egr-4, a target of EGFR signaling, is required for the formation of the brain primordia and head regeneration in planarians

After almost any type of amputation, freshwater planarians are capable of regenerating missing regions with the appropriate polarity. Once polarity is established, the appropriate morphogenetic and patterning programs must direct the differentiation of those blastemas into either anterior or posterior regions and their corresponding tissues and organs. Thus, cephalic ganglia develop de novo in decapitated planarians. The instrumental role of Wnt/β-catenin signaling in defining head versus tail identity during planarian regeneration is well documented (Gurley et al., 2008; Iglesias et al., 2008; Petersen and Reddien, 2008). However, less is known about the genes and/or pathways that mediate the patterning, growth and differentiation of blastemas after the re-establishment of polarity (Felix and Aboobaker, 2010).

Conserved signaling pathways play important roles in morphogenesis in all animals. One such pathway is the epidermal growth factor receptor (EGFR) pathway, which regulates multiple biological processes, including cell proliferation, differentiation, apoptosis and cell survival. We have previously shown that silencing of Smed-egfr-3, a planarian homologue of epidermal growth factor receptors, impairs regeneration and blastema growth in these organisms, probably by disrupting cellular differentiation (Fraguas et al., 2011). Similar effects have been reported after inhibition of ERK (MAPK extracellular signal-related kinase; Tasaki et al., 2011), a gene whose expression is regulated by EGFR signaling in many organisms. In the present study, we conducted digital gene expression (DGE) analyses to identify putative target genes of Smed-egfr-3 in order to better characterize the function of the EGFR signaling pathway during planarian regeneration. One of the isolated candidate genes that was downregulated after egfr-3 RNAi was Smed-egr-4 (Wenemoser et al., 2012), henceforth egr-4, a member of the early growth response (egr) gene family. egr genes were first characterized by their induction by nerve growth factor (Milbrandt, 1987) and other mitogens (Sukhatme et al., 1987), and are implicated in the regulation of multiple cellular processes (Calogero et al., 2004; Cole et al., 1989; Dussmann et al., 2011; Shafarenko et al., 2005; Sukhatme et al., 1988). Members of this family of zinc-finger transcription factors are considered immediate-early genes; they are rapidly induced by many environmental signals, including growth factors, hormones and neurotransmitters (Thiel and Cibelli, 2002), and are rapidly and transiently upregulated by a variety of signaling pathways, including the EGFR and MAPK/ERK signaling pathways (Aggeli et al., 2010; Cabodi et al., 2008; Kaufmann and Thiel, 2001; Ludwig et al., 2011b; Mayer et al., 2009; Tsai et al., 2000). EGFR signaling in mice stimulates the expression of egr2 to promote cell proliferation during bone formation (Chandra et al., 2013), while EGF induces Egr-1 expression in endothelial cells (Tsai et al., 2000).

We found that egr-4 was mainly expressed in the central nervous system (CNS) and was rapidly induced after different types of injury. Although early egr-4 expression after injury was independent of EGFR signaling, it became Smed-egfr-3 dependent from the second day of regeneration. Functional analyses based on RNA interference (RNAi) revealed that egr-4 was required for head regeneration but not for the regeneration of posterior regions. egr-4 silencing significantly impaired the formation of anterior blastemas; these animals exhibited either small mispatterned cephalic ganglia or a total absence of new brain tissue. The differentiation of other cell types normally present in the anterior region was also altered in egr-4(RNAi) animals. However, the early expression of polarity determinants required for the re-establishment of anterior polarity was unaffected. Simultaneous silencing of egr-4 and different elements of the Wnt/β-catenin pathway revealed that egr-4 is required for the differentiation of the brain primordia. Moreover, the results of these experiments suggest that egr-4 promotes head regeneration by antagonizing the inhibitory action of the Wnt/β-catenin pathway. Taken together, these findings identify egr-4 as one of the first known genes necessary for the differentiation of the brain primordia during planarian regeneration. In view of these findings, we discuss how the failure to differentiate proper brain primordia may lead to the blockade of blastema growth and head regeneration.

We have previously demonstrated that silencing of Smed-egfr-3, a planarian homologue of the epidermal growth factor receptor (EGFR) family, blocks regeneration, probably by disrupting cellular differentiation (Fraguas et al., 2011). To better characterize the functions of the EGFR pathway during planarian regeneration, we constructed and sequenced DGE (Digital Gene Expression) libraries to compare the transcriptomic profiles of control and egfr-3(RNAi) regenerating animals (supplementary material Table S1). Genes predicted to undergo up- or downregulation after Smed-egfr-3 RNAi on days 1 and 3 post-amputation are shown in supplementary material Table S2. After applying the cut-off criteria, 680 genes were upregulated on day 1 and 2403 on day 3 in Smed-egfr-3(RNAi) animals. A total of 655 genes were downregulated on day 1 and 800 on day 3 in Smed-egfr-3(RNAi) animals (supplementary material Fig. S1). All genes that were up- and downregulated on day 3 post-amputation (2949 and 921, respectively) were annotated and assigned Gene Ontology (GO) categories (supplementary material Fig. S2). Because EGFR signaling regulates multiple downstream targets and pathways, the annotated genes fell into a wide range of GO categories. Overrepresented categories included ‘signal transduction’, ‘cell differentiation’, ‘response to stress’, ‘catabolic process’, ‘nucleotide binding’ and ‘protein kinase activity’ (supplementary material Fig. S2). To identify putative target genes of Smed-egfr-3 that may be linked to the phenotypic defects observed after its silencing (Fraguas et al., 2011), we selected a group of genes for further characterization, based either on their proposed involvement in the EGFR pathway in other models or their role in processes such as neurogenesis, cell proliferation and differentiation, cancer and tumorigenesis, apoptosis and cell survival, and inflammation and the immune response (supplementary material Fig. S3).

