Cnidarians: An evolutionarily conserved model system for regeneration?



Cnidarians are among the simplest metazoan animals and are well known for their remarkable regeneration capacity. They can regenerate any amputated head or foot, and when dissociated into single cells, even intact animals will regenerate from reaggregates. This extensive regeneration capacity is mediated by epithelial stem cells, and it is based on the restoration of a signaling center, i.e., an organizer. Organizers secrete growth factors that act as long-range regulators in axis formation and cell differentiation. In Hydra, Wnt and TGF-beta/Bmp signaling pathways are transcriptionally up-regulated early during head regeneration and also define the Hydra head organizer created by de novo pattern formation in aggregates. The signaling molecules identified in Cnidarian regeneration also act in early embryogenesis of higher animals. We suppose that they represent a core network of molecular interactions, which could explain at least some of the mechanisms underlying regeneration in vertebrates. Developmental Dynamics 226:257–267, 2003.© 2003 Wiley-Liss, Inc.


The freshwater polyp Hydra, a member of the ancient phylum Cnidaria, is famous for its regenerative capacity. Just like the multiheaded monster in Greek mythology that grew two new heads for every one cut off, a cnidarian polyp can regenerate a new head after decapitation. Cnidarians are among the simplest living metazoans and evolved approximately 700 million years ago (Bridge et al., 1995; Conway Morris, 2000; Nielsen, 2001; Petersen and Eernisse, 2001). They consist of two body layers, an outer ectoderm and an inner endoderm, separated by an extracellular matrix (mesoglea), and they represent the first animals with a defined body axis and a nervous system.

The regenerative capacity of cnidarians is remarkable. Hydra polyps can be dissociated into single cells that can regenerate as reaggregates into an intact animal within a few days. Cnidarian regeneration occurs by morphallaxis, i.e., a process of repatterning of the existing tissue without the necessity of cell proliferation. This appears to be fundamentally different from regeneration in vertebrates, where wound closure is followed by blastema formation during which cells beneath the wound epidermis dedifferentiate, start to divide, and transdifferentiate (Lo et al., 1993; Brockes, 1997). However, recent data indicate that regeneration of cnidarian tissue shares more similarities to vertebrate (urodele) regeneration than previously thought. In this review, we focus on the molecular regulation of Hydra head regeneration in comparison to vertebrate systems. A comprehensive treatment of classic transplantation experiments, theoretical models, and molecular data in Hydra is found in the review of H. Bode (this issue).

At present, it is unclear to what extent “adult” stem cells are involved in the regeneration process. Such stem cells have been found even in some mammalian tissues, and they have a capacity for developing into a limited number of different cell types (for review, see Stocum, 2001). Of interest, stem cells in cnidarians also mediate the morphogenetic plasticity of the tissue. There are two epithelial stem cell populations, an ectodermal and an endodermal one, which continuously differentiate into head- and foot-specific tissue, which is sloughed off at both ends of the body axis (Campbell, 1967a,b; David and Campbell, 1972). A third stem cell system that is present at least in Hydra and some other Hydrozoans gives rise mainly to nerve cells, nematocytes, and gland cells, but it is not required for regeneration (David and Gierer, 1974).

In all regenerating tissue, a major question is what instructs the cells involved in regenerative processes and which gene products are responsible for the induction of regeneration. The application of molecular and genetic techniques has shown that several crucial genes of early embryogenesis is evolutionarily conserved between vertebrates and insects. Although little is known to date about cnidarian embryogenesis on the molecular level, new molecular data indicate that some of the homologous genes involved in bilaterian embryogenesis act during cnidarian regeneration. Therefore, the intriguing possibility exists that a common set of genes might control at least the early steps of the regeneration process in cnidarians and bilaterians. We presume that the high or even unlimited regenerative capacity characteristic for cnidarians reflects the properties of an ancient patterning system that can generate complete structures (whole organisms), starting from a broad range of initial conditions. It is plausible that molecular patterning systems capable of extremely robust and flexible self-organisation might have been selected during early metazoan evolution and became conserved in higher animals. This review, therefore, mainly emphasizes the cellular and molecular dynamics of this self-organisation system during regeneration, and it represents essentially one lab's view of the problem. We presume that the signaling molecules identified in cnidarian regeneration represent a core network of molecular interactions that could be responsible for at least some of the mechanisms underlying regeneration in vertebrates, e.g., limb regeneration (Brockes, 1997; Gardiner et al., 1999).


