An ancient Wnt-Dickkopf antagonism in Hydra

The dickkopf (dkk) gene family encodes secreted antagonists of Wnt signalling proteins, which have important functions in the control of cell fate, proliferation, and cell polarity during development. Here, we report the isolation, from a regeneration-specific signal peptide screen, of a novel dickkopf gene from the fresh water cnidarian Hydra. Comparative sequence analysis demonstrates that the Wnt antagonistic subfamily Dkk1/Dkk2/Dkk4 and the non-modulating subfamily Dkk3 separated prior to the divergence of cnidarians and bilaterians. In steady-state Hydra, hydkk1/2/4-expression is inversely related to that of hywnt3a. hydkk1/2/4 is an early injury and regeneration responsive gene, and hydkk1/2/4-expressing gland cells are essential for head regeneration in Hydra, although once the head has regenerated they are excluded from it. Activation of Wnt/β-Catenin signalling leads to the complete downregulation of hydkk1/2/4 transcripts. When overexpressed in Xenopus, HyDkk1/2/4 has similar Wnt-antagonizing activity to the Xenopus gene Dkk1. Based on the corresponding expression patterns of hydkk1/2/4 and neuronal genes, we suggest that the body column of Hydra is a neurogenic environment suppressing Wnt signalling and facilitating neurogenesis.


INTRODUCTION
Secreted Wnt glycoproteins constitute one of the major families of cell signalling molecules (Logan and Nusse, 2004;Nelson and Nusse, 2004;Nusse, 2003), and their appearance in early metazoan evolution was probably linked to the origin and evolution of multicellular animals (Kusserow et al., 2005). In vertebrate embryonic development, Wnt ligands initiate signalling by interacting with two types of receptor molecules: the seven-pass trans-membrane protein Frizzled and the low-density lipoprotein receptor Lrp5/Lrp6. It is currently assumed that Wnt ligands bind to both Lrp5/Lrp6 and Frizzled to form a functional ligand dual-receptor complex that activates the canonical Wnt-␤-Catenin pathway. Dickkopf 1 (Dkk1), a major secreted Wnt antagonist, binds to Lrp5/Lrp6 and its coreceptor Kremen to inhibit Wnt signalling Mao et al., 2002;Mao et al., 2001;Semenov et al., 2001). Inhibition of canonical Wnt signalling by Dkk1 is essential for head formation and limb patterning in the vertebrate embryo Glinka et al., 1998;Grotewold and Ruther, 2002b;Hashimoto et al., 2000;Kazanskaya et al., 2000;Mukhopadhyay et al., 2001;Shinya et al., 2000).
Vertebrate Dickkopf molecules consist of two cysteine-rich domains (CRD1 and CRD2), which are separated by a spacer region, diagnostic for grouping of Dkk proteins. It has been shown that CRD2 is necessary and sufficient to repress canonical Wnt signalling by competing with the Wnt-Frizzled complex for binding to the Lrp5/Lrp6 receptor. CRD1 is thought to have a modulating function on CRD2 (Brott and Sokol, 2002;Li et al., 2002).
Hitherto, it is completely unknown at what point in metazoan evolution Wnt-Dickkopf antagonism was established. Although the receptor proteins Frizzled and Lrp5/Lrp6 from vertebrates have homologues in insects, no Dickkopf protein could be identified in the insect and nematode genomes. A Dkk3-related protein was recently identified in Hydra (Fedders et al., 2004), and it was suggested that Dkk3 represents the ancestral Dickkopf type. It was furthermore proposed that subsequent genome duplication created the vertebrate Dkk1/2/4 subfamily (Fedders et al., 2004). According to this proposal, the Dickkopf-Wnt antagonism was established relatively late in metazoan evolution, i.e. concomitant with the radiation of vertebrates.
We describe here a new Dickkopf-related molecule from the freshwater polyp Hydra and the starlet sea anemone Nematostella. The novel Hydra Dickkopf protein was isolated as regeneration specific in a yeast signal peptide secretion screen. Structural and phylogenetic analysis indicates that this dkk gene is a dkk1/2/4 homologue, from which the vertebrate Dkk1, Dkk2 and Dkk4 arose by gene duplication. Moreover, our expression and functional analyses support the view that the Wnt-Dickkopf antagonism is phylogenetically very old, and probably had a major function in setting up a neurogenic environment in the body column of the prebilaterian Hydra.

Hydra culture and experiments
Hydra strains Hydra magnipapillata 105 and sf-1 (Sugiyama and Fujisawa, 1978), as well as Hydra vulgaris Basel and AEP (Martin et al., 1997), were cultured at 18°C (Takano and Sugiyama, 1983), fed five times a week, and used 24 hours after feeding for experiments. Gametogenesis was induced in polyps of Hydra vulgaris strain AEP as described (Rentzsch et al., 2005).

