Injury-induced MAPK activation triggers body axis formation in Hydra by default Wnt signaling

Given the importance of MAPK signaling in regeneration, we next set out to determine how MAPKs are activated in response to injury. We considered ROS because recent work in zebrafish and Drosophila showed the importance of ROS for regeneration and cell proliferation, probably by either activating or modulating different signaling pathways such as JNK or Hedgehog signaling (50, 51). To test whether ROS is produced at the site of injury, animals were incubated with CellRox prior to amputation (SI Appendix, Fig. S4). Indeed, bisected animals incubated with this dye showed a fluorescent signal at the wounded edge, indicating ROS production (SI Appendix, Fig. S4A). Incubation with the oxidizing reagent hydrogen peroxide (H₂O₂) increased the signal intensity at the injury site. Conversely, the signal was abolished when animals were incubated with reduced glutathione (GSH) or bisected without injury by tying (SI Appendix, Fig. S4). To test whether the redox state has an impact on the activation levels of MAPKs, animals were incubated in Hydra medium (HM) or in HM supplemented with either H₂O₂ or GSH to produce an oxidizing or reducing environment, respectively. The exposure to H₂O₂ resulted in elevated activation levels for all MAPKs tested (Fig. 2A). Incubation with reduced GSH resulted in decreased activation of p38 and JNK after wounding, while the effect was less pronounced for pERK (Fig. 2A). These findings were further supported by IF analyses using the same experimental setup. In the presence of H₂O₂, the signal intensity of activated ERK, p38, and JNK was increased and their distribution along the tissue was expanded (Fig. 2B and SI Appendix, Fig. S4C). Treatment with GSH reduced levels of p38 and JNK activation throughout the tissue, while effects on pERK were not evident (Fig. 2B), which is in line with our western blot analysis.

We next tested if manipulating the redox state has an effect on regeneration by incubating bisected Hydra in the presence of either H₂O₂ or GSH (Fig. 2 C and D). Animals regenerated normally under oxidative conditions. By contrast, on average, 65% of Hydra head regenerates and 60% of Hydra foot regenerates did not proceed under reducing conditions (Fig. 2 C and D). These data support a model according to which the activation of MAPKs and the initiation of regeneration are stimulated by ROS production after injury.

Calcium signaling is one of the fastest transduction pathways across cells and is involved in many processes, including wound closure and regeneration (52, 53). Given the fast kinetics of MAPK phosphorylation in response to wounding (Fig. 1B), we hypothesized that Ca²⁺ signaling is responsible for the gradient-like transmission of the injury signal. To test this, we performed live cell imaging of transgenic strains in which polyps express a genetically encoded Ca²⁺ indicator (GCaMP) in ectodermal or endodermal epithelial muscle cells. This allowed us to analyze the Ca²⁺ dynamics in intact and sectioned animals by measuring the fluorescence signal over time (Fig. 3 A and B). When intact animals were stimulated to contract by a short poke with tweezers, but without injury, we found a slight and transient Ca²⁺ release in both germ layers (Movies S1–S9). In contrast, the amplitude of the fluorescence signal was increased threefold in cut animals in both germ layers and at both section sites. Also, the duration of the increase lasted significantly longer in cut animals than in intact animals, with the exception of the oral section site in the endodermal strain. This tendency was, although less pronounced, also observable in polyps cut in calcium-free medium, indicating that the calcium signal is generated in part by intracellular release (SI Appendix, Fig. S5A).

To test whether intracellular Ca²⁺ levels affect MAPK activation, we either increased the intracellular calcium levels by treating polyps with the Ca²⁺ ionophore A23187 or reduced them by incubation with the Ca²⁺ chelator 1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA). SI Appendix, Fig. S5B shows that in bisected animals, exposure to BAPTA resulted in reduced and delayed Ca²⁺ release upon amputation, whereas treatment with A23187 resulted in elevated levels for more than 4 min (Movies S10 and S11). When polyps were exposed to BAPTA after injury, they showed decreased phosphorylation levels of p38, ERK, and JNK in both western blot analysis and IF staining (Fig. 3 C and D). Conversely, treatment with A23187 led to an increase of MAPK activation (Fig. 3C). Interestingly, exposure to A23187 did not necessarily result in enhanced staining signals at the wounded edge but rather resulted in uniform distribution of activated MAPKs along the entire body column (Fig. 3D).