Epithelial Stem Cells

Most cnidarian polyps and even some medusae propagate asexually, so that they are in a steady state of constant growth and tissue turnover. In Hydra polyps, it has been shown that both layers of the body wall, the ectoderm and the endoderm, are comprised by dividing epithelial stem cells in which newborn cells are passively displaced upward to form the stinging tentacles, downward to form the foot, or bud off at the sides to make replica animals (Campbell, 1967a,b; for review see Bode and Bode, 1984). An important consequence is that the passively displaced cells have to assess their relative position in the organism continuously. Hence, patterning systems necessary to provide this information are continuously active in Hydra polyps. By contrast, in mammals, most of these morphogenetic signals are mainly active only during embroygenesis.

During regeneration, these morphogenetic signals can be activated or enhanced at the site of wounding. Figure 1 summarizes the major events during the regeneration process in Hydra and other Cnidarians. When Hydra is cut in half, the upper half containing the head will regenerate a new foot, and the lower half containing the foot will regenerate a new head. Regeneration is a rapid process. After wound closure, which takes approximately 1–3 hr, the tentacles of a new head differentiate within 36 hr and a regenerating foot becomes sticky again within 30 hr (Hoffmeister and Schaller, 1985). Far less is known about foot regeneration, yet the mechanisms of head and foot regeneration are probably similar, although there is some evidence that the head system has some supportive function for the foot system (Müller, 1990, 1995, 1996; Lee and Javois, 1993; Forman and Javois, 1999; Javois and Frazier-Edwards, 1991; Schiliro et al., 1999). However, the molecular basis for this phenomenon is completely unclear. If a Hydra is cut into several pieces, the middle portions will regenerate both heads and feet at their appropriate ends, maintaining the initial polarity (Marcum et al., 1977). By comparison, an isolated foot or head alone cannot regenerate an intact animal, only if a head is transplanted on a foot, the missing body region will be intercalated (Holstein and David, 1990).

Figure 1.

Major processes during cnidarian regeneration on the cellular and molecular level. The minimal steps necessary for cnidarian regeneration (solid arrows) involve (1) wound closure of the epithelia followed by (2) the establishment of an organizer and the instruction of epithelial stem cells, and (3) subsequent morphogenesis. There are also nonobligatory steps in cnidarian regeneration (dotted arrows), e.g., dedifferentiation in terminally differentiated tissue of medusae (Schmid, 1992) and cell proliferation in Hydra polyps (Holstein et al., 1991), which could be related to the blastema formation in vertebrates.

Morphallaxis or Epimorphosis?

Classic experiments using Hydra polyps that were either x-ray irradiated (Hicklin and Wolpert, 1973; Noda and Egami, 1975) or treated with the S-phase blocking agent hydroxyurea before regeneration have shown that cell division is not required for the formation of a new head (Cummings and Bode, 1984). This finding has led to the conclusion that regeneration in Hydra is morphallactic (Gilbert, 2000; Wolpert, 2002). However, the cellular dynamics appear to be more complicated. Figure 2A shows that, 12 hr after head removal, the regenerating tip is completely free of S-phase cells (Park et al., 1970; Holstein et al., 1991; Fig. 2B), but by 30 hr, the pattern has completely changed, and the regenerating tip is more strongly labeled than the gastric tissue (Fig. 2C). This finding demonstrates dramatic effects on the cell cycle and proliferation at the regenerating site. Whether the resumption of mitosis is due to a decay of the inhibitory effect or a release of a stimulatory signal is not clear. Molecules of the extracellular matrix (ECM) also appear to contribute to the changes in the proliferation patterns during Hydra head regeneration (Sarras et al., 1991, 1993; Yan et al., 1995, 2000; Shimizu et al., 2002). Accordingly, a loss of the ECM could be related to the reduced mitotic activity and a restoration of the ECM to the delivery of mitogenic factors. Of interest, newborn buds also exhibit a dramatic increase in cell proliferation, and the head region is characterised by a continuous high level of proliferating cells (Holstein et al., 1991). This finding indicates that the loss and restoration of ECM upon wounding is not the only prerequisite for the effect on the cell proliferation during regeneration. It is worth noting that a similar correlation between regeneration and proliferation was also found in other cnidarians, e.g., the Cubopolyp Carybdea marsupialis (Holstein and Stangl, manuscript in preparation).

