In the freshwater hydra, the gastrodermis is made up of gland cells that produce zymogens and express Kazal1, a serine protein-kinase inhibitor Kazal-type (SPINK) gene, as well as endodermal epithelial cells that behave as digestive cells. The Kazal1 function was tested in hydra fed with bacteria expressing dsRNAs. Those showed a progressive decrease in Kazal1(–) transcript abundance, decrease that affected homeostatic conditions as evidenced by the low budding rate and the induced animal death, already detected after three feedings. Concomitantly, a dramatic disorganization of gland cells followed by their massive death was observed, while their neighbouring cells, the digestive cells, displayed a highly vacuolated cytoplasm. Those vacuoles were assigned as autophagosomes as they contained mitochondria and late endosomes.

Amputation of wild-type (wt) hydra leads to an enhanced Kazal1 expression in regenerating tips (Fig. 2, left panel). However, in Kazal1 knocked-down hydra (Fig. 2, right panel), autophagosomes were immediately detected in endodermal cells of the regenerating tips, and upon complete silencing, hydra no longer survived the amputation stress [11]. This first cellular phenotype resulting from a gene knockdown in cnidarians highlights the essential digestive and cytoprotective functions played by the Kazal1 serine-protease inhibitor activity in hydra. In mammals, autophagy of exocrine pancreatic cells is also induced upon SPINK1/SPINK3 inactivation, whereas SPINK3 expression is activated in injured pancreatic cells [13–15]. Hence, SPINKs, by preventing an excessive autophagy, appear as key players of the stress-induced self-preservation program. Enhancing the self-preservation program in injured tissues might therefore be the condition for unmasking their potential cell and/or developmental plasticity [16].

Following a distinct strategy, we used DNA-binding assays to screen for transcription factors that would exhibit modulations in their DNA-binding complexes after amputation. That way, we identified and cloned the cAMP Response Element Binding (CREB) protein, which shows three highly conserved domains: the kinase inducible domain (KID), the DNA-binding and dimerization leucine-zipper domains (bZIP) [17]. In vertebrates, the CREB transcription factor mediates the response to a large array of extra-cellular signals to the nucleus through post-translational modifications that involve multiple protein kinases [18], including the RSK kinase [19]. This kinase phosphorylates a particular residue, Ser133, located in the KID, an event that is critical for modulating CREB transcriptional activity, namely because the phosphorylated form of CREB specifically binds to the ubiquitous and multifunctional transcriptional co-activator CBP [20].

In hydra CREB, the Ser67 residue located in KID is a target for post-translational regulation, similarly to the Ser133 residue characterized in the CREB vertebrate proteins. By using the anti-phosphoSer133-CREB antibody together with the antihyCREB antibody, we noted that phosphoCREB-expressing nuclei were restricted to the endodermal layer, while CREB-expressing nuclei were distributed in both layers. Interestingly, immediately after amputation, the number of phosphoCREB-expressing nuclei increased significantly in the head-regenerating tips (Fig. 3A). Biochemical and immunological evidences identified a RSK-like kinase, which showed an enhanced activity and a hyperphosphorylated status during head but not foot regeneration [21]. Exposure to the U0126 MEK inhibitor, which prevents RSK phosphorylation, inhibited head but not foot regeneration (Fig. 3B), while in head regenerating tips, CREB phosphorylation was abolished and the early gene HyBra1 was not activated. These data support a role for the MAPK/RSK/CREB pathway in one specific cell lineage, the endodermal myoepithelial cells, likely linked to the reactivation of the developmental program leading to head regeneration [21].

The CREB transcription factor and the RSK kinase are indeed co-expressed in all three hydra-cell lineages including dividing interstitial stem cells, proliferating nematoblasts, proliferating spermatogonia and spermatocytes, differentiating and mature neurons as well as ectodermal and endodermal myoepithelial cells [22]. In addition, CREB gene expression is specifically up-regulated during early regeneration and early budding. Thus, in hydra, the CREB pathway appears already involved in multiple tasks, such as reactivation of developmental programs in an adult context, self-renewal of stem cells, proliferation of progenitors and neurogenesis. More recently, the CREB-Binding protein (CBP) gene was identified, shown to be ubiquitously expressed (SC, unpublished), and is currently tested together with CREB and RSK in RNAi experiments.