Because MAPK inhibition abrogated the initiation of regeneration, we tested whether reduction of Ca²⁺ levels and thus MAPK phosphorylation also affects the regenerative capacity in Hydra. We found that reduction of Ca²⁺ levels by BAPTA treatment resulted in a decrease of head regeneration capacity to 60% during the first 3 h after wounding and cotreatment with ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid) (EGTA) caused a further reduction to 32% (Fig. 3 E and F). Likewise, regeneration of the foot decreased to 65% in the presence of BAPTA and to 45% when a combination of BAPTA and EGTA was used. Taken together, injury promotes the release of both ROS and Ca²⁺ that in turn acts positively on the activation of MAPK-induced regeneration. We therefore next searched for the putative targets of MAPK activation.

Wnt signaling is an essential pathway for head regeneration and patterning in Hydra (23). Upon decapitation, multiple Wnt ligands are transcriptionally activated in the head regenerate in a temporal sequence (27). However, only little attention has been paid to factors governing foot regeneration. Recent results showed that expression of β-catenin is crucial for both head and foot regeneration—a novelty, as β-catenin had been assumed to act solely on head regeneration (32). We therefore asked whether Wnt signaling is likewise activated in response to injury in a position-independent manner, as observed for MAPKs in this study and as reported for β-catenin elsewhere.

For this, tissue of head and foot regenerating tips was collected at different time points to examine the expression levels of Wnt3, Wnt9/10c, and Wnt7 by qRT-PCR analysis (Fig. 4A). The selection was based on our previous in situ hybridisation and transcriptome studies, which revealed that Hydra Wnt9/10c and Wnt3 were expressed within the first 3 h of head removal, making them the earliest Wnt genes, while Wnt7 expression started much later, at 12 h (27, 37). In line with this, our qRT-PCR analysis showed an up-regulation of Wnt9/10c and Wnt3 after 3 hpa, while the response of Wnt7 was minimal during the first 6 hpa (Fig. 4B). This up-regulation was detectable in both head and foot regenerates, although it was less pronounced in the regenerating foot tissue. While Wnt3 and Wnt9/10c expression was sustained in regenerating head tissue at 6 hpa, it started to significantly drop in foot regenerates, suggesting that positional specification occurs as early as 6 hpa (Fig. 4B and SI Appendix, Table S1A).

We next wanted to clarify whether these diverging kinetics of Wnt expression upon amputation were associated with specific functions in head patterning. To this end, small interfering RNA (siRNA) mediated knockdowns of Wnt3, Wnt9/10c, and Wnt7 were performed in order to examine the role in head and foot regeneration upon bisection (Fig. 4 C–E and SI Appendix, Fig. S6 and Table S1B). The capacity to form a head or foot dropped to about 50% after knockdown of Wnt3-, Wnt9/10c-, and siWnt7-treated animals in both head and foot regenerates. While foot regeneration was completely abrogated independent of the targeted Wnt gene, silencing of Wnt9/10c in head regeneration resembled the phenotype observed upon MAPK inhibition; i.e., animals showed normal wound healing but no signs of regeneration (Fig. 4E). By comparison, siWnt3-treated animals appeared initially to regenerate but then exhibited a peculiar phenotype without mouth and tentacles; instead, they formed a large monotentacle (Fig. 4E). Knockdown of Wnt7 showed a less severe phenotype characterized by mispositioned tentacles along the hypostome. These data clearly indicate that early Wnt genes, particularly Wnt9/10c and Wnt3, are required for the onset of regeneration and head patterning, while later Wnts, such as Wnt7, are more involved in fine-tuned, small-scale patterning. The fact that the phenotype for Wnt3 knockdown is not a complete absence of regeneration, as observed for Wnt9/10c, suggests that Wnt9/10c is required for the onset of regeneration immediately after wound healing, while Wnt3 is involved in the initial specification of head identity. In contrast, expression of Wnt genes in foot regenerates seems to be essential to initiate the onset of regeneration, while later patterning and specification are accomplished by further factors.