Figure 2.

Changes in the pattern of epithelial cell cycling during head regeneration. A: Interstitial cell-free polyps were cut at 60–70% body length, pulse-labeled with bromodeoxyuridine (BrdU) at the times indicated, and processed for BrdU visualization by indirect immunofluorescence (camera lucida drawings). B,C: Photomicrographs illustrating the pattern of epithelial cell proliferation during head regeneration at 12 hr (b) and at 36 hr (c). © Academic Press (from Holstein et al., 1991).

Factors that stimulate mitosis of epithelial cells have been described. Schaller and coworkers (Schaller, 1976; Schaller et al., 1977, 1990) have shown that, if Hydra is treated with low concentrations of a neuropeptide, the head activator, mitosis is stimulated in epithelial cells. A local release of this factor could lead to an increased level of mitosis during regeneration. Hobmayer et al. (1997) demonstrated that head activator treatment stimulates epithelial cell division and induces the formation of more tentacle-specific epithelial cells during regeneration. Consistent with these findings, we also found that inhibition of cell proliferation by aphidicolin or hydroxyurea treatment leads to an incomplete regeneration of head structures after 36 hr (Hobmayer and Holstein, unpublished observations). Therefore, we presume that morphogenetic signals and growth factors that are released during head regeneration in cnidarians can also induce cell proliferation, which is required for a complete regeneration of the full-sized structure (Fig. 1).

Dedifferentiation as a Necessary Step in Regeneration?

Although most polyps grow constantly, medusae normally exhibit only a limited capacity to grow, which is related to the differentiation of a sexually mature animal (there is an exception from this rule, because in some hydrozoan life cycles, medusae can propagate asexually, e.g., Rathkea and Sarsia; for review, see Tardent, 1978). Nevertheless, even terminally differentiated medusa tissue can regenerate (Fig. 1). Pioneering studies of the jellyfish Podocoryne carnea showed that the muscle tissue of adult medusae transdifferentiates into several different cell types during regeneration (for review, see Schmid, 1992; Schmid and Reber-Müller, 1995). When striated muscle cells were explanted from the subumbrella and cultured in the presence of ECM degrading enzymes, diacylglycerol, or the phorbol ester 12-O-tetradecanoylphorbol-13-acetate, they dedifferentiated and started to proliferate 24–48 hr later (Schmid et al., 1998). Such cells formed flagellae first, which can be interpreted as a sign for the naive state of this tissue, and later they differentiated into smooth muscle cells, sensory cells, or nematocytes (Alder and Schmid, 1987). During transdifferentiation several mesodermal genes, which are specific for striated muscle differentiation, were turned off (Müller et al., 1999; Yanze et al., 1999; Spring et al., 2000). This finding is in accord with the morphologic data.

Cnidarian vs. Vertebrate (Urodele) Regeneration

A comparison of the basic features of cnidarian regeneration with the regeneration process in vertebrates indicates that both systems share some common principles, despite the prevailing view that one is morphallactic (Cnidaria) and one is epimorphic (vertebrates). The remarkable regeneration capacity of urodele amphibians involves the local dedifferentiation of stump tissue to form a blastema and new growth within the blastema to form distal structures (for review, see Gardiner and Bryant, 1996; Brockes, 1997; Gardiner et al., 1999; Wolpert, 2002). Yet, recent elegant molecular and genetic data suggests that limb bud formation and regeneration in vertebrates involves a prepatterning of the whole limb at an early stage in a small morphogenetic field (prespecification model), rather than a distal transformation (progress zone model) of the growing blastema (Sun et al., 2002; Dudley et al., 2002; Duboule, 2002). This finding suggests that, also in vertebrate limb formation and regeneration, patterning of a morphogenetic field occurs only at small scale and growth is needed only to add in cells to produce a structure of larger size. The reason for this may lie in the fact that morphogens can act only over a distance of several cell diameters and a maximum of 300 μm.

