Putrescine independent wound response phenotype is produced by ODC-like RNAi in planarians

Despite increasing evidence indicates polyamines as a convergence point for signaling pathways, including cell growth and differentiation, a unifying concept to interpret their role is still missing. The activity of ornithine decarboxylase (ODC), the rate-limiting enzyme in polyamine biosynthesis, is tightly regulated by a complex molecular machinery, and the demonstration of the existence of multiple ODC paralogs, lacking decarboxylation activity, suggests additional layers of complexity to the intricate ODC regulatory pathway. Because of their extraordinary regenerative abilities and abundance of stem cells, planarians have potential to contribute to our understanding of polyamine function in an in vivo context. We undertook a study on ODC function in planarians and we found six planarian ODCs (ODC1-6). Five out of six ODC homologs carry substitutions of key aminoacids for enzymatic activity, which makes them theoretically unable to decarboxylate ornithine. Silencing of ODC5 and 6 produced a complex phenotype, by prompting animals to an aberrant response, following chronic injury without tissue removal. Phenotype is neither rescued by putrescine, nor mimicked by difluoromethylornithine treatment. Moreover, the co-silencing of other genes of the ODC regulatory pathway did not modulate phenotype outcome or severity, thus suggesting that the function/s of these ODC-like proteins might be unrelated to decarboxylase activity and putrescine production.


Results
Identification of polyamine metabolic enzymes. We identified six different partial sequences (DjODC1, DjODC2, DjODC3, DjODC4, DjODC5 and DjODC6), coding for putative ornithine decarboxylase-like proteins in the species D. japonica, via bioinformatic analysis (Table S1, panel A). Multialignments of both nucleotide and predicted protein sequences revealed that the six genes were significantly different from each other, as indicated by the percentages of identity and similarity indicated in Table 1. Interproscan analysis of the predicted protein sequences 24 allowed us to identify each sequence as a member of the Ornithine/DAP/Arg decarboxylase family (IPR000183), Ornithine decarboxylase sub-family (IPR002433). In order to investigate whether the six putative ODC isoforms are conserved in the closely related planarian species S. mediterranea, we searched for transcripts annotated as "similar to ornithine decarboxylase" in the Schmidtea mediterranea genome database (SmedGD) 25 , and we managed to group the results into five (A to E) classes of sequences, each class representing a S. mediterranea homolog of ODC (Table S2). Representative sequences from each class were elongated by searching in the Planmine database 26 . According to the expected value (E-value) resulted from alignments of each D. japonica ODC nucleotide sequence, to S. mediterranea representative members of class A, B, C, D and E, we managed to match D. japonica ODCs with their corresponding putative homologs in S. mediterranea. The results obtained are shown in Supplementary Table 1 (panel B). A one-to-one correspondence exists for all ODCs but Smed-odc class A, that is equally close to DjODC1 and 2. Accordingly, we named the S. mediterranea ODCs with the progressive numbers of the corresponding D. japonica homologs (Smed-odc-3 to 6), except for S. mediterranea class A that has been called Smed-odc-A. We will refer to as "ODCn" when considering the same ortholog from both D. japonica and S. mediterranea species (i.e. "ODC5" for both DjODC5 and Smed-odc -5). Finally, we decided to compare the predicted protein sequence of each D. japonica and S. mediterranea ODC with the protein sequence of human ODC, with respect in particular to the conservation rate of the 18 selected key residues which have been described as critical amino acids for ODC catalytic activity 14 . As shown in Table 2, ODC2-6 have aminoacidic substitutions in at least one of the 18 key amino acids. Strikingly, both ODC5 and ODC6 display substitutions in one out of the two essential amino acids, Cys360Val (ODC5) and Lys69Ser (ODC6).