Elimination of interstitial cell lineage
Temperature-sensitive interstitial cells of Hydra magnipapillata strain sf-1 were eliminated by culturing polyps at 28-30°C for 5 days; afterwards starved polyps were kept at 18°C for up to six weeks (see Results). Loss of interstitial cells and derivatives was examined in macerates (David, 1973).

Regeneration and wounding experiments
Polyps were bisected at 20%, 50% or 80% body length by using a small scalpel. In some experiments, heads were removed without injury by tying a knot with a thin hair around the subhypostomal region of the polyp as described (Newman, 1974). This caused the head to be gradually pinched off without leaving an open wound with exposed endoderm. For wounding, one deep cut was set into the middle of the body column (30% body width). All animals were transferred into fresh medium after treatment.

Alsterpaullone treatment
Alsterpaullone (Calbiochem) was dissolved in DMSO and diluted with hydra medium to 5 M. Daily fed polyps were incubated in alsterpaullone for 24 or 48 hours.
Cell culture experiments TOPFLASH assay in HEK293T cells was performed as previously described . Transformation of cells was carried out in 96well plates, in triplicate, with 1 ng each of GFP, renilla and hfrizzled1, 10 ng TOPFLASH vector, 8 ng mwnt1, 3 ng human LRP6, 20 ng hydkk1/2/4, 5 ng xdkk1 and 3.125 ng hkrm1. Samples were supplemented with pCS2 + vector to 100 ng DNA. After 24 hours, the transformation efficiency was checked by monitoring GFP fluorescence. Firefly luciferase activity was normalized against renilla.

In situ hybridization (ISH)
Whole-mount ISH was performed as described (Grens et al., 1996;Martinez et al., 1997) at a probe concentration of 0.05 ng/l for 36 hours. Double ISH was performed as described (Hansen et al., 2000;Philipp et al., 2004), with NBT/BCIP substrate followed by Fast Red substrate. Sense and antisense riboprobes were produced from a 534 bp fragment of hywnt3a (bp 624-1158) and from full-length hydkk1/2/4 using the DIG or FITC RNA Labelling Kit (Roche).

Microscopy
Specimens were analyzed using a Zeiss Stemi SV 11 binocular, a Zeiss Axiovert 100, or a Nikon Eclipse-80i microcope, both equipped with interference contrast. For microphotography, either a Diagnostic Instruments Spot-II, a Canon PowerShot G5 or a Nikon DSL-1 camera were used. Micrographs were processed using the manufacturer's software in combination with MetaMorph or Adobe Photoshop software.

Xenopus experiments
In vitro fertilization, embryo culture, staging, microinjection and culture of explants were carried out as described (Gawantka et al., 1995). mRNA was produced with the MessageMachine SP6 Kit (Ambion) from the hydkk1/2/4 ORF and the hywnt3a ORF in the NotI linearized pCS2+ vector; mRNA was purified with P6 Spin Columns (Bio-Rad). Injections were done radially into all blastomeres of four-cell-stage Xenopus laevis embryos (100 pg), or in case of hywnt3a radially into ventral blastomeres (1 ng). Control mRNA (10 pg xdkk1, 12.5 pg xwnt8) was injected as well.

Animal cap assay
Xenopus laevis embryos were injected in the animal pole of all blastomeres at the four-cell stage: xwnt8 (100 pg), xdkk1 (200 pg), hydkk1/2/4 (6 ng). Animal caps were explanted at stage 8-9 and analyzed by RT-PCR at stage 10 for the induction of siamois expression. xbra expression validates mesoderm-free caps, histone-4 expression was monitored for normalization.

Identification of a Dickkopf-related molecule from
Hydra in a yeast signal peptide secretion screen In order to identify growth factors and their antagonists in Hydra, we performed a signal peptide secretion screen (Jacobs et al., 1999;Klein et al., 1996). Because Marcum and Campbell (Marcum and Campbell, 1978) have shown that Hydra lacking nerve cells, nematocytes and interstitial cells develop normally, we tried to eliminate the highly abundant transcripts of this cell line. We used heat-shocked animals of the mutant strain sf-1 from Hydra magnipapillata, which had lost their interstitial stem cells and nematocytes. After inducing head regeneration, we collected the heads and isolated regenerating stumps at various time-points after head removal (0.5 to 24 hours), and prepared a regenerationspecific cDNA library (Fig. 1A). The most abundant clone (10 %) in the signal peptide secretion screen contained a 328 bp fragment encoding a cysteine-rich protein. Using 5ЈRACE we completed the sequence obtaining a 395 bp transcript with short 5Ј and 3ЈUTRs (88 bp and 31 bp, respectively). The open reading frame encodes a protein of 73 amino acids, containing a cysteine-rich motif and a signal peptide of 19 amino acids (see Fig. S1A in the supplementary material). BLAST searches at NBCI revealed with highest score (E-value 0.005) similarity to the CRD2 of vertebrate Dickkopf subfamily members Dkk1, Dkk2 and Dkk4. The CRD2 motif of the Dkk family shares a high structural similarity with colipases (coenzymes of lipases) and has been assigned to the colipase fold (Aravind and Koonin, 1998). However, the similarity of the novel Hydra Dkk protein to colipases was much lower (Evalue 0.62).