In cnidaria, neither local dedifferentiation nor blastema formation are obligatory steps in the regeneration process, which is probably due to the high morphogenetic plasticity of the tissue. Epithelial stem cells are always competent for morphogenetic signals released at the site of wounding, and they can differentiate even when cell cycling is blocked. Hence, it appears that, in Hydra, cell division is not completely indispensable for the regeneration of the complete structure. On the other hand, in vertebrates, growth of the blastema might only be necessary to enlarge an already patterned field. Thus, Hydra and vertebrate regeneration might share more features than commonly thought. The crucial and obligatory step during regeneration in both systems is a prepatterning of the regenerating tissue. Regeneration always gives rise to structures with positional values proximal to the site of regeneration. In cnidarians there is substantial evidence that an apical signaling center, the head organizer, is the driving force for this process. In a first step, this organizer has to be re-established at the regenerating tip. Then, signals emanating from this organizing center pattern and respecify the tissue proximal to the wounding site. This emphasizes the importance of the reestablishment of an organizer during the initial steps in regeneration. In the following, we will discuss the molecular features of the Hydra head organizer and how it is reestablished during regeneration, particularly in reaggregates.


The capacity to regenerate a head is higher at the apical end than at the basal end of Hydra's body axis (Webster, 1966a,b; Wilby and Webster, 1970a,b; Wolpert et al., 1971, 1972; MacWilliams, 1983b; Technau and Holstein, 1995). Transplantation experiments have shown that the peak of this activity is localized in the hypostome. A small piece of tissue from the hypostome induces a secondary body axis when grafted laterally to another polyp (Browne, 1909; Mutz, 1930; Yao, 1945; Broun and Bode, 2002). Hence, in terms of organizer activity, Hydra's hypostomal tissue is equivalent to the dorsal lip of the frog embryo, the Spemann-Mangold organizer, which also induces a secondary body axis when grafted to the ventral side of the embryo (Spemann and Mangold, 1924).

Until recently, it was rather unclear when and how the organizer and its molecular composition arose during animal evolution (Harland and Gerhart, 1997; Knoll and Caroll, 1999). However, the discovery that the same set of genes is active in the organizer of all vertebrates suggested that basic features of signaling centers acting as an organizer might have arisen earlier in metazoan evolution. Potential signaling molecules that could act as diffusible morphogens similar to those in vertebrates have been identified recently in Hydra (Hobmayer et al., 2000). It is likely that the gradient of inductive capacity is mediated by a gradient of these signaling molecules released from the Hydra head organizer.

A Wnt ligand (HyWnt) and the cytoplasmic mediators Dishevelled (HyDsh), GSK3 (HyGSK3), and β-Catenin (Hyβ-Cat) were cloned from Hydra (Hobmayer et al., 2000). In a two-hybrid screen with Hydra β-Catenin as bait, the transcriptional coactivator Tcf (HyTcf) was also identified (Hobmayer et al., 2000). A Hydra member of the family of Frizzled receptors was identified by Minobe et al. (2000). Hence, the core Wnt pathway is present in Hydra. In situ hybridization revealed that Wnt signaling acts in axial patterning and during head regeneration in Hydra. HyWnt is expressed in a small number of ectodermal and endodermal epithelial cells in the apical tip of the hypostome, which represents the Hydra head organizer. HyTcf expression is also restricted to the hypostome of the polyp, but the HyTcf expression domain is broader than the HyWnt spot encompassing the entire hypostome and, thereby, possibly demarcating the range of action of the HyWnt ligand (Hobmayer et al., 2000). During head regeneration, HyWnt, HyTcf, and Hyβ-cat are among the earliest genes to be up-regulated, within 30–60 min after wound healing (Fig. 3). In the budding zone, where the new body axis of the daughter polyp is initiated (Otto and Campbell, 1977), activation of the HyWnt pathway also starts with an up-regulation of Hyβ-Cat and HyTcf and is followed by HyWnt expression in a spot of 10–15 cells (Hobmayer et al., 2000). These data indicate a pivotal role for the members of the Wnt-pathway in setting up the Hydra head organizer.

Figure 3.

Local up-regulation of Hyβ-Cat, HyTcf, and HyWnt expression in head regenerating tips. Whole-mount in situ hybridizations at 1 and 48 hr after head removal. The Wnt signaling cascade is up-regulated during head regeneration. HyWnt is expressed in a group of 15–50 cells defining the head organizer region at the tip of the hypostome. © MacMillan Press (from Hobmayer et al., 2000).