Among the candidate targets that exhibited interesting RNAi phenotypes and merited further characterization was a gene displaying sequence similarity to the early growth response (egr) gene family of transcription factors (Contig_2669_90e), a well-known family of genes that are regulated by EGFR signaling in other models (Cabodi et al., 2008; Kaufmann and Thiel, 2001; Mayer et al., 2009; Tsai et al., 2000). Although several members of this family have been described in planarians (Onal et al., 2012; Sandmann et al., 2011; Wenemoser et al., 2012; Wagner et al., 2012), Contig_2669_90e was the only egr gene downregulated after Smed-egfr-3 RNAi. Sequence analyses revealed that this gene corresponded to the previously annotated egr-4 (Wenemoser et al., 2012) and phylogenetic analyses confirmed that egr-4 belongs to this gene family (supplementary material Fig. S4).

Whole-mount in situ hybridization in intact planarians revealed egr-4 expression mainly in the cephalic ganglia, with weak expression in the ventral nerve cord (vnc) and the mesenchyme (Fig. 1A). We have previously reported Smed-egfr-3 expression in neoblasts and in the CNS (Fraguas et al., 2011). Here, we better characterized the expression of Smed-egfr-3 by fluorescent in situ hybridization (FISH) (Fig. 1B). Although we did not succeeded in performing double FISH of egr-4 and Smed-egfr-3, their similar expression patterns (compare Fig. 1A with 1B) suggest that these genes are probably co-expressed in the CNS. Moreover, egr-4 expression in the cephalic ganglia was strongly reduced after Smed-egfr-3 silencing, further supporting the DGE data (Fig. 1C). To investigate whether egr-4 expression in the mesenchyme corresponded to neoblasts we performed in situ hybridization on irradiated animals at different time points (supplementary material Fig. S5). Whereas Smed-egfr-3 expression in the mesenchyme disappears 1 day after irradiation (Fraguas et al., 2011), no change in the pattern of egr-4 expression was observed, even at 7 days after neoblast elimination (supplementary material Fig. S5), suggesting an absence of egr-4 expression in stem cells. During anterior regeneration, egr-4 was upregulated in the wound region from day 1 of regeneration (Fig. 1D). As regeneration progressed, egr-4 expression became restricted mainly to the differentiating cephalic ganglia (arrowheads in Fig. 1D). By contrast, no upregulation of egr-4 was observed during posterior regeneration at later time points (Fig. 1D). However, strong expression at earlier stages (1 h and 3 h post-amputation) was observed in both anterior and posterior blastemas. egr-4 has been recently categorized as a W1 (wound-induced class 1) gene (Wenemoser et al., 2012). In situ hybridization revealed rapid upregulation of egr-4 after a small incision (Fig. 1E), in agreement with previous reports for other planarian egr genes (Wenemoser et al., 2012; Sandmann et al., 2011). This strong upregulation of egr-4 was not dependent on neoblasts (Fig. 1E′) or on Smed-egfr-3 (Fig. 1E″), as no differences were observed in irradiated or Smed-egfr-3(RNAi) animals.

Finally, to better characterize the dependence of egr-4 expression on Smed-egfr-3 during regeneration we carried out a detailed time-course analysis of egr-4 expression after Smed-egfr-3(RNAi) (Fig. 1F). Normal egr-4 expression in the wound region was observed up to day 1 of regeneration. However, by day 2, egr-4 expression was downregulated (Fig. 1F), suggesting that egr-4 expression becomes Smed-egfr-3 dependent once its expression becomes restricted to the regenerating cephalic ganglia.

We conducted RNAi experiments to characterize the role of egr-4 in regeneration (see Materials and Methods). Both trunk and tail fragments regenerating a new head exhibited blastemas that were either very small or abnormally differentiated. After 8 days of regeneration, about half of the anteriorly regenerating trunks (n=18/34) formed normal-sized blastemas but with cyclopic eyes, and very few (n=3/34) regenerated small blastemas without eyes (Fig. 2A). By contrast, at the same time point, most tails (n=30/34) failed to properly regenerate a head. Of these 30 tails, some developed very small blastemas (n=12/30), whereas others appeared to only heal the wound (n=18/30) (Fig. 2A). These results indicate that the severity of egr-4 RNAi varied along the anterior-posterior (AP) axis. Posterior regeneration proceeded normally (Fig. 2A), indicating that egr-4 RNAi specifically blocked anterior regeneration.