In our search, we also identified in D. japonica EST collection a single sequence (FY937407.1), called DjOAZ, homologous to ornithine decarboxylase antizyme (OAZ) (best blast hit at aminoacidic level with: putative Ornithine decarboxylase antizyme [Daphnia magna] 4e −08 ), and a single antizyme sequence (dd_Smed_ v6_425_0_1) in S. mediterranea Planmine database 26 . For DjOAZ an in-silico analysis of the putative ORFs was undertaken. We managed to find the peculiar ORF1 and ORF2 distinguishing antizyme proteins (ref. 27, and references therein) (Supplementary Figure 1A); the site of +1 frameshift was characterized by the sequence AUU-UGA-UGA-G (Ile-Stop-Stop), which produced a shift of the reading frame, resulting in AUU-GAU-GAG  Table 1. Identity (bold) and similarity (Italic) percentages between D. japonica ODC putative proteins.
(Ile-Asp-Glu), allowing the synthesis of the full-length protein.
In addition, we also identified a distinctive mRNA secondary structure (hairpin) downstream of the frameshift site, which showed similar features to those of the human OAZ (Supplementary Figure 1B,C).
Expression of polyamine metabolism enzymes in intact planarians. The analysis of ODC transcript distribution in intact animals revealed a surprisingly complex scenario of expression patterns, involving different types of tissues. DjODC1 was found predominantly accumulated in the gut (Fig. 1A,B). However, a low widespread expression was observable in several body tissues. DjODC2 was expressed in the epidermis, with a homogeneous distribution on the ventral surface, and confined to the head on the dorsal surface, with a particularly intense signal in the auricle area ( Fig. 1B-D). Magnification of DjODC2 staining on the ventral surface shows that most epidermal cells are labelled, yet at different level (Fig. 1E). DjODC3, 4, 5, and 6 shared similar expression patterns ( Fig. 1F-I), being expressed in small cells spread in the parenchyma, both anterior to the photoreceptors, and in the periphery of the animal along the anteroposterior and dorsoventral axes. This expression pattern is apparently very similar to that of late epidermal committed neoblast progeny markers (Category 3 genes) 28 .
In order to assess the degree of differentiation of the cells expressing DjODC1-6, we analysed the expression levels of these genes by Real Time RT-PCR in intact 30 Gy-treated animals, which were sacrificed 2, 4, 7 and 10 days after X-irradiation. X-ray irradiation in planarians results in rapid and selective elimination of neoblasts, with differentiated cells remaining unaffected at the same time and dose 29 . After irradiation, the expression levels of DjODC1 and DjODC2 remained unchanged during the whole time-course (Fig. 1L), demonstrating that these genes are expressed in late neoblast progeny. On the contrary, the expression levels of DjODC3, 4, 5, and 6, decreased progressively starting 7 days after irradiation (Fig. 1L). The recent analysis of single planarian cell transcriptome 30 , and the availability of Planarian SCS database (https://radiant.wi.mit.edu/) allowed us to verify that Smed-odc-3/4/5 and 6 are all expressed in late epidermal-committed neoblast progeny. In addition, we found that expression of Smed-odc-3 -6 was dramatically reduced in Smed-zfp-1 RNAi animals (Fig. 1M), in which the formation of z-neoblast, the epidermal lineage precursors, is impaired 22 . This undoubtedly locates Smed-odc-3 -6 expression in epidermal precursors cells.
DjOAZ has a very wide expression pattern (Fig. 1N), consistent with that reported in Planaria SCS @ The Whitehead Institute database for its S. mediterranea homolog, suggesting an almost ubiquitous expression.