There is also evidence for a second major signaling system in cnidarians, i.e., the TGFβ/Bmp signaling pathway and its antagonist Chordin (Samuel et al., 2001; Lelong et al., 2001; Hobmayer et al., 2001; Hayward et al., 2002), which are involved in early embryonic axis formation of vertebrates. A Bmp ligand (Reinhardt and Bode, personal communication), a highly conserved receptor-regulated Smad1 homologue (Hobmayer et al., 2001a), and the Bmp antagonist Chordin (Rentzsch, Hobmayer, and Holstein, unpublished observations) have been found in Hydra. The expression patterns of HySmad1 and Chordin during regeneration are consistent with the hypothesis that Bmp signaling is suppressed by Chordin, which would indicate a conservation of the molecular interactions of dorsoventral patterning from Hydra to vertebrates (for review, see DeRobertis and Bouwmeester, 2001; Shilo, 2001).

These data demonstrate that at least two major signaling systems that are responsible for the function of the vertebrate organizer (DeRobertis and Sasai, 1996; DeRobertis and Bouwmeester, 2001) are already present in Hydra. This finding suggests that the core Wnt signaling pathway as well as the TGFβ/Bmp signaling pathway and its antagonist Chordin were present in the common ancestor of diploblastic cnidarians and the triploblastic Bilateria and, hence, most likely were a basic feature of early multicellular animals. Notably, a TGF-beta receptor was found in sponges (Suga et al., 1999), although its expression is unclear to date.

It should be also pointed out that transcription factors that play a role in the vertebrate organizer have been isolated from Hydra, such as the HNF3β homolog budhead (Martinez et al., 1997), the homeobox gene goosecoid (Broun et al., 1999), and the T-box gene Brachyury (Technau and Bode, 1999). These genes are all expressed in the organizer region in Hydra (for review, see Galliot, 2000) and may have a function in regulatory feedback loops together with the Wnt and TGFβ signaling cascades during head regeneration.


Hydra can be completely dissociated into single cells and will regenerate intact animals within 3 to 4 days (Fig. 4) (Noda, 1971; Gierer et al., 1972). After dissociation into a single cell suspension and subsequent reaggregation, all existing gradients of the polyp and any positional information are destroyed and have to be reestablished (Gierer et al., 1972; Sato et al., 1992; Technau and Holstein, 1992). This experimental system is unique in that it is possible to analyze regeneration from the very beginning and under conditions of de novo pattern formation on the cellular and molecular level.

Figure 4.

Regeneration of intact Hydra polyps from reaggregates of dissociated single cells. Formation of the tissue bilayer by ectoderm- and endoderm-specific cell sorting mechanisms is finished within 24 hr. Development of polyp structures occurs within 96 hr.

Reaggregation proceeds through a well-defined sequence of morphogenetic processes: initial cell adhesion, ecto–endo cell sorting, formation of the epithelial bilayer, differentiation of a head and foot, and finally separation into intact polyps. To establish the epithelial bilayer configuration, three interaction types are necessary: ecto–ecto, endo–endo, and ecto–endo cell interactions. The formation of homotypic ectodermal and homotypic endodermal aggregates was first observed during rotary culture of dissociated cell suspensions (Technau and Holstein, 1992) and confirmed by laser-cell trapping experiments, where the adhesive forces between individual cells were directly determined (Sato-Maeda et al., 1994). Pairs of endodermal cells exhibited stronger adhesive forces than pairs of ectodermal epithelial cells, and there was no initial heterotypic interaction between individual ectodermal and endodermal cells. Hobmayer et al. (2001b) used rotary culture of dissociated cell suspensions and found that aggregation of epithelial cells proceeded in two steps: first homotypic (ecto–ecto and endo–endo) interactions created small cell clusters, then heterotypic interactions between ectodermal and endodermal cell clusters led to the formation of larger aggregates. This switch from homotypic to heterotypic interaction occurred at a critical aggregate size of 10–20 epithelial cells and indicates that adhesive forces between ectodermal and endodermal cells became significantly stronger than adhesive forces between either ectodermal or endodermal cells (Hobmayer et al., 2001b). At present it is unclear whether this change in the cell–cell affinities in Hydra reaggregates can be explained by a depletion of a limited pool of cell adhesion molecules, a redistribution and clustering of preexisting heterotypic adhesion molecules (Grawe et al., 1996), or the new expression of heterotypic adhesion molecules due to an activation of intracellular signaling cascades in homotypic aggregates (Fagotto and Gumbiner, 1996).