Although egr-4 was not expressed in neoblasts, we investigated whether these defects in regeneration were due to a non-cell-autonomous effect of egr-4 RNAi on neoblast maintenance and/or proliferation. However, no differences in neoblast number or distribution were observed after immunostaining and in situ hybridization for the neoblast-specific markers anti-SMEDWI-1 antibody and Smed-histone-2B, respectively (Fig. 2B) (Reddien et al., 2005; Guo et al., 2006; Solana et al., 2012). Thus, neoblasts were normally found below the wound epithelium on day 3. Eight days after amputation, few Smed-histone-2B-positive cells were detected in the new head region of controls. Similarly, despite impaired blastema growth and regeneration, few Smed-histone-2B-expressing cells were observed after egr-4(RNAi), indicating that neoblast dynamics (in terms of distribution) was unaffected.

We next used an anti-phospho-histone H3 antibody to characterize the mitotic response after egr-4(RNAi) in both anterior and posterior regenerating fragments (Fig. 2C). In most cases, the first mitotic peak, which is associated with injury and occurs 6 h post-amputation, was significantly attenuated when compared with controls. Interestingly, the second mitotic peak at 48 h post-amputation, which is associated with tissue loss (Wenemoser and Reddien, 2010), was significantly attenuated in anterior, but not posterior, blastemas. However, despite this decrease in the rate of proliferation, a significant number of mitoses were observed at all stages after egr-4 RNAi (supplementary material Fig. S6), suggesting that the severe impairment of blastema growth and regeneration in these animals was not exclusively due to defective mitosis. Finally, we characterized neoblast progeny at different stages of differentiation using the lineage markers Smed-NB21.11e and Smed-AGAT-1, which are specific to early and late neoblast postmitotic progeny, respectively (Eisenhoffer et al., 2008). Although the early neoblast progeny were unaffected, the number of late progeny was significantly decreased after egr-4 RNAi (Fig. 2D). Taken together, these results indicate that egr-4 RNAi impairs anterior regeneration probably by affecting cell differentiation rather than neoblast pool maintenance or proliferation.

To further characterize these defects in regeneration, we first used the pan-neural marker anti-SYNORF-1 (Cebrià, 2008) to study CNS regeneration. Both control and egr-4(RNAi) head fragments regenerated normal pharynges and ventral nerve cords grew into the newly developed tails (Fig. 3A). By contrast, in anteriorly regenerating trunks, egr-4 RNAi generally resulted in the formation of blastemas with truncated ventral nerve cords and little cephalic ganglia differentiation, whereas normal cephalic ganglia regeneration was observed in controls (Fig. 3A). We next used additional neural (Fig. 3B) and anterior (Fig. 3C) markers to better characterize the egr-4(RNAi) phenotype. After 5-7 days of regeneration, most egr-4(RNAi) trunk fragments displayed significantly smaller or near-absent cephalic ganglia after in situ hybridization for Smed-gpas, a gene specific to the brain lateral branches (Iglesias et al., 2011). Moreover, the expression of markers of distinct neuronal populations such as Smed-tbh (octopaminergic neurons; Fraguas et al., 2012; Nishimura et al., 2008) and Smed-th (dopaminergic neurons; Fraguas et al., 2011; Nishimura et al., 2007) was significantly reduced (Fig. 3B). These markers revealed that egr-4 (RNAi) animals developed small and aberrant cephalic ganglia. To study the patterning of these cephalic ganglia, we analyzed expression of the otd/Otx family gene Smed-otxA and the homeobox-containing gene orthopediaSmed-otp (Umesono et al., 1997,, 1999; Iglesias et al., 2011) (Fig. 3B). otp is normally expressed in the most lateral region of the cephalic ganglia, whereas otxA is expressed more medially (Umesono et al., 1999; Iglesias et al., 2011). Although egr-4(RNAi) animals often displayed abnormal cephalic ganglia, these distinct domains along the mediolateral axis were clearly distinguishable (Fig. 3B). Taken together, our data indicate that egr-4 RNAi results in the development of small but well-patterned cephalic ganglia.

In addition to the aforementioned CNS defects, egr-4(RNAi) animals failed to properly regenerate the anterior gut branch into the blastema (Fig. 3C). Moreover, the expression of the anterior marker Smed-ndl-4, a FGF-receptor-like protein of the nou-darake family (Cebrià et al., 2002a; Rink et al., 2009), was either significantly reduced or completely absent in the majority of egr-4(RNAi) animals (Fig. 3C). The expression of Smed-sFRP-1, another anterior marker (Gurley et al., 2008; Petersen and Reddien, 2008), was not as strongly affected. Most egr-4(RNAi) animals (n=8/14) exhibited normal Smed-sFRP-1 expression; some (n=4/14) showed a very slight decrease (probably due to the smaller size of the regenerated anterior region) and the marker was completely absent in only 2 of the 14 animals (Fig. 3C). These results indicate that not all genes previously proposed as candidate anterior patterning genes (ndl-4 and sFRP-1; Chen et al., 2013) respond equally to egr-4 RNAi.