Phylogenetic analysis of the Dickkopf-related molecule reveals features of a putative Dkk1/2/4 precursor
Another Dkk-related molecule was recently identified in Hydra, which shares structural features with the vertebrate Dkk3 and was classified as a Dkk3 orthologue (Fedders et al., 2004). Because the novel Hydra Dkk-related molecule identified in the screen shows a higher similarity to CRD2 of the vertebrate Dkk1/2/4 subfamilies, we hypothesized that it might be a new member of the Dkk family in cnidarians. For a better overview, we performed a search for members of the Dkk family using EST data from several other invertebrates. We found ESTs from two other cnidarians, Nematostella vectensis and Acropora millepora, from the urochordate Branchiostoma belcheri tsingtaunese, and from the cellular slime mold Dictyostelium. No Dkk-like proteins were found in insect and nematode genomes.
The Nematostella contigs encode for two different Dkk-like proteins, each comprising two complete cysteine-rich domains (see Fig. S1A,B in the supplementary material). Alignments using TCoffee ( Fig. 2A), but also ClustalW and Muscle (data not shown), reveal a higher similarity of cnidarian Dkks to vertebrate Dkks than to vertebrate colipases. In cnidarian Dkks, all ten cysteine residues are completely conserved, and some other amino acids, like glycines and lysines, are partially conserved among the species. Vertebrate Dkk1, Dkk2, and Dkk4 share a number of motifs that cannot be found in the vertebrate Dkk3 subfamily. Cnidarian Dkks share a number of amino acids with both the Dkk1/2/4 and the Dkk3 subfamilies, but again not with the colipases (Fig. 2A).
The IGPNNI ML-tree ( Fig. 2C) of the CRD2 shows two distinct cnidarian Dkk subtrees: one containing the more dkk3-like sequences, the other containing the Hydra Dkk-related protein isolated in the secretion screen and the second nvdkk. The latter subtree is positioned closer to the highly supported cluster of the vertebrate Dkk subfamilies 1, 2 and 4. Unfortunately, neither TREE-PUZZLE nor bootstrap analyses were able to resolve any of the inner branches due to the high divergence of the Dkk family. However, a basal position between the Dkk3 and the Dkk 1/2/4 subfamilies was never contradicted. For clarity, we refer to the novel cnidarian Dkks as NvDkk1/2/4 and HyDkk1/2/4 (see Discussion).

hydkk1/2/4 is expressed in endodermal derivatives of the interstitial stem cell lineage
In order to unravel the function of the putative hydkk1/2/4 gene, we analyzed its expression pattern by in situ hybridization. Strikingly, Hydra's head was completely free of hydkk1/2/4-expressing cells, whereas the entire body exhibited a strong endodermal expression with a sharp boundary below the tentacle formation zone (Fig. 3A,B) (see also Hobmayer et al., 1990a;Hobmayer et al., 1990b). In many specimens, we found a graded expression, diminishing towards the peduncle region. This expression pattern is inversely related to that of hywnt3a, hytcf, brachyury and other head-specific genes (Hobmayer et al., 2000;Technau and Bode, 1999).
hydkk1/2/4 is not expressed in endodermal epithelial cells, but is present in gland cells. Gland cells are derivatives of the interstitial stem cell lineage (i-cells), which also gives rise to nerve cells, nematocytes and sex cells. To verify the i-cell nature of hydkk1/2/4expressing gland cells, we investigated the hydkk1/2/4-expression 903 RESEARCH ARTICLE Wnt-Dickkopf antagonism in Hydra Polyps of the temperature-sensitive strain Hydra magnipapillata sf-1 were exposed to heat shock for 3 days, causing the elimination of interstitial cells. Heads, as well as regenerating tips, were isolated from the heatshocked polyps at various times after head removal for mRNA and cDNA synthesis. (B) Cloning strategy. Size-fractionated cDNA was cloned adjacent to a signal peptide-deficient yeast invertase gene for expression in the yeast strain YTK12. Clones expressing a fusion protein with an intact Hydra signal peptide were selected by growth on raffinose plates. Invertase pattern in the temperature-sensitive Hydra magnipapillata strain sf-1. This mutant loses interstitial stem cells after heat shock at 28°C and, over time, all non-dividing derivatives, i.e. nematocytes and neurons, but also gland cells disappear (Sugiyama and Fujisawa, 1978). As expected, hydkk1/2/4-expressing gland cells were lost after a heat shock (Fig. 3C). Because gland cells have a limited proliferating capacity (Schmidt and David, 1986), they are lost more slowly than nematocytes and nerve cells. Thirty days after heat shock, the number of hydkk1/2/4-expressing gland cells diminished by more than 90% to about 50. These residual hydkk1/2/4 cells were full of vesicles (Fig. 3D). We also found a downregulation of hydkk1/2/4 expression in gland cells at sites of oogenesis ( Fig. 3E-G), but not spermatogenesis (Fig. 3H). This is in accordance with their i-cell origin, as oogenesis is accompanied by the differentiation of i-cells into nurse cells.