The formation of ecto–endodermal cell clusters finally leads to the formation of ectodermal and endodermal tissue layers. During this epithelial sheet formation, the ectodermal tissue layer begins an epiboly-like movement to spread over the endoderm (Kishimoto et al., 1996). In parallel, the endodermal layer organizes beneath an intact ectodermal layer (Murate et al., 1997), suggesting that the formation of both epithelial layers is driven by the ectoderm. This dramatic process of cell sorting and restoration of cell polarity are completed within the first 12 hr of the reaggregation process (Gierer et al., 1972; Technau and Holstein, 1992). Once the ectodermal and endodermal layers are established, no further rearrangement occurs. With respect to their original axial position in Hydra, no cell sorting has been observed (Sato et al., 1992; Technau and Holstein, 1992). This finding indicates that, after dissociation into single cells, there is no predisposition of erstwhile head cells to sort out into head tissue and that the formation of new activation centers and head organizers occurs by true de novo pattern formation (Gierer et al., 1972; Technau and Holstein, 1992).

In further experiments, it was shown that a community effect regulates the formation of activation centers in Hydra (Technau et al., 2000). Labeled cell clusters were produced from regenerating stumps that have a high competence for head induction (MacWilliams, 1983b). The regenerating tissue was dissociated into single cells, aggregated in rotary culture, and the resulting cell clusters were fractionated by size (Technau et al., 2000). Small labeled cell clusters consisting of 10–15 cells (60 μm in diameter) were added to an unlabeled cell suspension, and approximately half of them were found to be present in a developing head after reaggregation. The labeled cells were confined to the hypostome while the tentacles were formed by the host tissue (Fig. 5A,B). This finding shows that a cluster of only 10 to 15 cells is necessary and sufficient to instruct and recruit surrounding host tissue and initiate the formation of a new head, which is the definition of an organizer sensu strictu. Single cells or very small clusters (30 μm in diameter) consisting of a few epithelial cells have virtually no elevated capacity of induction. These data demonstrate that a community effect (Gurdon et al., 1993) between these cells is essential to create a stable signaling center. Further experiments indicated an activation range of approximately 45 μm (two to three epithelial cell diameters; Technau et al., 2000), which is in the estimated diffusion range of known morphogens, i.e., wnt/wingless in Drosophila (Gurdon and Bourillot, 2001; see below).

Figure 5.

Head induction by activated cell clusters. A: A two-headed aggregate (96 hr) with a head containing a green-labeled 60-μm cluster of aggregated cells from dissociated 12-hr regenerating tip tissue. B: Efficiency of cell clusters to induce head formation. Head formation frequency of single cells (30 μm) and different cell cluster sizes were scored 80 hr after aggregation in carrier tissue derived from whole polyps (filled circles, hatched line) and polyps lacking the upper fifth (filled triangles, solid line; P < 0.001; n = 28–53; means [SEM, three experiments]). Control clusters derived from the corresponding carrier tissue are indicated by open circles and open triangles. C: Effect of head inhibition in aggregates. Aggregates that contained competing 120-μm and 60-μm cell clusters in a single aggregate. A 120-μm cell cluster inhibited the formation of a head, the 60-μm cell cluster; the correlation of head formation frequency of 60 μm cell clusters with their distance from the nearest head is shown. © A–C from Proceedings of the National Academy of Sciences USA 2000;97:12127–12131. Scale bar = 200 μm in A.


That small clusters of cells can induce surrounding tissue to differentiate into head tissue suggests that diffusible morphogens like Wnt might play an instructive role in the activation process. The expression pattern of HyWnt was examined in early reaggregates (Hobmayer et al., 2000; Technau et al., 2000) and found to occur in small spots comprising only a few epithelial cells (Fig. 6) by 24 hr. At this time, cells have completely sorted out into ectodermal and endodermal layers (Gierer et al., 1972; Technau and Holstein, 1992), indicating that HyWnt activation requires intact epithelial tissue. By 96 hr, the HyWnt expression domains have enlarged to their final size in future hypostomes (Fig. 6). The size of early HyWnt spots is 50–60 μm, which corresponds to the minimal cluster size that can act as an organizer (see Fig. 5B).

Figure 6.

Expression dynamics of HyTcf,HyWnt, and HyBra1 during aggregate development. In situ hybridization reveals patterning events during head organizer formation. HyWnt and HyBra1 appear simultaneously in small spots (24 hr). which enlarge during later stages (96 hr), and precede formation of morphologic head structures by approximately 2–3 days. All spots eventually develop into heads. © Proceedings of the National Academy of Sciences USA 2000;97:12127–12131 and Nature 2000;407:186–189.