Taken together, our data confirm that egr-4 silencing impairs the proper differentiation of several cell types and organs (CNS and gut) during anterior regeneration. However, no defects were observed after egr-4 silencing in intact non-regenerating planarians; all CNS- and anterior-specific markers were normally expressed (supplementary material Fig. S7A). Moreover, neoblast proliferation was unaltered in intact egr-4(RNAi) animals when compared with corresponding controls (supplementary material Fig. S7B).

To determine whether the impaired head regeneration observed after egr-4 RNAi was due to defects in the establishment of anterior polarity, we analyzed the expression of the anterior polarity determinants Smed-notum, Smed-wnt1 and Smed-follistatin (Adell et al., 2009; Gurley et al., 2010; Petersen and Reddien, 2009, 2011; Yazawa et al., 2009; Roberts-Galbraith and Newmark, 2013). Control animals exhibited three typical phases of Smed-notum expression (Fig. 4). At 12 h post-amputation, Smed-notum expression was strongly upregulated in the wound region, giving rise to a dotted pattern of discrete cells. Between 18 h and 1 day after amputation, the number of Smed-notum-positive cells decreased (12 h to 1 day, phase I). After 3 days of regeneration, Smed-notum-positive cells coalesced at the tip of the anterior blastema (phase II). Finally, between 5 and 8 days post-amputation, the expression of Smed-notum became restricted to a small number of cells at the tip of the new head and along the midline, and to two bilateral groups of cells (arrowheads in Fig. 4; phase III), a pattern similar to that observed in intact planarians. No differences in the pattern of Smed-notum expression were observed between egr-4(RNAi) animals and controls up to 3 days post-amputation (Fig. 4). From this stage on, however, the progression of Smed-notum expression appeared to be delayed or arrested in egr-4(RNAi) animals (Fig. 4); after 3 days of regeneration, the clustering of Smed-notum-positive cells at the tip of the blastema seen in control animals was observed in only 3 out of 16 egr-4(RNAi) animals. By day 5 of regeneration, this clustering of Smed-notum-positive cells was observed in most egr-4(RNAi) animals. Normal Smed-notum expression was detected in several egr-4(RNAi) animals (n=7/22) by day 8; in others (n=9/22) the pattern of Smed-notum-positive cells was somewhere between that of phase II and III, while the remaining egr-4(RNAi) animals displayed no notum expression (n=6/22).

We observed no significant differences in the expression pattern of Smed-wnt1 between controls and egr-4(RNAi) animals (Fig. 4). Between 12 h and 1 day post-amputation, a dotted distribution of discrete Smed-wnt1-expressing cells was observed in the anterior wound. By 2 days post-amputation this anterior expression had been completely lost in both controls and egr-4(RNAi) animals, indicating that the lack of head regeneration after egr-4 silencing was not due to blastema posteriorization (Fig. 4). Smed-wnt1 expression in the tip of the tail was normal in all these samples (data not shown).

Smed-follistatin and Smed-notum co-expression was recently demonstrated in planarians in a small cluster of cells at the tip of the head (Roberts-Galbraith and Newmark, 2013). Functional analyses suggest that Smed-follistatin inhibits the Activin/ActR-1/Smad2/3 signaling pathway, which represses anterior regeneration. It has been proposed that Smed-follistatin and Smed-notum cooperate to promote anterior identity, thus acting as an anterior signaling center (Roberts-Galbraith and Newmark, 2013). We found that Smed-follistatin expression at the tip of the regenerating head re-appeared after egr-4 RNAi, although slightly later than in controls, as also observed for Smed-notum (data not shown). However, after 5 days of regeneration most animals displayed normal Smed-follistatin expression at their anterior-most tips (Fig. 4).

Overall, these results suggest that the polarity determinants Smed-notum, Smed-wnt1 and Smed-follistatin are normally expressed in the early regenerative stages during which AP polarity is established (Petersen and Reddien, 2009, 2011) and that egr-4 silencing impairs normal head regeneration without disrupting the establishment of anterior polarity.

As egr-4 was highly expressed in the mature and differentiating cephalic ganglia from very early stages of regeneration, we investigated its role in the regeneration of the cephalic ganglia by simultaneously silencing egr-4 and Smed-apc-1, an inhibitor of the Wnt/β-catenin pathway. apc-1 silencing results in the regeneration of a tail, rather than a head, from anterior wounds (Gurley et al., 2008). Remarkably, despite this polarity reversal, two small neuronal clusters known as ‘brain primordia’ differentiate within these posteriorized anterior blastemas (Evans et al., 2011; Iglesias et al., 2011). These brain primordia are equivalent to what it has also been termed ‘brain rudiments’ and correspond to the initial neuronal clusters that will differentiate into the new cephalic ganglia (Cebrià et al., 2002b; Kobayashi et al., 2007; Agata and Umesono, 2008).