Macroscopic effect of DjODCs silencing by dsRNA injection. As ODC is the rate limiting enzyme in polyamine biosynthesis, we decided to knock down the expression of each of these genes individually by RNAi, to better dissect their roles. dsRNA molecules are normally delivered to planarians by mixing them with liver paste and feeding. However, as liver contains high levels of polyamines, we decided to perform RNAi on both intact and regenerating planarians by injecting dsRNAs. Injections were performed in the anterior gut branch by punching the animals in the space between the eyes and the pharynx, a region that here will be referred to as "neck". The effect of gene silencing was investigated at different time points until day 21 after the first injection; regeneration rate was assessed by morphometric analysis in animals cut in the neck region 7, 11 and 14 days after the first dsRNA injection ( Fig. 2A). DjODC1, DjODC2, DjODC3 and DjODC4 RNAi animals were not affected by gene silencing, as both the intact and the regenerating individuals showed no substantial difference compared to controls (Fig. 2B,D). In contrast, DjODC5 and DjODC6 gene silencing caused deleterious effects in both intact and regenerating animals. Eleven days after the first dsRNA injection, both head and tail fragments of DjODC5 RNAi animals showed severe delays in the regeneration process. Fourteen days after the first dsRNA injection, DjODC5 RNAi animals produced small sized blastemas, less than half compared to controls; blastema size was particularly reduced in tail fragments ( Fig. 2B-E). In some cases (see below), the head fragments died after the cut. DjODC6 RNAi animals showed a statistically significant delay in regeneration (including some heads' death after the cut) at this time point as well (Fig. 2B-E). Co-silencing of both ODC genes did not produce a worsening of the regeneration delay (data not shown). After 4/5 dsRNA injections, intact DjODC5 and DjODC6 RNAi animals acquired a peculiar phenotype, marked by a progressive reduction of head size, with an evident disappearance of auricles protrusions (Fig. 3C,H), and constriction of the neck region. Later on, the head firstly became roundish, then it was retracted backwards into the "neck" region ( Fig. 3D,E,I,L). These particular morphological changes were displayed by 100% of DjODC5 RNAi intact animals (30/30). In rare cases (5%), head was eventually completely enclosed into the region anterior to the pharynx (Fig. 3F). This morphological change resulted in apparently headless animals, yet having the eyes still visible through the trunk, a unique phenotype that we called "blemmy". A similar phenotype was observed in DjODC6 RNAi intact animals (20/30 animals). Co-silencing of both ODC genes did not produce a worsening of the head regression phenotype. We verified the effectiveness and specificity of gene silencing 7 days after the first injection by Real-Time RT-PCR. A strong reduction of both DjODC5 and DjODC6 transcript levels was obtained, following dsRNA injection, and no cross-silencing was observed for other ODC transcripts (data not shown). Once established, blemmy phenotypes survived up to about 15 days, and then died. Supplementary figure 2 shows histological analysis of a representative blemmy phenotype, revealing   that the enclosed head was surrounded outwardly by parenchymal tissue near the site of retraction. Head retraction and the blemmy phenotype were specifically obtained when injections were performed in the neck region. Surprisingly, when injections were performed in one of the posterior gut branches, we observed thickening of the tail, followed by constriction towards the injection site, to produce arched planarians: the phenotype was therefore injection site-dependent (Fig. 3M,N). An injection-site directed impairment was also evident when the death rate of amputated fragments, subjected to injection in different body regions, was analysed. As shown in Supplementary Figure 3, a higher number of dead fragments was detected when the cut was performed in the same region of injection, while only sporadic dead fragments were found in control planarians injected with water. A similar analysis performed in S. mediterranea specimens revealed an identical scenario, in which only Smed-odc-5 and 6 produced a site injection-dependent "blemmy-like" phenotype (Supplementary Figure 4A-C).

Analysis of cell proliferation and death.
To understand the cellular events underlying this peculiar phenotype, we decided to evaluate cell proliferation and apoptosis rates. The amount of mitotic cells was unaltered in RNAi planarians, with respect to controls (Fig. 3P,Q). Strikingly, RNAi animals showed a significant increase of apoptotic cells, visualized by TUNEL reaction. Apoptosis activation was systemic, but TUNEL-positive cells were particularly concentrated in the body region where dsRNA was injected (Fig. 4). A similar activation of apoptotic response was also demonstrated in S. mediterranea specimes (Supplementary Figure 4D). Consistent with the activation of apoptosis, the expression of Smed-JNK 31 was significantly (p < 0.05) increased in RNAi-treated animals (1,8 ± 0.1 and 1,6 ± 0.08 folds vs control in Smed-odc-5 and Smed-odc-6 RNAi animals, respectively).