Reaction-diffusion models of pattern formation predict an autocatalytic feedback loop during the activation process. Preliminary data suggest a possible feedback control in the HyWnt pathway (Fig. 7). Hyβ-cat and HyTcf are expressed uniformly throughout aggregates and later become restricted to domains where new heads are being formed (Fig. 6). A uniform, but high level of HyTcf and Hyβ-Cat might provide a competence to cells to produce HyWnt. Activation of HyWnt might be a stochastic process which is initiated in single cells, but only maintained if, by chance, neighboring cells also express HyWnt. Alternatively, HyWnt might activate and stabilize its own expression directly by means of its transcriptional mediators Hyβ-Cat and HyTcf and later become restricted to domains where new heads are being formed. Notably, the expression of HyWnt always preceded the apparent restriction of domains in the initially symmetrical environment of an aggregate, and all HyWnt domains finally form a head (Technau et al., 2000). Both scenarios are consistent with the idea of an autocatalytic feedback loop and that HyWnt is a direct target gene of an active Hyβ-Cat/HyTcf complex. This finding is in line with findings from Drosophila, where autocatalytic self-activation of Wg and a functional Tcf-binding site in the Wg promoter have been demonstrated (van de Wetering et al., 1997; Lessing and Nusse, 1998).

Figure 7.

Model of putative positive feedback in Hydra Wnt signaling. Preliminary evidence and comparison with higher metazoans support the view that autocatalytic self-activation of the Wnt-pathway and a feedback between HyWnt and the transcription factor HyBra1 are involved in establishment and maintenance of the HyWnt signaling cascade (multiple arrows).

There is additional evidence that HyWnt might be coupled also by a positive feedback with another early head gene, HyBra1 (Fig. 7), a Hydra homologue of the T-box gene Brachyury (Technau and Bode, 1999). In aggregates, size and time of appearance of small HyBra1-positive spots are equivalent to the HyWnt expression dynamics. Interestingly, HyBra1 also shows synexpression with HyWnt during budding and head regeneration as well as in adult polyps, although the HyBra1-positive domain in the steady state hypostome is broader than the HyWnt-positive domain (Technau and Bode, 1999). A putative Tcf-binding site has been identified recently in the HyBra1 promoter (Technau, unpublished data), which supports the idea that Brachyury and Wnt are members of a synexpression group in Hydra. In mouse embryos and mouse cell lines, Brachyury is a direct target gene of Wnt3a signaling (Liu et al., 1999; Galceran et al., 2001), and Brachyury itself activates transcription of Wnt11 in Xenopus (Tada and Smith, 2000). Direct experimental proof for such a feedback loop in Hydra by testing the effect of exogenous HyWnt on HyBra1 expression or by loss-of-function experiments with the HyBra1 gene would be of particular importance.


Patterning processes have to be restricted to the regenerating tissue. On the theoretical level, reaction–diffusion mechanisms (Turing, 1952) predict that an inhibitor is produced by the activation center and transmitted to the surrounding tissue to prevent the initiation of another activation center (Gierer and Meinhardt, 1972; Meinhardt, 1982, 1993). Transplantation experiments using intact Hydra have provided strong evidence for such an inhibitory gradient extending from head into body column (MacWilliams, 1983a,b). The range of inhibition in regenerating aggregates was determined by introducing cell clusters of different size into a host aggregate where larger cell clusters (120 μm) exerted an inhibitory influence on the smaller clusters (60 μm). It was found that approximately 50% of the small clusters were not involved in head formation at a distance of 600 μm from the large clusters, whereas essentially all of them were in heads at 1,000 μm from the large clusters, indicating an effective range of inhibition of approximately 800–900 μm (Technau et al., 2000). By comparison, the range of activation was approximately 45 μm, hence, 20× shorter, which fits with theoretical predictions (Gierer and Meinhardt, 1972; MacWilliams, 1982).