As previously shown, egr-4 silencing inhibited the differentiation of normal cephalic ganglia (Fig. 5A). After apc-1 RNAi, approximately one-third of the tails regenerated small brain primordia with a similar morphology to that described previously (Evans et al., 2011; Iglesias et al., 2011). By contrast, no brain tissues were detected in most of the double egr-4(RNAi);apc-1(RNAi) regenerating tails, although in very few cases the differentiation of very reduced brain primordia was observed (Fig. 5A,B). Although in situ hybridization revealed no brain primordia in over half of the apc-1(RNAi) tails, differentiation of a pharynx was observed in most cases, indicating that the regenerative process had been somehow moved forward. By contrast, in egr-4(RNAi);apc-1(RNAi) animals, a greater number of tails failed to regenerate any brain primordia, when compared with either single apc-1 or egr-4 RNAi counterparts, and most of these failed to develop a normal pharynx (Fig. 5B). These results indicate that the differentiation of the brain primordia after apc-1 RNAi requires egr-4.

Given that egr-4 RNAi inhibits only anterior regeneration and the Wnt/β-catenin pathway mediates the specification of head versus tail regeneration, we further investigated the relationship between egr-4 and the Wnt/β-catenin pathway suggested by our data. Although most single egr-4(RNAi)-treated bipolar trunks displayed either smaller anterior blastemas with cyclopia or without eyes, treatment with double egr-4(RNAi);βcat1(RNAi) resulted in two-headed animals with normal anterior heads and two bilateral eyes, as also observed after βcat1 RNAi alone (Fig. 5C). The ability of βcat1 RNAi to rescue head regeneration in egr-4(RNAi) animals suggests that the egr-4(RNAi) phenotype requires βcat1, and indicates that in normal physiological situations egr-4 may antagonize β-catenin activity to allow head regeneration.

We next sought to delimit the period during which egr-4 is required for head regeneration. egr-4 was silenced by dsRNA injections 24 h and 48 h after amputation. Most of the animals injected 48 h post-amputation developed normal heads and only small number displayed defects in eye morphogenesis (Fig. 6A). By contrast, most of the animals injected 24 h post-amputation exhibited defects in the regenerated eyes (including ectopic and fused eyes; Fig. 6A). However, all of these animals regenerated normal-sized heads and brains (data not shown), in contrast to the severe impairment in head regeneration observed when egr-4 was silenced prior to amputation (Figs 2-5). Although it is unclear exactly how long after injection RNAi begins to exert its effect, these results suggest that egr-4 function is necessary during the first 2-3 days post-amputation, after which it may not be essential for proper head regeneration.

As this time window coincides with the development of the brain primordia (Cebrià et al., 2002b), we speculated that no phenotype is observed when egr-4 is silenced 48 h post-amputation because the brain primordia are already formed at this stage. We thus investigated whether egr-4 RNAi impaired head regeneration even when some brain tissue remained after amputation. Animals were injected with egr-4 dsRNA, amputated and allowed to regenerate. As expected, no defects in head regeneration were observed after this first round of amputation (see Materials and Methods). These animals were then re-injected (beginning 5 days post-amputation) and re-amputated at two different levels: level 1, immediately posterior to the newly regenerated eyes, leaving some of the regenerated brain tissue in the stump; and level 2, which involved decapitation and the removal of all brain tissue (Fig. 6B). Animals amputated at level 1 regenerated normally and differentiated normal brains as evidenced by anti-SYNORF-1 immunostaining and in situ hybridization with Smed-th (Fig. 6B). However, most of the decapitated fragments from level 2 amputations failed to regenerate proper heads or cephalic ganglia (Fig. 6B). Similarly, egr-4(RNAi) animals that were amputated sagittally along the midline regenerated properly (supplementary material Fig. S8). These results suggest that the presence of brain tissue is sufficient to rescue the blockade of head regeneration following egr-4 RNAi.

EGFR signaling plays important roles in many biological processes by activating or inhibiting many downstream pathways, including PI3K/AKT, MAPK (mitogen-activated protein kinase), PLCγ and JAK/STAT (signal transducer and activator of transcription) (Haley and Gullick, 2009). Previous reports suggest an important role for the EGFR signaling pathway in cell differentiation during planarian regeneration and homeostasis (Fraguas et al., 2011; Rink et al., 2011). Smed-egfr-1 regulates eye-pigment cell differentiation; Smed-egfr-3 seems to be necessary for blastema growth and cell differentiation (Fraguas et al., 2011); and Smed-egfr-5 is required for flame cell maintenance and regeneration, and for guiding branch extension in protonephridia (Rink et al., 2011). These observations are consistent with the demonstrated role of the EGFR signaling pathway in cell differentiation in other organisms (Jones et al., 2009; Lejard et al., 2011; Harris and Horvitz, 2011); for example, EGFR is required for the differentiation of mammary epithelial cells (Mukhopadhyay et al., 2013) and human neural progenitors (Lemcke and Kuznetsov, 2013). Here, we identify egr-4 as a putative target of Smed-egfr-3 (Fig. 1) that is required for anterior regeneration, likely through its regulation of neoblast differentiation.