Histological and ultrastructural analysis. To better understand the effects of gene silencing on planarian tissue organization, we decided to perform morphological investigation of the principal planarian tissue/ organs, in animals injected with dsRNA molecules in the neck region. Histological staining of D. japonica tissue sections revealed important structural alterations in epidermis, namely a reduction of thickness, a lower density of cilia, and an impairment of apical specialization in RNAi animals, with respect to controls (Fig. 5A,B and Supplementary Figure 5A). These observations were confirmed by ultrastructural analysis (Fig. 5C,F and Supplementary Figure 5B,C). Moreover, epidermal cells appeared disorganized, with changes in nuclear shape and localization. Basal lamina was highly disorganized, thinner (Fig. 5D,F), and in some cases absent. A reduction of epidermal nuclei density was also evident, although not as prominent as alterations in cell shape and ultrastructure (Fig. 6A,B). These effects were injection site-dependent, being maximal when injections were performed in the head region. Similar results were obtained in S. mediterranea (Supplementary figure 4E-G). With the aim to investigate possible alterations in other tissues, we decided to analyze the expression of specific tissue markers. Stem cells, early epidermal-committed stem cell progeny, late epidermal-committed stem cell progeny, nervous system, gut and protonephridia, were visualized by whole mount in situ hybridization using as probes DjPiwiA, DjNB21.11e, DjAGAT2, DjSyt, DjInnexin1 and DjInnexin 10 32 , respectively. Muscles were visualized using the D. japonica homolog of the S. mediterranea general muscle marker Smed-collagen 33 , here called DjCollagen. According to what observed above for the central nervous system marker ( Fig. 3G-L), and to data obtained with the other markers ( Fig. 7 for DjODC5 RNAi, data not shown for DjODC6 RNAi), we concluded that the morphology of protonephridia, gut, nervous system and muscle cells was not affected by RNAi, with the exception of structural reorganization due to head contraction. Conversely, either the expression level or the number of positive cells, for early and late stem cell progeny markers, was significantly increased, and a slight increase was detected in stem cell marker expression (Fig. 7). Effect of DjODCs silencing by dsRNA feeding. Some aspects of ODC5 and ODC6 RNAi phenotype are clearly the consequence of the synergistic effect of both ODC silencing and the chronic injuries (without tissue removal) produced by the repetitive injection in a specific body region. To identify which phenotype features are the exclusive consequence of ODC silencing and those that are due to knock-down and chronic punching we decided to deliver dsRNA molecules by feeding using egg yolk, which is known to contain low amounts of polyamines 34 , rather than liver paste as food. D. japonica and S. mediterranea were fed two times per week with dsRNA molecules. Effectiveness of RNAi was tested one week after the first feeding by Real Time RT-PCR analysis. dsRNA feeding was highly effective in reducing endogenous cognate mRNA level (80% of reduction with respect to planarians fed with egg alone). Planarian specimens were sacrificed 20 days after the first feeding to analyze morphological phenotype, apoptosis, cell proliferation, epidermal cell structure and density. As expected, in planarians fed with ODC5 or ODC6 dsRNA molecules, we never observed the appearance of the morphological phenotype; activation of apoptosis, JNK upregulation, changes in cell proliferation, alteration in epidermal cell structure and density were absent as well (Supplementary Figure 6). Analysis of the effect of OAZ silencing, ornithine decarboxylase inhibitor DL-α DiFluoroMethylOrnithine (DFMO), and putrescine treatment, on ODC5 and ODC6 RNAi phenotype. Table 3 summarizes the effect of different dsRNA or dsRNA combinations in accordance to the way of administration and the co-treatment with other substances. In addition to what has been described in the paragraphs above, we found that OAZ dsRNA, putrescine or DFMO injection and treatment with either putrescine or DFMO, in parallel with water injection to mime punching, did not produce macroscopic phenotypic changes. The ability of D. japonica to take up putrescine from the medium was tested evaluating its ability to rescue DFMO-induced phenotype during regeneration, as previously described by ref. 35 ( Supplementary Fig. 6I). The co-silencing of OAZ with either ODC5 or ODC6, did not rescue ODC5 and ODC6 RNAi macroscopic phenotype. Strikingly, the same results were obtained when putrescine or DFMO were added to the medium of ODC5 or ODC6 RNAi animals.

