High level of CpG methylation in P. dumerilii
In a first attempt to characterize DNA methylation in P. dumerilii, we used a computational approach that allows to evaluate the DNA methylation level and pattern of an organism based on the determination of normalized CpG content (e.g., [41,42,43]). Indeed, methylated CpGs are hypermutable compared to the other dinucleotides [44]. While deamination of non-methylated cytosine can be efficiently repaired, 5mC deamination gives rise to thymines which are less efficiently processed by DNA repair mechanisms [44, 45]. As a consequence, the mutation rate of 5mC into T is much higher than other transitions [46]. In species with high levels of 5mC in CpGs, there is an increase of the mutation rate from CpG to TpG or CpA, which leads, over several generations of germline mutation accumulation, to low contents of CpGs in the genomes of these species [47]. Determining the CpG observed/expected (o/e) ratios can thus be used to estimate 5mC levels: CpG o/e close to 1 means no methylation while CpG o/e far below 1 suggests that methylation of CpGs is present (e.g., [41,42,43]). As in non-vertebrates methylated CpGs are mostly found within gene bodies [11, 12], we calculated CpG o/e for P. dumerilii gene bodies, by applying Notos, a software that computes CpG o/e ratios based on kernel density estimations [43, 48], on a high-quality P. dumerilii reference transcriptome [49]. We found a CpG o/e distribution with a single mode at 0.55 (Fig. 1a), suggesting high-level gene body methylation in P. dumerilii. Indeed, based on a large-scale analysis of 147 species from all major eukaryote lineages, four types of gene body methylation have been defined and P. dumerilii fits into type 3 to which belong species with high gene body methylation, which is the case for most vertebrates [48]. We used the same approach to calculate CpG o/e for additional species used for phylogenetic analyses of methylation machinery proteins (Additional file 2: Table S1; Additional file 3: Fig. S2)—see below for further discussion.
To further assess CpG methylation in the P. dumerilii genome at the experimental level, we performed genomic DNA (gDNA) digestion with the methylation-sensitive enzyme HpaII and its methylation-insensitive isoschizomer MspI, which target CCGG sites [50]. If portions of genomes are methylated, different profiles of restriction fragments are expected from the two enzymatic digestions. To facilitate interpretation of profiles obtained with P. dumerilii gDNA, we included gDNA from species with known methylation patterns in our experiment (Fig. 1b). Drosophila melanogaster do not have 5mC methylation and, as previously reported [51], similar profiles are obtained for both HpaII and MspI enzymatic digestions. The cephalochordate Branchiostoma lanceolatum and the cnidarian Nematostella vectensis have a mosaic pattern of methylation (type 4 in [48]), characterized by the presence of a large number of different cleaved fragments in both digestions and a high molecular weight fraction only found with HpaII [50]. Vertebrates such as Homo sapiens have global CpG methylation in their genome (type 3 in [48]), and accordingly their gDNA is largely resistant to HpaII digestion [50]. An exception are naïve mouse embryonic stem cells (mESC) [52] where similar restriction profiles with both enzymes were observed. In the case of P. dumerilii, we found a restriction pattern that is remarkably similar to that of H. sapiens, further supporting the hypothesis of high levels of CpG methylation in this species (Fig. 1b).
We next performed LUminometric Methylation Assay (LUMA) [53, 54] to obtain a quantitative assessment of CpG methylation and information about its dynamics during Platynereis’s life cycle. LUMA is an efficient method to measure global CpG methylation, based on gDNA digestion (at CCGG sites) by methylation-sensitive restriction enzymes followed by pyrosequencing. LUMA was performed on P. dumerilii gDNA extracted from ten different stages (Fig. 1c). Very high and similar methylation levels were found during embryonic/larval development (from 12 to 72 h post-fertilization, hpf; about 80% of CCGG sites are methylated; Additional file 2: Table S2). This level significantly decreases after the end of larval development, as shown in juvenile worms (early stages of post-larval growth; 4, 5, and 15 days post-fertilization, dpf), but nevertheless remains quite high (about 60–65% of methylated CCGG sites). The methylation level further decreases in older juvenile worms (3 months post-fertilization, mpf; about 27–32%) and subsequently increases when worms become sexually mature, significantly more in males (about 67-70%) than in females (about 43–56%).