Although Smed-egfr-3 is expressed in neoblasts and in the CNS of planarians (Fraguas et al., 2011; Fig. 1B), we found that egr-4 was mainly expressed in the CNS and in the mesenchyme in irradiation-insensitive cells, suggesting that egr-4 and Smed-egfr-3 are co-expressed in the CNS (compare Fig. 1A and 1B). Indeed, silencing of Smed-egfr-3 resulted in marked downregulation of egr-4 expression in the cephalic ganglia (Fig. 1C).

Like other planarian egr homologues, egr-4 expression was rapidly and locally upregulated after any small incision or during regeneration. Remarkably, this effect was independent of Smed-egfr-3 (Fig. 1E″,F). However, after 2 days of regeneration, the expression of egr-4 in the blastema became Smed-egfr-3 dependent (Fig. 1F). These results suggest that egr-4 is expressed in two distinct phases during regeneration, an early, Smed-egfr-3-independent, phase, and a subsequent phase that is controlled by EGFR signaling. These findings are supported by previous studies demonstrating that the expression of egr genes is regulated by the EGFR pathway in different organisms and contexts (Kaufmann and Thiel, 2002; Tsai et al., 2000, Lindzen et al., 2012).

The silencing of egr-4 impaired head regeneration, resulting in the development of either extremely reduced blastemas or small blastemas with aberrant photoreceptors and cephalic ganglia (Figs 2 and 3). Planarian regeneration is dependent on pluripotent stem cells known as neoblasts. Specific anti-SMEDWI-1 and Smed-histone-2B markers revealed no differences between control and egr-4(RNAi) animals (Fig. 2B). We did, however, observe differences in neoblast mitotic activity between these two groups. In egr-4(RNAi) animals, the first, wound-related mitotic peak (at 6 h) was markedly attenuated in anterior stumps and in posterior stumps from head sections, when compared with controls. This decrease may reflect a role of early egr-4 expression in this initial proliferative response. Interestingly, the second mitotic peak (associated with tissue loss and appearing at 48 h) was significantly attenuated only in anterior blastemas. This decrease may be a consequence, at least in part, of the marked attenuation of the first mitotic peak. However, because the decrease in the second peak was specific to anterior regeneration, we cannot rule out a potential role of later egr-4 expression in regulating neoblast proliferation (directly or indirectly) at this stage. A normal bimodal proliferative response was observed after egr-4 RNAi. Together with the slight attenuation of neoblast proliferation seen during posterior regeneration and the presence of a large number of mitotic cells in anterior blastemas at all stages (supplementary material Fig. S6), these findings suggest that the severe phenotypes observed after egr-4 silencing are not solely due to defects in proliferation. Although egr-4 silencing had no effect on early neoblast differentiation (normal anti-SMEDWI-1 and Smed-NB21.11e expression was observed), it resulted in a significant decrease in late neoblast progeny (Fig. 2D), suggesting that egr-4 RNAi impairs late neoblast differentiation.

Taken together, these results suggest that egr-4 is expressed in two distinct phases during regeneration. Early egr-4 expression may participate in the initial (injury-induced) proliferative response, whereas the late expression, mediated by EGFR signaling, appears to regulate cell differentiation during anterior regeneration. These data provide the first evidence of a conserved EGFR/egr pathway in planarians. Further experiments will be necessary to better understand the relationship between egr-4 and Smed-egfr-3, and to elucidate their specific roles during anterior and posterior regeneration.

egr genes have been implicated in cell proliferation, differentiation, inflammation, apoptosis, wound healing and liver regeneration (Thiel and Cibelli, 2002; Dussmann et al., 2011). Several egr genes have been identified in planarians (Sandmann et al., 2011; Wagner et al., 2012; Wenemoser et al., 2012).

Many of these genes, including egr-4, are upregulated immediately after injury or amputation and are barely expressed (if at all) in intact planarians (Sandmann et al., 2011; Wenemoser et al., 2012). To date, no functional characterization of the injury-induced expression of planarian egr genes has been reported. However, given the decrease reported here in the first mitotic peak after egr-4 RNAi and the demonstrated role of several egr genes in wound healing, tissue fibrosis and inflammatory responses in other models (Schmidt et al., 2008; Wu et al., 2009; Chen et al., 2006), it seems plausible that early egr genes expression in planarians is involved in the initial stages of regeneration.

egr-4 was also expressed in the mature and differentiating CNS (Fig. 1). Given the conserved function of egr genes in neural development (reviewed by O'Donovan et al., 1999; Pérez-Cadahía et al., 2011) and the severe phenotypes observed in the regenerating CNS after egr-4 RNAi, we investigated whether this gene is required for the regeneration of cephalic ganglia in planarians. The results of double RNAi of egr-4 and Smed-apc-1 (Fig. 5) suggest that egr-4 is necessary for the early differentiation of the brain primordia. However, egr-4 was not required for the maintenance of the CNS either in intact planarians (supplementary material Fig. S7) or in regenerating sections in which brain tissue remained after amputation (Fig. 6). These findings identify egr-4 as the first gene known to be involved in the initial stages of neural regeneration in planarians. This role in neural differentiation is in agreement with functions attributed to egr genes in other organisms. For example, egr-1 regulates astrocyte growth through via activation of the ERK signaling cascade triggered by EGF receptors (Mayer et al., 2009; Biesiada et al., 1996). In cultured PC12 cells, p35 is essential for neurite outgrowth and is induced via the egr-1-mediated activation of ERK (Harada et al., 2001). Similarly, egr-1 regulates neurite extension during development in Xenopus (Anelli et al., 2013), and ERK-dependent egr-4 expression is required for the maturation of GABAergic neurons (Ludwig et al., 2011a).