Discussion
Very little is known regarding ODC function and activity in invertebrate animals. As shown by previous bioinformatics studies investigating the variability of putative AZIs and ODCs in vertebrates and invertebrates 14 , most  of invertebrates have only a single copy of ODC-like genes in their genome. However, a few exceptions exist. Mosquito species like A. aegypti, C. pipiens, and A. gambiae were shown to be characterized by no less than 6 ODC homologs. However, as reported by the authors, only one of these homologs seems to be highly transcribed and (potentially) catalytically active, suggesting that this is the main ODC enzyme in mosquitoes. In contrast, the authors state that only one or zero ESTs were found to support the other 5 homolog expression, suggesting very low levels of transcription. Other invertebrates were found to have more than one ODC-like gene in their genome. This is the case of the chordate species C. intestinalis and B. floridae, as well as the cnidarian H. magnipapillata. All of them have at least one homolog with an unaltered set of the 18 key amino acids; however, no expression or functional studies have been carried to characterize the activity of ODC-like genes in these organisms. Planarians possess multiple homologs of putative ODC. The presence of substitutions in the set of 18 key amino acids essential for vertebrate ODC activity suggests that most planarian ODCs would have theoretically lost their catalytic activity, thus resembling, at least from this point of view, the vertebrate AZI. The presence of multiple copies of ODC genes in invertebrate species evolutionarily far from each other suggests that in these groups some ODC homologs could have diverged from the original decarboxylase function, especially those displaying amino acid substitutions in the key 18 residues necessary for ODC activity.
Planarian putative ODCs are expressed in post-mitotic differentiated and undifferentiated tissues but not in stem cells. This suggests that they might not be necessary for stem cells maintenance, but rather in differentiation of daughter cells.
Among the planarian ODC paralogs that we found, only one (namely DjODC1 for D. japonica and Smed-odc-A for S. mediterranea) does not display substitutions in any of the 18 key amino acids, suggesting a fully functional ornithine decarboxylase activity. Surprisingly, RNAi of ODC1/A did not produce altered phenotypes, both in regenerating and intact animals. Similarly, silencing of OAZ, the direct inhibitor of ODC, does not produce any detectable macroscopic phenotype. These two findings suggest that the production of polyamines by ODC pathway could have only a secondary role in planarian homeostasis; if this would be the case, polyamines could be easily supplied directly from food, and subsequently stored within the cell. Alternatively, the presence of a parallel pathway of polyamine production (i.e. agmatinase-dependent production of putrescine), could be responsible for the lack of ODC1/A RNAi-phenotype in planarian homeostasis. Further loss of function studies for this isoform will have the potential to reveal interesting polyamine functions in planarians, as suggested by the finding that DFMO inhibits regeneration 35 .
Conversely, ODC5 and ODC6 silencing resulted in a peculiar complex phenotype, leading animal to death, which is why we focused our attention on these paralogs.
ODC5 and ODC6 RNAi leads the animal to acquire deleterious morphological alterations that were observable only when dsRNA molecules were delivered by microinjection, a condition in which a kind of chronic injury without loss of tissue is produced by the repeated punching. Indeed, the macroscopic phenotype was injection-site dependent, meaning that the animal underwent severe changes only in those regions where the injection was performed, a condition never observed in water-injected control animals.
The dramatic effect observed in epidermis morphology, as well as the upregulation of epidermal precursor markers produced by ODC5 or 6 dsRNA injection, led us to hypothesize that ODC5 and 6 could play a role in late phases of epidermal differentiation. However, epidermal aberrations were not observed in animals in which dsRNA had been delivered by feeding, suggesting that epidermal aberrations are not a consequence of impaired tissue turnover. Head or tail regression observed in ODC5 and ODC6 RNAi-injected animals is caused by a massive activation of apoptosis in the body region in which injections have been performed. Moreover, we observed a significant increase in expression in JNK transcript only in ODC5 or 6 RNAi-injected animals. Previous studies demonstrated that JNK is only essential for responses that require the regeneration of missing tissues and it is required during the first stages of regeneration, to properly induce the expression of early wound-induced genes, the initiation of the wound-associated apoptotic peak and the controlled onset of mitosis 31 . In conclusion, we can hypothesize that silencing of ODC5 or 6 leads to no consequence if the animal is not injured, but causes serious morphological aberrations upon a chronic injury like injection. For example, the process of epidermal wound healing upon continuous punching may be impaired, making the animal to improperly perceive it as absence of tissue. This, in turn, might lead to the activation of an unnecessary regenerative response 36 , starting with the induction of apoptosis 37 , and followed by the appearance of the phenotype. Despite this appealing hypothesis, we cannot exclude the existence of other reasons that might be at the basis of differences between fed and injected organisms. For example in fed organisms, a higher tissue turn-over rate may counteract ODC5/6 RNAi effect in epidermal differentiation.