Given our results that MAPK activation occurs in a position-independent manner in response to injury, we asked whether ERK, p38, or JNK affects early expression of Wnt3 and Wnt9/10c. To this end, animals were bisected at 50% body length and evaluated by qRT-PCR in controls and upon MAPK inhibition at different time points postamputation. We decided to investigate the expression levels of Wnt3 and Wnt9/10c genes only, since they were most implicated in the onset of regeneration (Fig. 4). Upon inhibition of the p38 or JNK pathways (i.e., the stress-induced MAPK pathways), Wnt3 was strongly up-regulated in a position-independent manner in head and foot regenerates (Fig. 5 A and B, Left and SI Appendix, Table S1 C and D). This up-regulation of Wnt3 was, however, accompanied by strong down-regulation of Wnt9/10c in head and foot regenerates (Fig. 5 A and B, Right). Thus, p38 and JNK have opposing effects on early Wnt9/10c and Wnt3 gene regulation. By comparison, ERK inhibition had only a mild repressive effect on Wnt3 in head regenerates, while foot regenerates were unaffected. Also, there was a discernible difference in the repression of Wnt9/10c in foot and head regenerates, suggesting a different responsiveness of the tissues for ERK-induced Wnt activation (Fig. 5 A and B and SI Appendix, Table S1). Because animals that were inhibited for p38 or JNK do not show any signs of regeneration despite the increased expression levels of Wnt3, we conclude that Wnt9/10c acts as the primary Wnt gene in the transcriptional cascade during regeneration. This is also confirmed by the more pronounced phenotype of the Wnt9/10c knockdown as compared to the other Wnts (Fig. 4 C–E).

To test whether the regeneration deficiency upon MAPK inhibition, as shown in Fig. 1F, is a direct consequence of aberrant Wnt signaling, we repeated the regeneration assay in the presence of MAPK inhibitors and examined whether recombinant Hydra Wnt3 or Wnt9/10c protein is able to restore the regeneration capacity (Fig. 5C and SI Appendix, Fig. S7). Our previous work has shown that treatment of a regeneration-deficient strain with recombinant Wnt3 can rescue regeneration, likely by the activation autocatalytic Wnt activation loop in regenerates (26, 27). In Dimethyl sulfoxide (DMSO) treated control polyps, neither Wnt3 nor Wnt9/10c treatment showed any effect on the progression of regeneration. In ERK- and p38-inhibited polyps, we found an increase of regenerating animals after Wnt treatment, with Wnt9/10c consistently having a stronger effect than Wnt3 (SI Appendix, Fig. S7). This is line with our previous finding that Wnt9/10c is more strongly repressed by p38/ERK inhibition than Wnt3. Neither Wnt3 nor Wnt9/10c was able to restore regeneration upon JNK inhibition (Fig. 5C and SI Appendix, Fig. S7 and Table S1E). This might be due to JNK’s pleiotropic functions in noncanonical Wnt signaling, e.g., during tentacle morphogenesis (17).

The fact that Wnt3 and Wnt9/10c are both required to initiate regeneration raised the question of whether Wnt3 and Wnt9/10c solely act on the onset of regeneration or also fulfill instructive functions in the patterning process, as initially postulated for Wnt3 (23). We tested this assumption by treating gastric pieces with the recombinant Hydra Wnts. To do so, we cut out the middle gastric pieces without head and peduncle and incubated them with recombinant Hydra Wnt3, Wnt9/10c, a 1:1 combination of both, or Bovine serum albumin (BSA) as a control (Fig. 5D). While BSA-treated pieces regenerated normally, 35% of the regenerates that were treated with recombinant Wnt formed heads at both ends of the body column (Fig. 5 E and F and SI Appendix, Table S1F). Of those regenerates, 32% showed ectopic tentacles or double heads. The fact that at the aboral end a head instead of the presumptive foot was formed suggests a change in the axial identity of the regenerating tissue similar to a polarity reversal induced by parabiosis with head tissue in transplantation experiments (54). This effect was stronger with Wnt3 than with Wnt9/10c and strongest with the combination of both (Fig. 5E). The foot was formed centrally and thus most distally in relation to the heads, an effect that likely corresponds to proportion regulation (55–57).