The silencing of different genes required for the respecification of anterior polarity or pole formation produces phenotypes similar to those described here after egr-4 RNAi (Petersen and Reddien, 2009, 2011; Almuedo-Castillo et al., 2012; Roberts-Galbraith and Newmark, 2013; Blassberg et al., 2013; Chen et al., 2013). It is thus possible that the regeneration defects observed after egr-4 RNAi are due to dysregulation of the respecification of polarity, given that double egr-4(RNAi)/β-catenin(RNAi) rescues proper head regeneration. However, although the knockdown of polarity or pole determinants completely silences Smed-notum (Roberts-Galbraith and Newmark, 2013; Blassberg et al., 2013; Chen et al., 2013), the expression of Smed-notum and Smed-follistatin, genes that are required for anterior polarity (Roberts-Galbraith and Newmark, 2013; Petersen and Reddien, 2011), is upregulated and largely maintained in regenerating egr-4(RNAi) animals. After normal upregulation of Smed-notum in egr-4(RNAi) animals during the first 24 h of regeneration (neoblast-independent upregulation; Chen et al., 2013), the recovery of the normal expression pattern in the days that follow is delayed when compared with controls. This delay may be associated with the attenuation of the second mitotic peak induced by egr-4(RNAi). By 5-8 days of regeneration, most animals exhibited Smed-notum expression at the tip of the blastema. Moreover, egr-4(RNAi) did not affect the normal expression of Smed-follistatin at the blastema tip. We cannot rule out the possibility that the delay in restoring Smed-notum and Smed-follistatin expression at the tip of the blastema in egr-4 knockdowns in turn influences polarity respecification (and/or maintenance), thus giving rise to the observed defects. Nonetheless, our findings suggest that the impaired head regeneration observed after egr-4 RNAi is caused by disruption of egr-4-mediated differentiation of the cephalic ganglia.

The results of the double egr-4(RNAi)/β-catenin(RNAi) suggest that the impairment of head and cephalic ganglia regeneration after egr-4 silencing is β-catenin dependent. Thus, in wild-type planarians, egr-4 may inhibit β-catenin activity to allow head regeneration. In support of this hypothesis, defects in brain regeneration were more severe in egr-4(RNAi);apc-1(RNAi) animals, in which β-catenin activity is augmented. These results suggest that the effect of egr-4 on cephalic ganglia regeneration is at least partially mediated by antagonism of the Wnt/β-catenin pathway. Although the Wnt/β-catenin pathway has been primarily associated with the establishment of AP polarity (Iglesias et al., 2008; Gurley et al., 2008; Petersen and Reddien, 2008), other studies suggest that this pathway may regulate the differentiation of cephalic ganglia independently of axial polarity (Iglesias et al., 2011; Blassberg et al., 2013).

Overall, our data point to egr-4 as a unique factor that plays a key role in the differentiation of the brain primordia by antagonizing β-catenin function downstream of the polarity determinants Smed-notum and Smed-wnt1. Further experiments will be required to unravel the exact relationship between egr-4 and Wnt/β-catenin signaling.

If egr-4 is required for the early development of the cephalic ganglia, how does the inhibition of CNS differentiation block head regeneration? Previous studies have shown that the inhibition of Smed-egfr-3 (Fraguas et al., 2011) and ERK signaling (Tasaki et al., 2011) yields phenotypes similar to those obtained after egr-4 RNAi: blastema growth is severely affected due to impaired cell differentiation. Tasaki and co-workers (2011) have suggested that ERK activation within the blastema is necessary for the early differentiation of an initial cohort of cells that is subsequently required to induce neoblast proliferation in the stump, and is probably necessary for the migration of neoblast progeny into the blastema, ultimately promoting blastema growth. We propose that this cohort of cells that is necessary to maintain blastema growth might be the initial brain primordia, the differentiation of which is dependent on egr-4. Under normal physiological situations, the brain primordia could emit some form of signal to promote neoblast proliferation in the stump and/or the migration of neoblast progeny cells into the blastema, where they subsequently differentiate. However, this mechanism may be dysregulated in egr-4(RNAi) animals, in which differentiation of the brain primordia is impaired (Fig. 7). Future experiments will be required to investigate this hypothesis. Further suggesting that brain cells are required for blastema growth, regenerating egr-4(RNAi) planarians in which some brain tissues was retained following amputation (Fig. 6; supplementary material Fig. S8) were capable of regenerating normal heads with proper brains. Given that EGFR/ERK signaling regulates egr genes in several other contexts (Mayer et al., 2009; Harada et al., 2001; Tarcic et al., 2012; Mukhopadhyay et al., 2013), planarians provide a unique opportunity with which to study the function of this conserved pathway during regeneration.