To confirm the existence of gene body methylation in P. dumerilii, we performed bisulfite pyrosequencing [55] on CpG-rich parts of the coding region of two different genes, Pdum-histone H4 and Pdum-14-3-3-like. These two genes were selected because they display stretches of CpGs in their coding region (7 and 16 CpGs for Pdum-histone H4 and Pdum-14-3-3-like, respectively). Additionally, orthologs of these genes in other lophotrochozoan species were shown to have gene body methylation [28, 30]. Using DNA extracted from 72hpf larvae, we found high levels of methylation (between 65 and 87%) for 6 of the 7 CpGs of Pdum-histone H4, and low levels for all CpGs of Pdum-14-3-3-like (< 10%; Fig. 1d). These data therefore indicate that gene body methylation does indeed occur in P. dumerilii and that the level of methylation strongly differs in the two studied genes. In contrast, the level of methylation in the coding region of these two genes remains almost constant throughout the life cycle of the worm, as shown by bisulfite pyrosequencing using DNA extracted from five additional stages (Additional file 2: Table S3).
Taken together, these data indicate high-level CpG methylation in the P. dumerilii genome. In addition, the 5mC level is dynamic along the P. dumerilii life cycle and is significantly higher during embryonic/larval development as compared to post-larval stages. Striking changes in methylation level also occur during post-larval growth and when the worms reach sexual maturity. We also obtained evidence for gene body methylation in P. dumerilii and found that the level of CpG methylation in gene bodies is not uniform from one gene to another.
P. dumerilii possesses a full ancestral-like DNA methylation and NuRD toolkit
Having established the existence of 5mC in P. dumerilii, we next aimed to identify proteins involved in writing, modifying, and reading this epigenetic mark, as well as putative NuRD components, in this species. For that purpose, we searched for P. dumerilii orthologs of proteins known to exert these functions in mammals (Additional file 1: Fig. S1) through a sequence-similarity approach using reciprocal best BLAST with Homo sapiens and Mus musculus sequences as queries. We found putative P. dumerilii orthologs for all investigated proteins/protein families (Additional file 4: Fig. S3). Sequences of all the identified proteins can be found in Additional File 5. As 5mC and NuRD proteins are often characterized by the presence of particular domains or association of domains, we searched for conserved domains present in the retrieved P. dumerilii proteins. In most cases, we found domains that are consistent with orthology relationships inferred from BLAST analyses (Additional file 4: Fig. S3).
Since defining orthology relationships only on BLAST analyses can be misleading, in particular when numerous paralogs are present, we turned to phylogenetic analyses to ascertain these relationships. To perform these analyses on a firm basis and to get insight into the evolution of the DNA methylation and NuRD toolkit in animals, we retrieved, by reciprocal BLAST searches using mouse and human sequences as queries, putative orthologs from 51 additional species from diverse animal phylogenetic groups. We ended up with a sample of 54 species from all major animal lineages (Fig. 2). Maximum likelihood (ML) trees were constructed for each protein family and are shown in Additional file 6: Fig. S4. We also searched for members of the different families in species from choanoflagellates, the sister group to animals, and used, when possible, these sequences as outgroups to root the phylogenetic trees. These phylogenetic trees allow us to confirm orthology relationships for all P. dumerilii proteins and to define the number of members of all protein families in the 54 investigated animal species (Fig. 2). We summarized the number of members for each defined family and, based on parsimony, we inferred the presence or absence of each protein family in the last common ancestors of animals and bilaterians (Fig. 3). All the identified proteins are listed in Additional file 2: Table S4 and their sequence can be found in Additional file 5.
The P. dumerilii genome encodes three Dnmt proteins that can be clearly assigned to the Dnmt 1, 2, and 3 subclasses (Additional file 6: Fig. S4A). The presence of these three subfamilies appears to be ancestral to animals (Fig. 3), as these three subfamilies are found in most non-bilaterians and in many species in the three bilaterian evolutionary lineages (Fig. 2). While only dnmt2 genes were found in choanoflagellates, dnmt1 and dnmt3 genes have been reported in other eukaryotic groups, suggesting an early diversification of the Dnmt family during the evolution of eukaryotes [9]. Only a few gene duplications occurred for dnmt1 (in particular in some arthropod species) and for dnmt3 (in particular during vertebrate evolution in agreement with published studies; e.g., [56]). dnmt gene losses occurred in some species or lineages such as nematodes, rotifers, tardigrades, placozoans, and platyhelminthes. Absence of both dnmt1 and dnmt3 is correlated to the absence or very low abundance of cytosine DNA methylation as shown by CpG o/e ratio calculation (Fig. 2). P. dumerilii also possesses single tet, tdg, and uhrf genes, which likely corresponds to the ancestral situation in animals (Figs. 2 and 3; Additional file 6: Fig. S4B-D). Duplications of tet genes are infrequent, but two duplications nevertheless occurred in vertebrates [56]. tet genes are only absent in species that lack 5mC methylation and dnmt1 and dnmt3, with the exception of dipterans which possess one tet gene. tdg is present in almost all investigated species in one copy, as expected for a gene involved in DNA repair. uhrf has a similar distribution to tet, being absent in species lacking 5mC, but in this case including dipterans. A single uhrf gene is found in most other species, with the notable exception of euteleostomes (bony vertebrates) that possess two genes (Fig. 2).