When we blocked β-catenin signaling by (5Z)-5-{[2,5-Dimethyl-1-(3-pyridinyl)-1H-pyrrol-3-yl]methylene}-3-phenyl-1,3-thiazolidine-2,4-dione (iCRT14) treatment at various stages after amputation in order to elucidate early and late Wnt/β-catenin effects of head and foot regenerates, we found strong dependence of late head regeneration on Wnt-dependent transcription and dependence of early head and early foot regeneration (SI Appendix, Fig. S6 B and C). In line with these data, it was also shown that Wnt3, Wnt9/10c, and Wnt7 decrease in expression in foot regenerates by 8 to 12 hpa (31). Thus, late patterning of regenerating feet must be accomplished by factors other than Wnts. These data clearly suggest that Wnt3 and Wnt9/10c possess stage-specific dual functions that comprise the position-independent onset of regeneration and the instructive function for head patterning. The latter was also evident from previous work linking Wnt signaling and oral pattern formation in Hydra (26–28).

Our finding that ERK and the stress-induced MAPK pathways exhibit complementary effects on Wnt expression suggests that these pathways are integrated by positive and negative feedback mechanisms during Hydra regeneration. We therefore tested such putative feedback interactions by using pathway-specific inhibitors. For this purpose, animals were incubated with one of the inhibitors for 1 h and bisected. At different time points after sectioning, the regenerates were then tested for the levels of activated MAPKs by IF and western blot analyses. Our data indicate positive feedback between the stress-activated p38 and JNK pathways and antagonism between stress-activated MAPKs and the ERK pathway, as summarized in Fig. 6A. IF analysis reveals that ERK phosphorylation was inhibited by pretreatment with the MEK/ERK inhibitor; however, pERK was activated in response to treatment with the p38 inhibitor or, to a lesser extent, the JNK inhibitor (SI Appendix, Fig. S8A). The stress-activated MAPKs, i.e., p38 and JNK, showed increased activation levels upon ERK inhibition, whereas a reduction was obtained for pJNK in the presence of the p38 inhibitor and likewise for pp38 upon exposure to the JNK inhibitor. IF-based inhibitor analysis was fully confirmed by the corresponding western blot data (SI Appendix, Fig. S8B).

Strikingly, the mutual antagonism between the p38 and ERK pathways exhibited the strongest regeneration phenotypes upon inhibition (Fig. 1 F and G). Although MAPKs control many cellular processes, one major antagonistic function is their role in cell survival and apoptosis (58). We therefore next analyzed whether there is a link between MAPK activation and the induction of apoptosis. We studied apoptosis by using a standard Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining protocol, thereby avoiding a heating step to increase signal intensity, which in our hands led to staining artifacts (Materials and Methods and Fig. 6B). Apoptotic cells were located close to the amputation site, whereas the remaining tissue was virtually free of apoptotic cells (Fig. 6C). In bisected animals, apoptotic cells were found in tissue regenerating a head and foot, i.e., at both sites of cutting. This finding is consistent with recent work of the Juliano laboratory (31), which used acridine orange staining to detect apoptotic events in regenerates. We detected apoptosis as early as 60 min after amputation, reaching its maximum at 1 to 2 hpa and dropping again by 3 hpa (Fig. 6C). The number of TUNEL-positive cells increased strongly upon inhibition of the JNK pathway but was completely diminished after ERK inhibition, while no significant change was detectable after inhibition of p38. This indicates an early antiapoptotic function of JNK and a proapoptotic function of ERK (Fig. 6D).

The antagonistic role of the stress-inducible MAPKs and ERK in apoptosis is part of a general injury response during Hydra regeneration. This is reminiscent of signal integration by MAPKs in development and cancer formation of vertebrates (58–60). We assume that the antagonism between ERK and JNK/p38 pathways in apoptosis and Wnt signaling has a balancing function in regulating cell fitness and survival to ensure tissue health at the onset of regeneration (Fig. 6A).