The molecular nature of the inhibiting gradient is currently unclear. Using antibodies to the gap junction proteins, Fraser et al. (1987) could perturb the head inhibition gradient in grafting operations, suggesting that the inhibition gradient is mediated by cell–cell communication by means of gap junctions. However, the inhibition gradient might also involve long-range morphogens regulating the properties of epithelial cells. In the Drosophila wing disc and the amphibian blastula animal cap (Day and Lawrence, 2000; Lawrence, 2001), members of the TGFβ/Bmp family act as long-range morphogens up to 300 μm and concentration-dependent effects have been confirmed (Gurdon and Bourillot, 2001). Changes in production of Dpp, the Drosophila Bmp homolog, can substantially redesign the Drosophila wing, indicating a long-range action. However, recent studies using GFP-Dpp constructs suggest a more complex mode of gradient formation, including endocytotic trafficking and degradation (Entchev et al., 2000). The antagonistic factor to Dpp/BMP2-4 is Sog/Chordin, which forms an opposing gradient. In Xenopus embryos the Chordin gradient can have a range of at least 450 μm when overexpressed, although its in vivo range, which is restricted by the metalloprotease Xolloid, appears to be smaller (Blitz et al., 2000). Recently, it has been shown directly in Drosophila that Sog forms a protein gradient in dorsal cells of the embryo (Srinivasan et al., 2002). On the dorsal side, Tolloid (Tld) degradation and a dynamin-dependent retrieval of Sog act as a dorsal sink for active Sog (Srinivasan et al., 2002). This long-range activity of Sog/Chordin and the related degradation by Tolloid/Xolloid could be an important component of the long-range inhibition phenomena and size control of Hydra during regeneration.


Earlier work has suggested that regeneration by morphallaxis (as found in Hydra) and epimorphosis (as found in vertebrates) are fundamentally different. New experimental results from Hydra and vertebrates reveal, however, that regeneration in these evolutionarily extremely distant phyla share some similarities (see Fig. 1).

In an initial phase of regeneration, after wound closure, epithelial stem cells can respond to changes of patterning signals at the wounding site. If the cells are differentiated, as is the case in medusae of Podocoryne (Schmid, 1992), they first have to dedifferentiate to adopt a new fate. However, this dedifferentiation appears not to be an obligatory step, as there is no evidence for it in Hydra. Because all epithelial stem cells along the body column of Hydra are competent to respond to the regeneration signal, it is an important and unsolved question whether this locally restricted response is due to a positive stimulatory signal or to a release of an inhibitory signal at the regenerating site.

In the next phase of regeneration, i.e., the formation of an organizer and the establishment of a prepattern, substantial progress has been made on the molecular level. It is striking to note that a set of highly conserved genes, i.e., the Wnt and TGF-beta pathways as well as members of the T-box gene family, are involved in cnidarian regeneration. The data reviewed here indicate that Hydra, a representative of one of the oldest metazoan phyla, uses these genes in a signaling center for regulating the establishment and regeneration of its major body axis. These genes also have a crucial role in the patterning of higher animals. This finding indicates the antiquity of this patterning system and points toward an origin of signaling centers in the earliest multicellular animals. Eventually, patterning signals have to be translated into morphogenesis and differentiation of cells. (Non-canonical) Wnt-signaling and the T-box transcription factor Brachyury are good candidates for mediating patterning to morphogenesis. In chordates, Brachyury is a target gene of Wnt, TGF-β, and FGF signaling and a transcriptional activator of many genes involved in convergence and extension, cell adhesion, and cytoskeleton (Tada et al., 1998; Takahashi et al., 1999; Tada and Smith, 2000).

At present, we are far away from a comprehensive view of the genetic network controlling regeneration and the reestablishment of a body axis in cnidarians. Genomic approaches, screens to identify the extracellular antagonists of signaling molecules, and promotor analyses of the involved genes will help to understand this genetic network. This progress will lead to the identification of cell-type–specific downstream genes, and to a better understanding of the question to what extent cell proliferation is involved in the cnidaria regeneration.

Another open question, not addressed in this review, is how peptide signaling molecules are related to the known signal transduction pathways. These peptides also affect cnidarian regeneration (Schaller, 1973; Endl et al., 1999; Hampe et al., 1999; see the review of Fujisawa in this issue) and may represent a phylogenetically ancient feature of cnidarians. However, it is totally unclear at present to what extent these cnidarian peptides have homologues in vertebrates. A gene encoding members of the LWamide peptide family, one of the peptide families identified by the Hydra Peptide Project (Takahashi et al., 1997; Bosch and Fujisawa, 2001), has been identified recently in Caenorhabditis elegans (see review of Fujiswa in this issue). Thus, signaling peptides easily could have been overlooked with algorithms that are normally used in sequencing projects.


We thank C.N. David (Munich) for his critical comments on the manuscript.