Phylogenetic analysis shows the existence of two large groups of Mbd proteins, one which contains Mbd1, Mbd2, and Mbd3 proteins from vertebrates (hereafter named Mbd1/2/3 group) and the other which contains vertebrate Mbd4 and MeCP2 proteins (Mbd4 group; Additional file 6: Fig. S4E). P. dumerilii’s genome encodes two Mbd proteins, one belonging to the Mbd1/2/3 group (putative NuRD component) and the other to the Mbd4 group. Presence of both Mbd1/2/3 and Mbd4 is observed in many species belonging to most animal lineages, including non-bilaterians such as sponges and cnidarians, strongly suggesting that the last common ancestor of animals possessed at least two mbd genes (Fig. 3). mbd gene losses occurred in few species, mainly in those that also lack cytosine DNA methylation. A few gene duplications also occurred, in particular in vertebrates in which both mbd1/2/3 and mbd4 ancestral genes underwent gene duplications.
The phylogenetic tree of Chd proteins comprises three large groups: one that includes vertebrate Chd3/4/5 proteins (hereafter named Chd3/4/5 group), the second vertebrate Chd1/2 (Chd1/2 group), and the third vertebrate Chd6/7/8/9 (Chd6/7/8/9 group; Additional file 6: Fig. S4F). Six chd genes have been found in P. dumerilii, one belonging to the Chd1/2 group, one to the Chd6/7/8/9 group, and four to the Chd3/4/5 group (putative NuRD components). Members of these three groups are found in almost all studied species, including non-bilaterians, indicating that presence of three different types of CHD proteins is ancestral to animals (Fig. 3). Only very few gene losses occurred. Gene duplications are more frequent, in particular in vertebrates and lophotrochozoans, including in annelids in which two to four members are found in the three studied species (Fig. 3).
Previous phylogenetic studies classified Hdac proteins into four classes (I, IIA/B, III, and IV) [57, 58]. Here we focused on class I to which belong Hdac1, Hdac2, Hdac3, and Hdac8, genes that encode members of the NuRD complex. Phylogenetic analysis showed the existence of three subgroups, Hdac1/2, Hdac3, and Hdac8 (Additional file 6: Fig. S4G). We found one member of each subgroup in P. dumerilii, as well as in almost all other investigated species, indicating that at least three class I hdac genes were present in the last common ancestor of all animals (Fig. 3). We found only very few gene losses (e.g., in nematodes). Duplications mainly occurred in arthropods and vertebrates. Single rbbp4/7, mta1/2/3, and gatad2 genes are found in P. dumerilii (Additional file 6: Fig. S4H-J). At least one member of each of these subfamilies is found in all studied species, indicating that their presence is ancestral to animals (Fig. 3). Gene duplications occurred in vertebrates, ecdysozoans, and lophotrochozoans.
In conclusion, we have identified in P. dumerilii a complete set of writers, modifiers, and readers involved in 5mC methylation, as well as putative NuRD components. We additionally provide an animal-wide view of the evolution of the corresponding gene families (Fig. 2), which suggests that the last common ancestor of animals already possessed a complex repertoire of 5mC and NuRD toolkit genes (Fig. 3). Our analysis also indicates that the P. dumerilii repertoire is mostly composed of single-copy genes and likely close to the one present in the last common ancestor of bilaterians.
