing embryo (Martin et al., 1997; Freeman, 1981). This implies that during oogenesis an anterior-posterior polarity is set up early, plausibly in the oocyte. However, the very late expression of HyBra1 suggests that formation of the head and the adult body axis is only initiated in a second later step and much later than the establishment of the anterior-posterior polarity in the embryo.

Another issue that is poorly understood is the relationship between the patterning events that occur in the sexual (embryogenesis) and asexual (budding) modes of reproduction. Certainly the early stages are quite different. Embryogenesis in hydra proceeds as is common in animal embryos starting with cleavage divisions and gastrulation, eventually leading to the formation of two embryonic layers, the ectoderm and endoderm (e.g. Martin et al., 1997). In contrast, budding begins with the two layers already formed. Thereafter, the processes are similar in that elongation of the spherical shell of the embryo into a cylinderical shell is fairly similar to the evagination and subsequent elongation of the body wall into a cylindrical protrusion during budding. In both processes, HyBra1 expression first occurs in a small spot in the part of the developing embryo or bud that will be the apical tip of the elongating cylindrical shell. Hence, this pattern provides the first information that the initial stages of elongation and head formation are similar in the two modes of reproduction.

The identification of a Brachyury orthologue in hydra shows that Brachyury is a member of an ancient gene family. The high degree of conservation in the T-box including the dimerization and DNA-binding domains points to a strong selection for this functional part of the gene. However, the function of the homologues in different phyla is not obviously conserved since the homologues are expressed in different tissues and locations. In vertebrates, Brachyury initially has a pan-mesodermal expression in the marginal zone of the blastula, but later becomes restricted to the axial mesoderm during gastrulation (Wilkinson et al., 1990; Smith et al., 1991; Schulte-Merker et al., 1994). Expression is related in the protochordate ascidians, where Brachyury is expressed in the precursors of axial mesoderm, but not in an early pan-mesodermal fashion (Yasuo and Satoh, 1994; Corbo et al., 1997). However, in another group of deuterostomes, the sea urchins, the differences are greater. The Brachyury homologue is expressed in the precursors of secondary mesenchyme, which are derived from the anterior end of the invaginating endoderm during gastrulation and give rise to a variety of cell types (Harada et al., 1995). Among the arthropods, the expression patterns are different again. In Drosophila, the T-related gene (Trg) is expressed in the hindgut primordia during blastoderm stage and later in the differentiating hindgut (Kispert et al., 1994; Singer et al., 1996). Similarly, the Brachyury homologue in Locusta, a short-germ insect, is expressed in the posterior tip of the gastrulating embryo, a region that continues to grow and add segments at the posterior end (Kispert et al., 1994). Clearly, these patterns are quite different from the anterior expression of the hydra homologue in the developing head.

However, the gene may participate in a common molecular pathway in all these organisms. It is striking that, where examined in vertebrates, ascidians, sea urchins, insects and hydra, the expression of the Brachyury homologue coincides with the expression of a fork head/HNF-3β homologue (Weigel et al., 1989; Dirksen and Jamrich, 1992; Harada et al., 1996; Shimeld, 1997; Shimauchi et al., 1997; Terazawa and Satoh, 1997; Olsen and Jeffrey, 1997). In the several organisms examined, this coordinated expression occurs in a variety of developmental circumstances such as different tissues, at different times during embryogenesis, as well as in tissues composed of single cells or epithelial layers. The same pairing of expression holds true for hydra. HyBra 1 and the hydra fork head/HNF-3β homologue budhead are expressed in an overlapping pattern in the endodermal epithelial cells of the hypostome. Whereas HyBra1 is expressed throughout the hypostome, budhead is expressed in the lower half of this structure (Martinez et al., 1997).

Not only are these two genes expressed together in many organisms, there is some evidence that they function in the same pathway. In animal cap assays in Xenopus, it has been shown that Brachyury and pintallavis, the fork head/HNF-3β homologue, have a synergistic effect on the differentiation of dorsal mesodermal fates (O’Reilly et al., 1995). These considerations suggest that, early in metazoan evolution, a molecular pathway involving a Brachyury homologue and a fork head/HNF-3β homologue arose. Subsequently, this pathway was conserved, but co-opted for a variety of functions throughout metazoan evolution.

Finally, the expression of HyBra1 in the endoderm provides evidence for the question of the evolutionary origin of the mesoderm. From comparative embryological studies, it has been proposed that the mesoderm arose from the endoderm in the diploblast-to-triploblast transition (e.g. Siewing, 1969). We note that the expression of the hydra orthologue in the endoderm is consistent with this idea.

We thank Ira Blitz for expert technical advice and helpful discussion, and Michael Sarras for the hydra regeneration cDNA library. This research was supported by grants from NIH (R01-HD08086 and P01-HD27173) and NSF (IBN-9723660) to H. R. B., and an EMBO post-doctoral fellowship to U. T.