hydkk1/2/4 expression during budding
As HyDkk1/2/4 is a putative antagonist to the Wnt signalling pathway in Hydra, we analyzed the expression pattern of hydkk1/2/4 in budding polyps. Surprisingly, there were no changes in the expression level of hydkk1/2/4 in the entire bud, neither in early, nor in mid-bud stages (Fig. 4A). Only in late stages, just before the tentacles emerge, was hydkk1/2/4 downregulated at the site of head formation, reflecting the expression pattern in detached polyps. It remains unclear whether this decrease in hydkk1/2/4 signal is due to transcriptional downregulation, the active retreat of gland cells or apoptosis.

hydkk1/2/4 expression during regeneration
During regeneration the transcriptional regulation of hydkk1/2/4 is highly dynamic. In animals bisected at 80% body length, hydkk1/2/4 expression was markedly upregulated at the site of cutting within 30 minutes (Fig. 4). High levels of expression were sustained up to 6 hours after head removal in 50% of all animals (Fig. 4B,C), which is significantly longer than the time required for wound closure (1 hour). Afterwards, hydkk1/2/4 was completely downregulated in the presumptive head (Fig. 4B). With the emergence of tentacles, 30 hours after head removal, the apical region of all regenerates was free of hydkk1/2/4 expression.

RESEARCH ARTICLE
Development 133 (5   The region of stimulated hydkk1/2/4 expression measured roughly 5-10 epithelial cell diameters, which corresponds to 10% of the entire body length (Fig. 4B, 0.5 and 4 hours). ISH on macerated cells revealed that the number of hydkk1/2/4-expressing gland cells increased at the site of regeneration. We determined the ratio of hydkk1/2/4-expressing cells to epithelial cells in uncut control animals and in isolated regenerating tips (correspond to 10 % of the entire body length) 1 hour after head removal. This ratio doubled from 0.33±0.12 in uncut control animals to 0.62±0.06 at the site of regeneration. The expression level in small and large gland cells was also markedly elevated in the regenerating tip when compared with control tissue from the body column (Fig. 4D,E). hydkk1/2/4 expression was also upregulated in foot regeneration, but the number of hydkk1/2/4-expressing cells was not as high as in head regenerates (data not shown). Thus, regeneration induces a dramatic increase in hydkk1/2/4 expression at the site of tissue repair.

hydkk1/2/4 is involved in the injury response
The rapid increase of hydkk1/2/4 message during head regeneration could be either caused by the removal of inhibiting signals emanating from the head or by the injury stimulus itself. We tested both possibilities. (1) Animals were injured by making a deep cut into the body column and leaving the head intact. Injured animals exhibited a strong stimulation of hydkk1/2/4 transcripts for at least 6 hours at the site of injury (Fig. 5A). Even after wound closure, increased levels of hydkk1/2/4 transcripts were still recognizable in cells at the site of the wound. (2) Heads of polyps were removed without wounding by tying a knot with a thin hair around the polyp's subhypostomal region (Newman, 1974). No significant upregulation of hydkk1/2/4 expression occurred in such polyps (Fig. 5B). In agreement with Newman's observations, tied regenerates did not regenerate normally. In conclusion, our results suggest that it is the injury signal itself that causes a stimulation of hydkk1/2/4 expression, and not the removal of an inhibitory signal from the head.