According to several independent datasets, neurons are supposed to play a minor role in de novo head patterning; for instance, nerve-free hydra mutants can regenerate their heads [23,24]. We recently readdressed this question in wt hydra by testing the function of the ParaHox Gsx-homolog gene, cnox-2. Cnox-2 expression is restricted to fast-cycling interstitial cells that give rise exclusively to sensory mechano-receptor cells (nematocytes) in the body column and apical multipolar neurons. Therefore, cnox-2 is a marker for a subset of interstitial cells that corresponds to bipotent neuronal progenitors [12]. Upon partial cnox-2 silencing, the apical nervous system (ANS), which can be visualized with the anti-β-tubulin antibody, appears reduced and disorganized; when silencing is complete, apical neurons are no longer detected and the body size is drastically reduced (Fig. 4).

During regeneration, a massive de novo neurogenesis was observed in the head-regenerating tip, starting 24 h post-amputation. The cnox-2 expressing cells, which start to appear in the presumptive head region at the same time, correspond to dividing neuronal progenitors and differentiating neurons. As anticipated, cnox-2 RNAi knockdown alters de novo apical neurogenesis and delays significantly head formation. Similarly, in the sf-1 nerve-free temperature sensitive mutants [25], cnox-2 expression is abolished at restrictive temperature and head regeneration is far less efficient, as 50% of the animals remain unable to regenerate their heads after six days. These results indicate that de novo head patterning in wt hydra polyps depends on cnox-2 promoted neurogenesis [12]. Alternatively, when neurogenesis is missing, a slower and less efficient head developmental program is possibly activated [2].

What is the role of stem cells in the cellular remodelling underlying regeneration? The classical views of regeneration imply a clear distinction between morphallactic regeneration, occurring in the absence of any cell proliferation, and epimorphic regeneration, relying on the formation of a proliferating blastema [26,27]. The first type would concern mostly hydra, and partially planarians [28], whereas the second would correspond to vertebrate regeneration. The recent results obtained in our laboratory suggest that the endodermal myoepithelial cells of the tip undergo a transient phenotypic transition, i.e. they loose their epithelial polarity, while the interstitial cells located immediately underneath re-enter the cell cycle (SC, unpublished). The combination of these two cellular events is highly reminiscent of the blastema formation in urodele regenerating limbs [29] or zebrafish regenerating fins [30]. If confirmed, this would suggest that hydra regeneration in wt conditions might be closer to epimorphic regeneration than anticipated, sharing some regenerative mechanisms at work in amputated urodeles limb or zebrafish fin.

One key aspect in the control of the blastema growth is the presence of neurotrophic factors that play an essential function in the urodele limb [31]. Nerve-free hydra obtained either chemically or genetically provide a useful context to test the nerve dependence. Indeed, those nerve-free hydra can regenerate their head, although with a much weaker efficacy and at a slower pace. At the early phase of head regeneration, the proliferating zone in wt hydra is located in an area that is neuronal rich (SC, unpublished). Therefore, the putative neurotrophic function of these neurons could be tested by comparing the cellular remodelling and the expression profiling in wild-type and nerve-free contexts. At the early–late stage, a de novo neurogenesis was observed at the regenerating tip, which is the site of an intense concomitant cell proliferation. Again, a functional link between these two cellular events is not established. Nevertheless, in nerve-free hydra, an intense proliferation of the myoepithelial cells was also observed at the early–late stage [32]. Are the signals that trigger this cell proliferation in both contexts identical? Are the nerve-free hydra turning on an alternative pathway? Is there any analogy between the nerve-free hydra and the urodele aneurogenic limb [33], i.e. a limb that develops in the absence of any neuronal support and later on does not require any neurotrophic factors for its regeneration? Thanks to the functional tools now available in the hydra model system, these questions should be reconsidered in qualitative and quantitative terms at both cellular and genetic levels.

In the coming years, functional studies will tell us how similar are the molecular and cellular processes that drive regeneration in hydra, planarians, annelids, urodeles, and zebrafish. This aspect of regeneration is currently a dark box. To achieve a full understanding of the regenerative potential of adult tissues, their respective regenerative programs have to be elucidated. Nevertheless, more importantly, it will be necessary to understand the permissive context(s) that keeps accessible such developmental programs that de novo give rise to appropriate form and function. Then, a comparative view at the level of gene regulation will be a first step towards the identification of the key regulators of regeneration. The hydra model system provides a unique entry point to identify what does animal regeneration indeed mean, i.e. the capacity to use a minimal number of cellular processes that, given the cell types available, will reactivate a developmental program, which ultimately leads to the reestablishment of the missing part, identical to the amputated one. Surprisingly, the CREB pathway, which was identified as a key signalling pathway for consolidating long-term memory in aplysia, drosophila, and mammals [34], appears as an essential player in the reactivation of the developmental program in hydra. Although we anticipate from our preliminary analyses that this reactivation process in hydra is likely not unique, but rather multiple, it is tempting to speculate that regeneration might reflect the potential for long-term memory mechanisms of developmental processes.