DNA methylation and NuRD toolkit genes are dynamically expressed during development and regeneration in P. dumerilii
We next aimed to characterize the expression of DNA methylation and NuRD genes in P. dumerilii. We first took advantage of two previously published transcriptomic datasets corresponding to various developmental stages and adult conditions of P. dumerilii, available in a public database (PdumBase) [59]. The first dataset corresponds to embryonic developmental stages, ranging from 2 to 14hpf, with a time point every 2 h [49]. The second dataset comprises major larval stages (24 to 4dpf; five time points), juvenile stages (10dpf to 3mpf; five time points), and adult reproductive stages (males and females) [60]. Altogether, expression data for a total of 19 stages during embryonic and post-embryonic development, as well as male and female adult stages, are available.
Expression values for most genes (exceptions are Pdum-dnmt3 absent in the two sets of transcriptomic data and Pdum-gatad2 and Pdum-rbbp4/7 only found as chimeric transcripts) were recovered and can be found in Additional file 7: Fig. S5. High transcript levels are found for many studied genes in the earliest developmental stages (2–6hpf) and several of them belong to co-expression clusters defined by Chou et al. [49] as maternal gene clusters (clusters 1–4; Additional file 7: Fig. S5). This indicates that the P. dumerilii egg contains a large pool of maternal transcripts coding for DNA methylation proteins that could be used for embryonic development. To further analyze these expression data, we studied changes in expression during the main steps of P. dumerilii life cycle (Fig. 4). From 2hpf to 14hpf, a decrease in quantity of transcripts of about half of the genes, including genes coding for DNA methylation maintenance (Pdum-dnmt1 and Pdum-uhrf), as well as putative members of the NuRD complex (Pdum-chd3/4/5A-B and Pdum-hdac8), is observed. From 24hpf to 4dpf, this decrease is found for most genes, including Pdum-dnmt1 and Pdum-uhrf. In contrast, expression of Pdum-tet and Pdum-tdg is increased or stable, respectively. This is consistent with the decrease of the CCGG methylation level that we observed at the end of larval development (Fig. 1c). From 4dpf to 3mpf, a majority of genes have stable expression with the exception of the upregulation of every chd gene except chd3/4/5B, which is downregulated (Fig. 4). Transition from 3mpf to the adult stage is strikingly gender-specific: in males, most genes have stable or downregulated expression, while about 80% of the genes are strongly upregulated in females (Fig. 4; Additional file 7: Fig. S5), suggesting different occurrence and importance of DNA methylation during sexual maturation and gamete production between males and females.
We next studied the expression of DNA methylation and NuRD genes during P. dumerilii posterior regeneration. To characterize in which part(s) and tissue(s) of the regenerated region these genes are expressed, we performed whole-mount RNA in situ hybridizations (WMISH) at all five stages of posterior regeneration (a schematic representation of regeneration stages can be found in Additional file 8: Fig. S6) [37], focusing on a set of ten genes that encode putative writers/modifiers/readers of 5mC or NuRD components. Representative expression patterns are shown in Fig. 5 and Additional file 9: Fig. S7. A schematic representation of the expression patterns can be found in Additional file 10: Fig. S8. We also tried to define the expression of the studied genes in non-amputated worms (to compare to the expression during regeneration) but failed to obtain any signal above the background level with our WMISH protocol, likely due to the presence of a thick cuticle around the fully differentiated segments of these worms [36]. As a proxy of non-amputated worms, we therefore used worms that have regenerated for 15 days (15 days post-amputation, dpa) and which show many well-differentiated segments lacking the thick cuticle that hampers WMISH in non-amputated worms. Representative expression patterns of the studied genes in these worms are shown in Additional file 11: Fig. S9. We were also able to detect the expression of some genes in worms fixed immediately after amputation (hereafter named stage 0), as the wound probably favors the penetration of the probes used for WMISH. Only very weak expression was however observed for most studied genes in stage 0 worms (Additional file 12: Fig. S10), and these expressions will not be further discussed.