hydkk1/2/4 is involved in the loss-of-regeneration capacity of epithelial Hydra
To assess further the relevance of HyDkk1/2/4 for regeneration, we analyzed the mutant strain sf-1, which loses hydkk1/2/4-expressing gland cells after long-term starvation (Fig. 6). We found that the regeneration capacity of heat-shocked, starved sf-1 animals was strongly correlated to the depletion of hydkk1/2/4-expressing gland cells. The number of hydkk1/2/4-expressing cells was reduced from an average of 270 cells at t 0 to four cells at t 46 (Fig. 6A). The regenerative capacity of these animals was severely reduced: the number of regenerated tentacles dropped from 6.1 at t 0 to 0.4 at t 46 (Fig. 6A). Non-heat-shocked control polyps lost only about 20% of their hydkk1/2/4-expressing cells and continued to regenerate fairly normally (at t 35 , 3.9±1.3 tentacles in sf-1 and 3.9±1.1 in 105 animals). Figure 6 shows representative examples of 35-day-starved animals at 4 hours after head removal (Fig. 6B-D) and at 9 days of regeneration ( Fig. 6E-G). Most regenerating pieces had no hydkk1/2/4 + cells (Fig. 6B), although a few had up to 50 hydkk1/2/4 + cells (Fig. 6C,D). Following regeneration, almost all 'regenerates' lacked hydkk1/2/4 + cells and failed to regenerate properly. Very few pieces regenerated a head (Fig. 6G). Although most of these regenerates lacked hydkk1/2/4 + cells (Fig. 6G), we presume that such successful regenerates had hydkk1/2/4 + cells at the start of regeneration.
We also tested wnt expression in hydkk1/2/4-depleted animals. We found a significant number of ectopic hywnt3a expression domains in such animals ( Fig. 6H-J), but never in non-heat-shocked control animals that were starved for the same time (data not shown). In 47% of all analyzed hydkk1/2/4-depleted sf-1 polyps (n=55), several different-sized patches of hywnt3a-expressing cells were formed along the body column. This suggests that in a hydkk1/2/4-depleted background hywnt3a expression can be spontaneously activated. These are unexpected data, indicating that hydkk1/2/4 gland cells are required for normal regeneration in Hydra and have a function in the regulation of hywnt3a gene expression (see Discussion).
Alsterpaullone-treated polyps (24 hours) formed numerous spot-like hywnt3a-expression domains after 3 days (see Fig. S2 in the supplementary material), followed by ectopic tentacles after 3-4 days (Fig. 7A,B) and ectopic head-like structures after 8 days (Fig.  7A,B). hydkk1/2/4 transcription levels successively decreased 24-48 hours after the onset of alsterpaullone treatment, and were completely absent when ectopic tentacles were formed (Fig. 7B,C). Double ISH indicates that the downregulation of hydkk1/2/4 starts from the centre of the hywnt3a-expression domains in the body column (Fig. S2 in the supplementary material). At later stages, hydkk1/2/4 expression was restored in the tentacle-free tissue between distinct heads (Fig. 7C). Thus, hydkk1/2/4 expression is negatively regulated by canonical Wnt signalling.

HyDkk1/2/4 is functionally active in Xenopus
Overexpression is an instructive experiment to elucidate protein function. Because hydkk1/2/4 overexpression in Hydra was not possible, we have chosen a heterologous approach in Xenopus laevis. As demonstrated by Glinka et al. (Glinka et al., 1998), endogenous xdkk1 overexpression in the embryo leads to enlarged head structures and reduced posterior structures (Dickkopf phenotype) due to enhanced inhibition of late canonical Wnt signalling. Injection of hydkk1/2/4 mRNA in each blastomere of four-cell embryos resulted in severe morphological defects comparable to the Dickkopf phenotype (Fig. 8A): anterior structures, like the cement gland, were enlarged, whereas posterior trunk regions were severely reduced if not abolished. To further demonstrate the Wnt-inhibitory activity of hydkk1/2/4, we co-injected hydkk1/2/4 with xwnt8, which induces secondary axis formation in Xenopus embryos. This secondary axis formation could be blocked by the co-injection of hydkk1/2/4, even though it was less effective than xdkk1 (Fig. 8B). These experiments clearly show that HyDkk1/2/4 is a functional homologue of vertebrate Dkk1 and Dkk4. To determine the inhibiting effect of HyDkk1/2/4 on downstream targets of canonical Wnt signalling, an animal cap assay was performed using siamois, a direct Wnt target gene, as readout (Fig. 8C). Embryos were injected with xwnt8 mRNA in presence or absence of xdkk1 and hydkk1/2/4 mRNA. RT-PCR was then carried out on animal caps to measure the expression of siamois.
Detection of brachyury and histone-4 served as controls for mesodermal contamination and internal normalization, respectively. Non-injected control embryos showed no detectable siamois transcripts. siamois was upregulated upon xwnt8 injection, and this upregulation was abolished by co-injection with xdkk1 or hydkk1/2/4. This result was further verified using the TOPFLASH Wnt reporter (Korinek et al., 1997) (Fig. 8D). In this assay, HyDkk1/2/4 showed a dose-dependent, Wnt-inhibitory activity. These findings strongly indicate that HyDkk1/2/4 is a Dickkopf homologue capable of inhibiting Wnt signalling.