A recent study proposed that during planarian regeneration neoblasts follow a default program, triggered by ERK activation, to differentiate into head tissues (Umesono et al., 2013). These authors suggest that in posterior regions the Wnt/β-catenin pathway inhibits this ERK activation, resulting in regeneration of a tail. It will thus be of interest to analyze the relationship between egr-4 and ERK in future studies to determine to what extent egr-4 mediates the differentiation of head tissues after ERK activation.

Finally, the requirement of the brain primordia for blastema growth supports an evolutionarily conserved role of the nervous system in animal regeneration (Kumar et al., 2007; Kumar and Brockes, 2012; Miljkovic-Licina et al., 2007; Singer and Craven, 1948). Although several planarian studies have proposed a role of the nervous system in regeneration (Baguñà et al., 1989; Cebrià and Newmark, 2007; Oviedo et al., 2010; Stéphan-Dubois and Lender, 1956), the underlying molecular basis remains unknown. Accordingly, egr-4 may be the first identified gene linking the differentiation of the brain primordia with head regeneration in planarians.

Schmidtea mediterranea from the BCN-10 clonal line were used in all experiments. Planarians were starved for at least 1 week prior to experiments. Genes and RNAi experiments were named using the nomenclature proposed by Reddien et al. (2008).

For the DGE experiment, total RNA from control (GFP) and Smed-egr-3(RNAi) planarians was extracted 1 day and 3 days after amputation using TRIzol (Invitrogen). Libraries were generated by the Skuldtech transcriptomic service (Montpellier, France). Sequence tag preparation was performed using Illumina's Digital Gene Expression Tag Profiling Kit according to the manufacturer's instructions (version 2.1B). Cluster generation was performed after applying 4 pM of each sample to the individual lanes of the Illumina 1G flowcell according to the manufacturer's instructions. Image analysis and base calling were performed using the Illumina Pipeline, from which sequence tags were obtained after purity filtering. This was followed by sorting and counting the unique tags.

Four libraries were obtained from control and Smed-egfr-3(RNAi) animals on days 1 and 3 post-amputation (henceforth referred to as C1, C3, R1 and R3) (supplementary material Table S1). For comparison between libraries, tags with a sum of occurrences below 10 in the two compared sets or with a P>0.005 were discarded. P-value determination was performed as previously described (Piquemal et al., 2002). Tags were mapped using the short sequence mapping tools SeqMap (Jiang and Wong, 2008) and MPscan (Rivals et al., 2009) against the Smed454_90edb transcriptome dataset (Abril et al., 2010). Annotation of the sequences and assignation of Gene Ontology (GO) categories was performed using the Blast2GO suite (Conesa et al., 2005).

Double-stranded RNAs (dsRNA) were delivered into the planarian digestive system for three consecutive days as previously described (Sánchez-Alvarado and Newmark, 1999). All controls were injected with GFP dsRNA. After one round of dsRNA injections and amputation all egr-4(RNAi) planarians regenerated well. Defective phenotypes were identified after two rounds of injections and amputation. To increase the penetrance of the phenotypes the animals were injected with dsRNA for 3 consecutive days, cultured un-cut for 5 more days, re-injected for 3 consecutive days and then amputated pre- and post-pharyngeally. All experiments were performed following this protocol unless otherwise specified. In all figures showing a quantification of the RNAi phenotypes when the fractions do not add up to 100% the missing animals are normal, unless specified in the main text.

Gene expression analysis was carried out by whole-mount in situ hybridization, as previously described (Molina et al., 2007; Umesono et al., 1997; Cebrià et al., 2007; Pearson et al., 2009). All samples were observed through a Leica MZ16F stereomicroscope and images from representative organisms of each experiment were captured with a ProgRes C3 camera from Jenoptik. Confocal laser scanning microscopy was performed with a Leica SP2. A lethal dose of 100 Gy was used in irradiation experiments, and animals were fixed and hybridized at the indicated time points.

Immunostaining was carried out as described previously (Cebrià and Newmark, 2005). The following antibodies were used: anti-SYNORF-1, a monoclonal antibody specific for synapsin, which was used as a pan-neural marker (Cebrià, 2008) (1:10; Developmental Studies Hybridoma Bank); anti-SMEDWI-1 (1:1500; Guo et al., 2006; März et al., 2013); and anti-phospho-histone H3 (H3P), which was used to detect mitotic cells (1:300; Cell Signaling Technology). Alexa 488-conjugated goat anti-mouse (1:400) and Alexa 568-conjugated goat anti-rabbit (1:1000; Molecular Probes) were used as secondary antibodies. Confocal laser scanning microscopy was performed using a Leica TCS-SPE and a Leica SP2. Confocal stacks from representative organisms in each experimental condition are shown.

The authors thank Chema Martín-Durán for his help with the phylogenetic analyses and Josep Francesc Abril for providing the bioinformatics resources for analysis of the DGE data. We also thank Emili Saló and Teresa Adell for helpful comments and discussions. We thank Kerstin Bartscherer for the anti-SMEDWI-1 antibody. We thank Owen Howard for advice on English style.