Pdum-dnmt1 and Pdum-dnmt3 are weakly expressed in the wound epithelium at stage 1 (Fig. 5a1, b1). At stage 2, Pdum-dnmt1 is strongly expressed in two internal groups of cells and in the lateral ectoderm (Fig. 5a2). Its expression extends in almost the whole blastema at stage 3 and is found in the regenerated growth zone (Fig. 5a3). At the same stages, Pdum-dnmt3 is very weakly expressed in both mesodermal and ectodermal cells of the regenerated region (Fig. 5b2, b3). At stages 4 and 5, Pdum-dnmt1 is expressed in the mesoderm of the developing segments, mesodermal and ectodermal growth zone, at the base of the anal cirri, and weakly in the lateral/dorsal ectoderm (Fig. 5a4, a5; Additional file 9: Fig. S7A). Pdum-dnmt3 is expressed in the ventral ectoderm and at the base of the anal cirri (Fig. 5b4, b5). Broad and diffuse expression in the developing segments was observed for both genes in worms at 15dpa (Additional file 11: Fig. S9A, B). Pdum-tet expression is not reliably detected at stage 1 (Fig. 5c1). At stages 2 and 3, a weak expression is found in both internal and superficial blastemal cells (Fig. 5c2, c3), which continues at stages 4 and 5 and at which expression is also observed at the base of anal cirri (Fig. 5c4, c5). At 15dpa, Pdum-tet expression is found in both the mesoderm and ectoderm of the developing segments (Additional file 11: Fig. S9C). Pdum-tdg is strongly expressed at stage 1 in the wound epithelium and internal cells of the segment abutting the amputation plane (Fig. 5d1). At stages 2 and 3, it is broadly expressed in the whole blastema (Fig. 5d2, d3). From stage 3, Pdum-tdg is expressed in mesoderm and ectoderm of the developing segments, and in ectodermal and mesodermal growth zones, as well as weakly at the base of the anal cirri (Fig. 5d4, d5; Additional file 9: Fig. S7B). Pdum-tdg is expressed in the ventral ectoderm and in the developing parapodia at 15dpa (Additional file 11: Fig. S9D).
Pdum-mbd1/2/3, Pdum-hdac3, and Pdum-hdac8 have roughly similar expression during posterior regeneration, Pdum-hdac3 being expressed at most stages weaker than the two other genes. At stage 1, an expression is detected in two lateral patches of cells in and close to the wound epithelium (Fig. 5e1, f1, g1). The three genes are widely expressed in the blastema at stages 2 and 3 (Fig. 5e2, e3, f2, f3, g2, g3). Expression in the mesoderm and ectoderm of the developing segments, growth zone, and at the base of the anal cirri is observed at stages 4 and 5 (Fig. 5e4, e5, f4, f5, g4, g5; Additional file 9: Fig. S7C, D). Broad and diffuse expression patterns in the developing segments were observed for the three genes at 15dpa (Additional file 11: Fig. S9E-G). Pdum-chd3/4/5B expression is found at stage 1 in four small patches of internal cells close (but not adjacent) to the wound epithelium, two located ventrally and two dorsally (Fig. 5h1; Additional file 9: Fig. S7E). From 2dpa, we observed an intense expression in the mesodermal part of the regenerated region, including the mesodermal growth zone (Fig. 5h2–h5; Additional file 11: Fig. S9H). Pdum-chd1/2 and chd6/7/8/9 are expressed in cells in and close to the wound epithelium at stage 1, the latter having a much broader expression (Fig. 5i1, j1). At stage 2, both genes are expressed in superficial and internal cells of the regenerated region, in most or all cells for Pdum-chd6/7/8/9 but only in a few cells for Pdum-chd1/2 (Fig. 5i2, j2). Broad expression in mesodermal cells, including the growth zone, is observed at later stages (Fig. 5i3–i5, j3–j5). At stage 5, Pdum-chd1/2 is also weakly expressed in the ectodermal growth zone (Additional file 9: Fig. S7F). At 15dpa, Pdum-chd6/7/8/9 is weakly expressed in the developing segments (Additional file 11: Fig. S9I). We failed to detect significant expression of Pdum-chd1/2 at 15dpa.
Altogether, we found that P. dumerilii DNA methylation and NuRD genes are dynamically expressed during embryonic, larval, and post-larval development, as well as during sexual maturation and regeneration. During this latter process, most genes are expressed from its earliest stages and their expression is later mostly found in blastemal cells, putative mesodermal and ectodermal stem cells of the growth zone, and cells of the developing segments (Additional file 10: Fig. S8). Observed patterns of expression show striking similarities with those previously reported for proliferation (cycB and pcna genes) and stem cell genes (e.g., piwi, vasa, nanos, and myc genes) [37], suggesting that DNA methylation and NuRD genes are mainly expressed in undifferentiated proliferating cells, including stem cells of the regenerated posterior growth zone.