DISCUSSION
The origin of the Dkk proteins in metazoan evolution Dickkopf proteins have been identified as major inhibitors of Wnt signalling in vertebrates. Among the four subfamilies only Dkk1, Dkk2 and Dkk4 are inhibitory ligands of the Lrp5/Lrp6-Wnt-Frizzled complex; Dkk3 does not inhibit Wnt signalling Brott and Sokol, 2002). Here, we describe, for the first time in an invertebrate, the identification of a novel Dickkopf-related protein, which has an inhibitory function in Wnt signalling.  Structurally, the novel protein corresponds to the CRD2 of vertebrate Dkk1, Dkk2 and Dkk4; the similarity is highest to mammalian Dkk1 and Dkk4. There is a second hydkk1/2/4 gene in the Hydra EST database (CN559480; Hydra EST-Kiel) with more than 79% nucleotide identity (probably a product of a recent hydrozoan-specific gene duplication). In Nematostella vectensis there is a dkk1/2/4 gene that contains the N-terminal CRD1. We presume that the CRD1 of this gene is lost in the Hydra hydkk1/2/4. There is also a dkk3 gene in Nematostella, which is related to the nematocyte-specific hydkk3 in Hydra (Fedders et al., 2004).

Wnt-Dickkopf antagonism in Hydra
Based on their conserved CRD2, all cnidarian dkk genes share features of chordate dkk3 and dkk1/2/4 genes. The significant diversification of cnidarian dkk genes suggests a deep split into dkk3 and dkk/1/2/4 gene families during early metazoan evolution. Although this could not be clarified with certainty by phylogenetic analysis because of the high divergence of the sequences, this notion is supported by the chromosomal location of vertebrate Dkks. An analysis of paralogy regions in the human genome using the paralogon database Human 5.28 of Wolfe and McLysaght (Luke et al., 2003;Lundin, 1993;McLysaght et al., 2002) (http://wolfe.gen.tcd.ie/dup/human5.28/) shows that human DKK1, DKK2 and DKK4 are located within related chromosomal regions (DKK1 maps to 10q11, DKK2 to 4q25 and DKK4 to 8p11). These sites lie within the well-characterized 4/5/8/10 paralogy group, which also contains FGF receptors and NK homeobox genes (Birnbaum et al., 2000;Coulier et al., 2000a;Coulier et al., 2000b;Leveugle et al., 2004;Pollard and Holland, 2000), and numerous other duplicated gene families. Genes within this paralogy region were duplicated early in vertebrate evolution (Pollard and Holland, 2000;Luke et al., 2003). Hence, we conclude that dkk1, dkk2 and dkk4 most likely originated by gene duplication. By contrast, human DKK3 maps to 11p15.3, which is not part of the same set of paralogy regions. This finding is consistent with the hypothesis that the features common to the Hydra, Nematostella and vertebrate dkk1/2/4 subgroup were likely to have been present in the common ancestor of cnidarians and bilaterians.