Our data show that injuries quickly lead to phosphorylation of MAPKs in ectodermal and endodermal epithelial cells. We were able to demonstrate that the injury signal is composed of Ca²⁺ and the production of ROS, which has not been studied in early Hydra regenerates so far (32). Both signals transmit information about tissue damage intracellularly by activating ERK, JNK, and p38. The activation by ROS is likely achieved by inactivation of counteracting protein tyrosine phosphatases via oxidation of regulatory cysteines, while calcium has been frequently reported to activate MAPK via calmodulin-dependent kinases, Protein kinase C (PKC), and Ras activation (69–72). Although there is increasing evidence for the importance of either ROS or calcium signaling upon injury, it is unclear whether ROS production is Ca²⁺ mediated or whether both messengers act independently in the regenerative context, as was shown for zebrafish fin regeneration (53).

MAPK phosphorylation was among the earliest regeneration responses in our previously shown proteomic/transcriptomic profile of Hydra head regeneration (37). In the present study, we show that the activation of ERK, p38, and JNK is essential for both head and foot regeneration, with ERK showing the most sustained activation kinetics (SI Appendix, Fig. S2). Since MAPK phosphorylation was observed upon both incision and amputation, irrespective of localization, it implicates MAPKs as part of a generic wound response. This term was initially brought up by Reddien and colleagues (11, 73), who showed that both forms of injury promote a position-independent up-regulation of the same set of genes in planarians. The finding was further specified by studies showing a sustained increase in pERK in response to incision and amputation, which was required for Wnt activation (74). In contrast, the role of stress-induced MAPK pathways in regeneration is only poorly investigated. Joint activation of both stress-inducible MAPK pathways was so far only reported for axonal regeneration in Caenorhabditis and imaginal disk regeneration in Drosophila (75, 76). While most studies focus on either ERK activation during regeneration (74, 77–79) or p38/JNK (9, 80), our data indicate that these pathways do not work independently of one another but are based on extensive cross talk. Inhibitor experiments revealed that p38/JNK interacts with ERK as a potential antagonist.

This antagonism of p38/JNK and ERK was also reflected on the level of apoptosis induction. Surprisingly, JNK exhibited antiapoptotic and ERK exhibited proapoptotic properties (Fig. 6). While the two-sided face of JNK in apoptosis has been extensively described in various systems, the proapoptotic role of ERK has only started to raise awareness. It has been suggested that the antiapoptotic role of JNK is based on the phosphorylation of the proapoptotic Bcl-2, which leads to its sequestration and thus inactivation (81). Interestingly, this family has been recently identified in Hydra and demonstrated to execute similar functions (82).

Conversely, although ERK has been frequently associated with antiapoptotic functions, a constantly growing number of studies have demonstrated ERK promotes proapoptotic processes as well (59). This phenomenon was observed in particular after ROS release, which leads to sustained ERK activation (83), as was also observed in this study. A sharp decrease in the number of apoptotic cells was also described by Chera et al. (40) for Hydra head regeneration. However, our data differ in that ERK-mediated apoptosis is not restricted to head regenerates but also occurs in foot regenerates (Fig. 6). The fact that ERK-mediated apoptosis occurs in response to injury, regardless of the amputated axial position, also challenges the earlier report by Chera et al. (29), in which apoptotic cells were described to occur exclusively at the oral-facing wound sites. We show that applying a standard protocol that omits a high-temperature incubation step, as used by Chera et al. (29), is not prone to yield unspecific signals, as shown for fixed samples (Fig. 4B). Our findings are also in full accordance with recent results by Cazet et al. (31), showing that the response to amputation is identical at both wounds and includes widespread apoptosis. Thus ERK-mediated apoptosis must be considered as an essential part of the generic wound response in Hydra.