Previous studies showed a rescue of the phenotype produced by DFMO when the medium was supplied with putrescine 35 . After silencing of ODC5 and 6, if RNAi phenotype had been the consequence of reduction in polyamine synthesis, it should have been rescued by the addition of putrescine to the planarian medium or injection of this molecule in the gut. As this was not the case, we suggest that the phenotype cannot be ascribed to a putrescine production-dependent role for these proteins. This possibility is also confirmed by the inability of DFMO treatment or injection to mime ODC5 or 6 RNAi phenotype. These observations also allow us to exclude the possibility that ODC5 and 6 behave as an AZI protein, in rescuing an enzymatically active paralog (i.e. ODCA or 1) from OAZ-dependent degradation. A further hypothesis to be tested is the possibility that ODC5 and 6 form heterodimers with an enzymatically active paralog and inactivate it, thus functioning as ornithine decarboxylase inhibitors. In this case, the consequence of RNAi should be an increase in putrescine concentration that might be the cause of the phenotype. However, planarian raised in excess of putrescine and injected with water to mime the punching or directly injected with putrescine, did not develop the phenotype and the treatment of ODC5 or 6 RNAi animals with DFMO did not rescue the phenotype. Considering the numerous reports on avid polyamine uptake systems in Eukarya 4, 28-41 and our evidence about the ability of planarian species used in this paper to take up putrescine from the medium, a very conceivable hypothesis is that ODC5 and 6 proteins play an additional polyamine-unrelated function, the loss of which produces the aberrant wound response phenotype displayed by RNAi-treated animals.
In both S. mediterranea and D. japonica species, ODC5 and 6 are expressed in epidermal-committed neoblast progeny cells, and their almost complete co-localization indicates that these two isoforms are expressed in the same cells. This striking result can account for either a sub-cellular compartmentalization of the two enzymes within the cell (cytoplasmic and nuclear localization for ODC has been indeed demonstrated; see 42 and references therein), or a completely new condition in which they form a heterodimer. In support of the latter hypothesis, we found that single ODC5 or ODC6 RNAi and co-silencing of the two isoforms produced the same phenotype, at least for the parameters considered in this paper. Further studies will be necessary to elucidate whether these two isoforms may interact with each other. However, if this is the case, given that ODC5 and ODC6 show complementary aminoacidic substitutions in the two key positions for catalytic activity, this would lead to the formation of two differentially active catalytic pockets, with one of them bearing both loss-of-function substitution of Lys69 (N-terminus of ODC6) and Cys360 (C-terminus of ODC5) and the other one bearing a fully-functional active site (N-terminus of ODC5 and C-terminus of ODC6). A previous study demonstrated that when mutations in Lys69 and Cys360 are introduced in two different ODC cDNAs, the catalytic activity of the dimer is maintained, yet performed by only one functional active site 43 .
The discrepancy between the phenotypes observed in ODC5/6 RNAi animals and those described for DFMO by 35, led us hypothesize that ODC5/6 might have functions, additional to the putatively halved decarboxylase activity, located in a domain that has nothing to do with ornithine decarboxylation. In fact, when using the  inhibitor, the protein is still present in the cell, yet inhibited to perform decarboxylation; in this case, if a particular domain is involved in performing another function, this would be unlikely to be disrupted. Conversely, with RNAi the whole protein is depleted from the cell, leading to the absence of both decarboxylase domain and the domain supposedly involved in another activity; only the actual absence of the latter would lead to the appearance of the polyamine-unrelated phenotype. Future studies would be of great interest to elucidate these putatively additional, polyamine-independent roles of planarian ODC-like genes.