hydkk1/2/4 is an early regeneration-responsive gene
The most obvious role HyDkk1/2/4 plays in Hydra is its function during early regeneration. We found a rapid and dramatic increase of hydkk1/2/4 message at the site of injury within the first hour after head removal. hydkk1/2/4 upregulation was clearly related to the injury stimulus, as it also occurred by simply cutting the animals at any site in the body column. This early upregulation seems to be indispensable for regeneration, because animals whose heads were removed by means of the ligature technique (Newman, 1974), with minimal or even without injury, exhibit no hydkk1/2/4 upregulation and could not regenerate normally. Furthermore, animals that have lost hydkk1/2/4-expressing cells also lose their regeneration capacity.
The regeneration deficient mutant strain reg-16 is also sensitive to the injury stimulus: regenerates develop normally after setting a second cut at the site of injury (Sugiyama and Fujisawa, 1977). We therefore presume that an essential trigger for head regeneration in Hydra is the early release of Dickkopf proteins at the site of cutting. This role of gland cells in patterning processes has certainly been underestimated so far.
Interestingly, Prockop et al. (Prockop et al., 2003) found in the process of mammalian tissue repair that marrow stromal cells (MSCs) from the bone marrow secrete and require Dkk1 for cell expansion in vitro and during the process of tissue repair. It probably interacts with Wnt5a in the growth regulation of MSCs (Gregory et al., 2005;Gregory et al., 2003;Horwitz, 2004;Prockop et al., 2003). Dkk1 has also been found to be strongly upregulated at the sites of apoptosis during vertebrate limb development, and in UV-irradiated tissue (Grotewold and Ruther, 2002a;Grotewold and Ruther, 2002b). Because regeneration in Hydra is accompanied by dramatic changes in the pattern of cell cycle and proliferation at the site of cutting (Holstein et al., 1991;Holstein et al., 2003), we propose that HyDkk1/2/4 in Hydra has a similar function in the response to stress signals and the initiation of tissue repair. It is as yet unclear which signal actually causes the extremely fast upregulation of hydkk1/2/4 transcripts. One molecular trigger could be Jun (also known as c-Jun), a stress-responsive transcription factor and activator of dkk1 that is upregulated during embryonic wounding (Grotewold and Ruther, 2002a;Grotewold and Ruther, 2002b). Another candidate is ␤-Catenin, which can also activate dkk1 expression in a dosedependent manner (Gonzalez-Sancho et al., 2005;Niida et al., 2004). Both genes have been identified in Hydra (Hobmayer et al., 2000).
Our data show that there exists an additional level of complexity in the regulation of hydkk1/2/4 expression. We found during late regeneration, i.e. in late bud stages and after activation of the Wnt/␤catenin pathway by treatment with alsterpaullone, a complete silencing of hydkk1/2/4 expression. This clearly indicates that the early activation of hydkk1/2/4 expression during regeneration is only transient. The shift in the transcriptional regulation of hydkk1/2/4 found during Hydra regeneration shares similarities with the transcriptional downregulation of dkk1 in human colon tumours. Colon tumours exhibit activated Wnt/␤-Catenin signalling and downregulation of dkk1 expression (Gonzalez-Sancho et al., 2005). It was proposed that hypermethylation of the dkk1 promotor leads to dkk1 silencing, similar to the silencing of other Wnt inhibitory genes and to dkk3 Caldwell et al., 2004;Gonzalez-Sancho et al., 2005). Thus, in Hydra, hydkk1/2/4 expression appears to be stably silenced by Wnt/␤-Catenin signalling in the head, while it might suppress Wnt/␤-Catenin signalling in the body column, generating the distinct compartments of Hydra's body, as implied by the sharp boundary underneath the head.

HyDkk1/2/4 is an evolutionary conserved antagonist of Wnt signalling
We tested the function of HyDkk1/2/4 as a putative Wnt inhibitor in Xenopus embryos. In this heterologous system, HyDkk1/2/4 has similar Wnt-inhibitory properties to endogenous XDkk1. (1) HyDkk1/2/4 and the endogenous XDkk1 have the same anteriorizing capacity in Xenopus embryogenesis. (2) HyDkk1/2/4 can block XWnt8-induced secondary axis formation (Fig. 8B). (3) HyDkk1/2/4 blocks the induction of the downstream target gene of the canonical Wnt pathway siamois in animal cap assays. The fact that HyDkk1/2/4, corresponding to the carboxy-terminal cysteinerich domain of Dkks, can exert an inhibitory effect on Wnt signalling is consistent with similar findings for the CRD2 of vertebrates. It was also shown that the CRD2 of Dkk4 is proteolytically cleaved from the full-length protein (Krupnik et al., 1999), which might represent an ancient feature of the basal Dkk4 within vertebrates (Fig. 2C). By comparison, Dkk3 has a linker sequence between CRD1 and CRD2 without a proteolytic cleavage site, and has not been shown to inhibit or modulate Wnt signalling at all (Krupnik et al., 1999;Mao and Niehrs, 2003).
The evolutionary origin of Dickkopf proteins remains unclear, but cnidarians suggest a possible scenario. The CRD2 is similar to colipases, which facilitate the interaction of pancreatic lipases with