Our previous data have revealed not only unexpected complexity between the early proteomic responses and the first transcriptional activation of the patterning GRNs (37) but also dominance of Wnt signaling in the formation of the early Hydra head organizer (21, 23, 24, 27, 37). It was therefore striking that two of the earliest and one of the latest Wnt genes (27) were activated in both head and foot regenerates. This is consistent with recent findings in Hydra (31, 32) indicating position-independent activation of Wnt signaling during regeneration, and it challenges the concept of an initial asymmetry of canonical Wnt signaling (29). As in planarians (11), it also raises the general questions of how the positional information is established and what function early Wnt genes have.

The fact that MAPKs are essential for Hydra regeneration (Fig. 1) and the discovery in planarians that Wnt1 expression decreases with attenuated ERK activity (74) led us to systematically investigate the relationship between MAPKs and Wnts. When we tested the influence of ERK, p38, or JNK activation on the expression of the early regeneration genes Wnt3 and Wnt910c, we found strong evidence that early Wnt expression is highly dependent on MAPK activation and largely independent of the axial position, i.e., the regeneration of a head or foot (Fig. 5 A and B). Consequently, MAPKs and early Wnts are part of a generic wound response that only becomes head organizer specific at about 6 hpa (Fig. 5). This is also supported by recent RNA sequencing (RNA-seq) data demonstrating that early gene expression is indistinguishable in presumptive head and foot tissue and only becomes position specific at 8 hpa (31). It also coincides with less pronounced effects of MAPK inhibition on Wnt expression at later stages (SI Appendix, Table S1 and Fig. 5 A and B).

Because prominent downstream effectors of MAPK signaling are transcription factors (TFs), we performed an in silico promotor analysis of Wnt3, Wnt9/10c, and Wnt7 for TF binding sites associated with MAPK signaling (SI Appendix, Fig. S9A). All Wnt promotors showed high probability scores for the bZIP family of TFs such as cFos and different other CREB/ATF members (SI Appendix, Fig. S9A and Table S2). While cFos is a downstream target of the ERK pathway, ATFs can be activated by the JNK or p38 pathway (84–86). This hypothesis is further supported by recent assays for transposase accesible chromatin sequencing (ATAC-seq) studies in Hydra demonstrating that early stages of injury-induced Wnt expression likely depend on bZIP TFs such as cFos and ATFs (31). In silico analysis revealed that ATF2/CREB can bind mainly to the Wnt9/10c promotor and ATF3 can bind mainly to the Wnt3 and Wnt7 promoters (SI Appendix, Table S2). ATF3 is activated in response to stress stimuli (87–90) but has not been described in cnidarians (SI Appendix, Fig. S10) (91). In electromobility shift assays (EMSAs), we found that recombinant human ATF3 binds to the putative ATF3 binding sites of the HyWnt3 promoter (SI Appendix, Fig. S9B). Although we cannot rule out the possibility that ATF2/CREB also binds to the putative ATF3 sites, differential binding of ATF2/CREB and ATF3 to the promoter region of Wnt9/10c and Wnt3 could explain the antagonistic effects of MAPKs on Wnt expression. We therefore propose a model in which differential binding of ATFs promotes the expression of Wnt9/10c to initiate regeneration and prevent premature up-regulation of patterning Wnts such as Wnt3 and Wnt7. To our knowledge, there are currently only two reports of ATF2-driven expression of Wnts, i.e., in colon cancer cells (92) and in hematopoietic tumor cells (93), and none on ATF3. For us, they represent a level of Wnt regulation that has received little attention but is important for integrating Wnts in the generic wound response.

While the differential activation of Wnts by ATFs is still part of the injury-induced cascade, the question remains how the transition from a generic wound response to position-specific gene expression is achieved on the cellular level. Previous work in Hydra (23, 94, 95) and Nematostella (22, 96), as well as related studies in planarians (65, 97, 98), has convincingly shown that β-catenin is crucial for the overall polarity of the body (99). We thus assume that β-catenin is intimately linked to the long-term storage of the body axis gradient, as already postulated (100, 101). This gradient is known as head-forming competence and shown to be different at various axial levels along the oral-aboral body axis in Hydra (95). It was introduced as the source density in the mathematical model by Gierer and Meinhardt (102) and is proposed to be essential for establishing the previous positional identity. Since the source density seems to encode a distinct responsiveness/competence along the body axis, we hypothesize that the specification of head and foot tissue occurs by the interplay of the injury signal with the source density.