Methods
Animals, treatments, regeneration and morphometric analysis of blastema size. Planarians used in this work belong to the species Dugesia japonica, asexual strain GI 44 and to an asexual strain of Schmidtea mediterranea (a kind gift of Professor Renata Batistoni, University of Pisa). Additional information on planarian rearing can be found in Supplementary methods. Regenerating fragments were obtained by transection between auricles and pharynx (in the planarian neck) or immediately below the pharynx. For morphometric analysis, animals were cut at the level of the neck, killed 4 days later in 2% hydrochloric acid for 5 min at 4 °C, and then fixed in 100% ethanol. For additional information, please see Supplementary methods. For irradiation experiments, intact planarians were exposed to 30 Gray (Gy) (with uncertainty of ±2%) single-dose of hard X-rays using a 6 MV beam of a VARIAN Medical System accelerator for radiotherapy at the dose rate of 6 Gy/min. Putrescine hydrochloride was purchased by Sigma and diluted at 1 M in distilled water. Putrescine treatment was performed at 0.5 mM final concentration. DFMO was purchased from Cayman Chemicals, diluted at 42 mM in planarian water, and used at a final concentration of 10 mM. DFMO and putrescine media, or clean water for controls, were changed every second day. In some experiments 0.5 mM putrescine or 42 mM DFMO were directly injected in the planarian gut every two days.

Identification of D. japonica and S. mediterranea ODC sequences and in silico analysis. Short
sequences of DjODC1, DjODC2 were retrieved from a DGE library containing differentially expressed TAGs obtained by comparison of a wild-type D. japonica DGE library and a 5 Gy-treated animal DGE library during the repopulation process that occurs after low-dose X-ray treatment (unpublished data). For the analysis, each transcript was compared to nucleotide databases ("nucleotide collection" database and "expressed sequence tags" database) by one-against-all alignments on BLAST (blastn function). DjODC1, DjODC2 sequences met at least one perfect-matching transcript, and this allowed us the expand their nucleotide sequences. In addition, imperfect matches led us to find other nucleotide sequences, similar but not identical to those of DjODC1 and DjODC2. These transcripts, were named DjODC3, DjODC4, DjODC5 and DjODC6, and their sequences are significantly different from each other, at both nucleotide and protein level; their nucleotide sequences were expanded using blastn, as described for DjODC1 and DjODC2. In some cases, nucleotide sequences were fully expanded by comparison with a transcriptome assembled from sequence reads obtained from wild-type D. japonica 45 .
S. mediterranea ODC homologous sequences, were obtained by searching the Schmidtea mediterranea genome database (SmedGD) 25 and the Planmine database 26 .

Analysis of mitosis.
Mitotic analysis was performed using the anti-phosphorylated histone-H3 (H3P) antibody (Upstate), both by Western blot and whole-mount immunofluorescence. For Western blot analysis, proteins were obtained by homogenizing 4 planarians per sample in lysis buffer (Tris-HCl 20 mM pH 7.4, 150 mM NaCl, 1 mM DTT, 5 mM EDTA, 5 mM EGTA, 0.1% TRITON X-100). Two independent samples were processed for each experimental condition. For a detailed protocol, please see Supplementary methods.
For immunofluorescence analysis, planarians were treated the same as for in situ hybridization (see below), and then incubated with 1:500 dilution of rabbit anti-H3P antibody. After several washes, specimens were incubated in alexafluor 488 anti-rabbit secondary antibody, and scanned under a Leica TCS SP8 confocal microscope.  (Table S3). RNA interference (RNAi) was performed by dsRNA injection with the Nanoject Microinjectior (Drummond), or by the feeding procedure described by Rouhana and colleagues 47 . Specific downregulation of the corresponding cognate mRNA, was tested by Real Time RT-PCR 7 days after the first injection or feeding, for each dsRNA molecule. For a detailed protocol, please see Supplementary methods.