RESEARCH ARTICLE
Development 133 (5) lipid micelles (Krupnik et al., 1999). Based on this structural similarity, it was proposed that the CRD2 of Dkk could help other proteins to interact with lipids in order to regulate Wnt function (Aravind and Koonin, 1998). Wnt proteins are indeed tightly associated with the cell surface (Nusse, 2003;Smolich et al., 1993). Wnts are palmitoylated proteins and are therefore much more hydrophobic than is predicted from their primary amino acid sequences (Nusse, 2003;Willert et al., 2003). Thus, lipid binding of Dkk may have initially served to tether Wnt ligands to target membranes. From such a facilitated interaction of Wnt and Dkk at the plasma membrane, the Dkk/Wnt antagonism may have evolved.
In accordance with this idea, we found that HyDkk1/2/4 enhances human LRP6-and mouse Wnt1-induced Wnt signalling in 293T cells 1.5-to 2-fold (see Fig. S3 in the supplementary material), similar to Dkk2 in Xenopus embryos (Brott and Sokol, 2002). The evidence for an inhibitory function in Hydra is less clear. Nevertheless, the expression patterns of hydkk1/2/4 and hywnt3a/hy␤-catenin/hytcf are mutually exclusive and suggestive: hywnt3a and other Wnt genes from Hydra are expressed in the hypostomal region around the mouth of the Hydra (F. Rentzsch, C.G., B.H. and T.W.H., unpublished) (Hobmayer et al., 2000), whereas hydkk1/2/4 is uniformly expressed in the entire body column, but not in the hypostomal region (Fig. 3). This is consistent with the idea that HyDkk1/2/4 is involved in a mechanism that might help to suppress the expression of Wnt genes in the body column. It is also consistent with the fact that, in polyps that were depleted of hydkk1/2/4-expressing cells, hywnt3a is expressed in small cell clusters all over the body column. This patchy upregulation of hywnt3a was never observed in normal polyps except in the budding region. It should be emphasized, however, that this interaction might be more complicated. During budding, hywnt3a-expression spots regularly arise in a tissue that strongly expresses hydkk1/2/4 (Hobmayer et al., 2000). hywnt3a and hydkk1/2/4 are also co-expressed during early regeneration. Thus, although our functional assays in Xenopus provide clear evidence that the HyDkk1/2/4 molecule has the ability to antagonize Wnt signalling, it remains unclear as to what extent HyDkk1/2/4 actually antagonizes Wnt signalling in Hydra.

Evolutionary considerations
Although hydkk1/2/4 is evidently required for the regeneration process in Hydra, it remains to be clarified whether the molecule has an additional role in steady-state animals. At this point, we can only speculate, as no experiments addressing this question have been done. However, the characteristic expression pattern of hydkk1/2/4 and the fact that hydkk1/2/4 can induce the Dkk1 phenotype in Xenopus suggests that hydkk1/2/4 has a similar function in steadystate Hydra polyps as during head and neuronal induction in vertebrates. In vertebrates, Dkk1 can induce secondary heads (Glinka et al., 1997), and is necessary to activate neural genes in the anterior and dorsal region of the vertebrate brain (Glinka et al., 1998;Kazanskaya et al., 2000;Mukhopadhyay et al., 2001;Niehrs, 2004). In Hydra, hydkk1/2/4 is only expressed in the body column. Figure  9 shows schematically that in this region the neuronal differentiation from multipotent interstitial stem cells takes place (Grens et al., 1996;Lindgens et al., 2004;Smith et al., 1999;Technau et al., 1996). We therefore favour the definition of the body column as being the neurogenic region of the Hydra. We presume that the release of HyDkk1/2/4 from gland cells in the endoderm promotes neuronal differentiation by inhibiting the ␤-Catenin/Wnt signalling pathway in the ectoderm of the Hydra body column (Fig. 9A). The BMP antagonist Chordin is also expressed in the same tissue (F. Rentzsch, C.G., B.H. and T.W.H., unpublished), supporting the hypothesis that a primary function of Bmp and Wnt antagonism in metazoan evolution is the facilitation of neuronal differentiation. This hypothesis is supported by the expression patterns of cnidarian Wnt genes (Hobmayer et al., 2000;Kusserow et al., 2005) and by theoretical considerations on midline formation during the evolution of bilaterians (Meinhardt, 2002;Meinhardt, 2004). According to that idea, the Hydra body column is the counterpart to the vertebrate brain, where proneuronal and neuronal genes are expressed, whereas the bilaterian trunk evolved later, and intercalated between the posterior and anterior end.

Conclusion
The vital role of the expression of hydkk1/2/4 in gland cells for head regeneration in Hydra was unexpected. Furthermore, despite the lack of phylogenetic information because of the high divergence of Dkks, the chromosomal location of vertebrate Dkks, the inversely related expression patterns of hydkk1/2/4 and hywnt3a in Hydra, and the functional antagonism of HyDkk1/2/4 and canonical Wnts in Xenopus and Hydra are all consistent with the hypothesis that the Dkk/Wnt antagonism was already present in the last common ancestor of cnidarians and bilaterians. It thus appears that Caenorhabditis and insects have lost the Dkk-Wnt-antagonism, together with other developmental genes.
We thank Charles N. David (Munich) for critically reading the final version of this manuscript. We also thank Rebecca Furlong and Peter Holland (Oxford) for their essential help with the phylogenetic analysis and for providing the chromosome analysis for mammalian Dkks; Uli Technau (Darmstadt, Bergen) for sharing unpublished data and helpful discussions; Anne Lehmkuhl (Darmstadt) and Kirsten Wehner (Darmstadt) for technical assistance; and Hans Meinhardt for critically reading an early version of the manuscript. This work was funded by the DFG.