We tested the role of injury-induced self-amplification of Wnt/β-catenin signaling in the context of bud formation. Since this process requires high levels of β-catenin (94), we placed an incision in the budding region at a fluorescently labeled side to examine whether we are able to artificially induce budding (Fig. 7A). Compared to a noncut control, the frequency of bud development at the site of incision increased from 20 to 60% (Fig. 7A and SI Appendix, Table S1G). This is in line with recent findings by Cazet et al. (31), who were able to induce ectopic head organizers by puncture wounds.

We also attempted to override a presumed existing positional gradient of the source density by incubating gastric pieces with the GSK3 inhibitor azakenpaullone, which generates high levels of β-catenin throughout the whole tissue (28, 33, 103). Strikingly, these regenerates formed heads at both ends of the body column to about 85% (Fig. 7B), which is twice as high compared to gastric tissue treated with recombinant Wnts (Fig. 5 E and F). This is strikingly similar to alsterpaullone- or azakenpaullone-treated embryos of the sea anemone Nematostella, which also form polyps with a head at both ends of the body axis (22), an approach that has also recently been used to identify β-catenin downstream targets patterning the oral-aboral axis in these animals (96).

These experiments demonstrate that Wnt/β-catenin signaling is intimately linked to the level of the source density, which in turn determines the fate of the regenerating Hydra tissue as initially shown for planarian regeneration (Fig. 7C). Recently proposed mathematical models support the importance of an initial injury signal and β-catenin for robust patterning in Hydra and animal regeneration (104–106). Given that injury-induced Wnt expression has already been described for the regeneration of fish fins, amphibian limbs, and mammalian intestine and heart, this activation loop seems to be a common strategy across metazoan phyla to overcome tissue damage in both regeneration-competent and regeneration-noncompetent organisms (107). We thus conclude that the injury response is essential to promote both wound healing and regeneration, while the decision to regenerate or simply heal the tissue is determined by its competence. The nature of the source density, which encodes this competence, is still elusive, but our data stress the importance of identifying the molecular cues that constitute the source density and determine whether wound healing or regenerative pathways are activated upon injury.

Polyps were kept in HM (1 mM NaHCO₃, 0.1 mM KCl, 0.1 mM MgCl₂, 1 mM CaCl₂, and 1 mM Tris, pH 6.9) at 18 °C. Animals were fed three times a week with Artemia salina nauplii. The medium was exchanged daily. The strain Hydra vulgaris AEP was used for all experiments if not otherwise indicated. Animals were starved for 24 h prior to experiments.

Animals were pretreated with inhibitors for 1 h prior to the start of the respective experiment. 25 µM U0126 (MEK/ERK inhibition), 50 µM SB 203580 (p38 inhibition) (SI Appendix, Fig. S11), and 1 µM SP600125 (JNK inhibition) (21) were diluted in HM. All reagents were purchased from Cell Signaling Technology. All inhibitors were dissolved in DMSO as 20 mM stock solutions. For calcium experiments, 2 µM A23187 (Sigma-Aldrich) in HM was used if not otherwise indicated. 50 µM BAPTA and 0.5 mM EGTA were used in calcium-free HM. Both reagents were purchased from Sigma-Aldrich. BAPTA and A23187 were dissolved in DMSO as 50 mM stock solutions. For redox experiments, 100 µM freshly purchased GSH (Merck Millipore) and 500 µM H₂O₂ (Sigma-Aldrich) were diluted in HM. As shown before for U0126 (42), the effects of inhibitors are dose dependent and require activity control for long-term treatment of head and foot regenerates. Inhibitors and GSH were exchanged every 12 h. For reagents dissolved in DMSO, the solvent only served as control by diluting an equal volume of DMSO in HM. For multiple inhibitor experiments, the highest volume of the inhibitor solution was used as reference for the control.

Full details of methods are given in SI Appendix, SI Material and Methods.