In situ hybridization. DNA templates for, Djinnexin1, DjPiwiA, DjNB. 21.11.e, DjAGAT2, DjODC1-6, Smed-odc-5 and 6, DjOAZ and DjCollagen were prepared by RT-PCR from wild-type planarian cDNA, using T7 promoter adapted primers, as described in Table S3. DjSyt DNA template was obtained as described by Salvetti and colleagues 48 . DjInnexin10 DNA template was a kind gift of Professor Batistoni (University of Pisa). Purified amplification products were in vitro transcribed in the presence of DIG-labelling mix (Roche) to obtain digoxigenin (DIG)-labeled RNA probes or in the presence of fluorescein-labelling mix (Roche) to obtain fluorescein-labeled RNA probes. For D. japonica colorimetric in situ hybridization was performed according to Rossi and colleagues 49 . At least ten specimens were tested for each probe. For a detailed protocol please see Supplementary methods. Double fluorescent in situ hybridization in S. mediterranea, was performed according to King and Newmark 50 . Histological analysis. For histological examination of phenotypes, planarians were killed in 2% hydrochloric acid as described above, and fixed in 2% formaldehyde 0.2% saturated aqueous picric acid in 5/8 Holtfreter for 30 minutes at 4 °C. Specimens were then post-fixed in 2% formaldehyde, 0.1% glutaraldehyde, 0.2% saturated aqueous picric acid in 5/8 Holtfreter for 60 minutes at 4 °C. After embedding in paraffin, 6 μm thick slices were stained with Sirius red and fast green. For a detailed protocol, please see Supplementary methods. Slides were analyzed under a Zeiss Axioplan microscope. Five animals were analyzed for each experimental condition. For morphometric analysis of epidermis, 5 sections per animal were analyzed, and the thickness of both ventral and dorsal epidermis was measured in at least 6 different points per section.
Ultrastructural microscopy. Planarians were killed in 2.5% glutaraldehyde in 0,1 M cacodylate buffer pH 7.2 for 2 hours at 4 °C, after washing in cacodylate buffer specimens were post-fixed in 0,1 M osmium tetroxide in cacodylate buffer for 2 h at room temperature. After washing in cacodylate buffer, samples were dehydrated by a graded series of ethanol and then embedded in epoxidic resin. Ultrathin sections were placed on formvar carbon grids, stained with uranyl acetate and lead citrate and analyzed under a Jeol electron microscope. TUNEL analysis. TUNEL assay was performed in S. mediterranea as described by Pellettieri and colleagues 36 and Tu and colleagues 35 . In D. japonica, the assay was performed with a unique modification, namely the pre-treatment of planarians with 2% HCl / 5/8 Holtfreter solution, instead of N-acetylcysteine/PBS solution. Specimens were mounted in 80% glycerol, and scanned under the Leica TCS SP8 confocal microscope. Three planarians were analyzed for each experimental condition. A single composite image was taken for each specimen. The number of apoptotic cells was evaluated using the find maxima option of the Image J software.
Propidium iodide staining and analysis of epidermis nuclear density. Planarians were fixed, bleached, rehydrated as for in situ hybridization and incubated for 30 minutes in 0.5 mg/ml RNAse A in PBS. Specimens were then stained with 2 µg/ml propidium iodide over night at 4 °C. After a quick wash in PBS, specimens were mounted in 80% glycerol and scanned under the Leica TCS SP8 confocal microscope. Three planarians were analyzed for each experimental condition, and composite images of epidermal nuclei were taken for both the head and tail region of each animal. The number of nuclei was counted manually in each picture.
Statistical analysis. For comparison of blastema size a visual inspection was performed to confirm a normal distribution. Unpaired student T-test was applied to evaluate statistical significance (p < 0.05) in each single experiment, considering as matrix 1 the fifteen animals of control and matrix 2 the fifteen animals of each experimental class.
For Real Time RT-PCR, epidermal thickness and epidermal nuclei density, the non-parametric U-Mann-Whitney test was applied to evaluate statistically significant differences (p < 0.05).