Decitabine reduces DNA methylation and impairs development, regeneration, and post-regenerative posterior growth in P. dumerilii
To test a possible role of DNA methylation during P. dumerilii regeneration, we tried to reduce 5mC levels using two well-known and widely used hypomethylating agents: Decitabine (5-aza-2′-deoxycytidine) and RG108 (N-Phthalyl-L-Tryptophan) [61,62,63]. Decitabine is incorporated in DNA and binds Dnmt1 irreversibly, leading to a progressive loss of DNA methylation through cell divisions. RG108 is a specific non-nucleoside inhibitor of Dnmt1, which acts by binding in a reversible manner to the active center of the enzyme. As these two drugs have never been used in P. dumerilii, we first tested their activity by treating larvae continuously from 1 to 3dpf with Decitabine or RG108 (Fig. 6a). Neither drugs caused significant lethality during treatment. DNA was extracted from larvae at 3dpf and CCGG methylation level measured using LUMA (Fig. 6b): Decitabine treatment leads to a 2.5-fold decrease of CCGG methylation (from 81.5 to 32.4%) while no significant effects were found for RG108. We also checked for morphological defects (Fig. 6c): larvae were observed either immediately after treatment (at 3dpf) or after washing out the drug and putting larvae in normal sea water until 5 or 14dpf. Larvae that had been treated with Decitabine presented morphological abnormalities at 3dpf, in particular reduced parapodia (worm appendages) bearing very few chaetae (extracellular chitinous structures) and a reduced pygidium (Fig. 6c). While abnormal, these larvae were alive and survived for a few more days. All animals did however die in the following days, possibly because of feeding defect (during this period normal young worms start to eat, dead Decitabine-treated worms consistently showed an empty gut). In contrast, RG108 treatment did not affect larval morphology (Fig. 6c). We therefore conclude that Decitabine can affect DNA methylation levels and larval development in P. dumerilii.
To investigate potential consequences of a decrease of DNA methylation on regeneration, we treated worms with three concentrations of Decitabine (10 μM, 50 μM, and 100 μM) immediately after amputation for 5 days and scored the worms every day for the stage that had been reached based on a previously established staging system (Additional file 8: Fig. S6) [37]. We found a small number of deaths at 10 μM and 50 μM concentrations, while a 100 μM concentration appears to be much more harmful to worms (Additional file 2: Table S5). Some worms also underwent spontaneous amputation of their posterior part (autotomy) at some time points (Additional file 2: Table S5). These worms were excluded from the analysis. We found that Decitabine significantly delayed regeneration as compared to controls (DMSO 0.5 % and sea water), in a concentration-dependent manner (Fig. 7a). At 5 days post-amputation (dpa), while most control worms reached stage 4 or more, worms treated with Decitabine were mostly at stages 2 to 3 (Fig. 7b). No major abnormalities were observed at the morphological level in Decitabine-treated worms (not shown). To better understand how regeneration proceeds in the presence of Decitabine, we did Decitabine treatments (at 50 μM as this concentration shows low toxicity and pronounced effect on regeneration) from 0dpa to 5dpa, fixed treated worms at 5dpa, and performed WMISH for some of the genes whose expression was previously studied during normal regeneration [37]. The analyzed genes showed expression at 5dpa in Decitabine-treated worms that are similar to those of stage 2 or 3 in non-treated worms [37], indicating that regeneration is blocked in the presence of Decitabine (Fig. 7c). Abnormal expression patterns, never observed in non-treated animals, were nevertheless found in some treated worms for Pdum-hox3 (growth zone marker; extended and/or mis-located expression domain), Pdum-piwiB (stem cell marker; no or reduced expression), and Pdum-engrailed (segment marker; incomplete expression stripes).
It has been shown that mammalian cells treated with Decitabine only partially recover their initial methylation level, leading to an epigenetic “imprint” of drug exposure [64]. We hypothesized that Decitabine treatment could have long-term impacts in P. dumerilii and affect segment formation that follows regeneration (post-regenerative posterior growth [36, 37]). To test this hypothesis, we treated worms with Decitabine from 0 to 5dpa, then washed out the drug, put worms in normal sea water until 25dpa, checking their morphology and counting the number of segments that have been produced at six time points (Fig. 8a). As for the previous experiment, Decitabine treatments induced few worm deaths and autotomies (Additional file 2: Table S6). Most Decitabine worms recovered from the treatment and were able to reach stage 5 and undergo posterior growth (Fig. 8b). Decitabine-treated worms continued to be delayed as compared to control worms, had a reduced number of newly added segments at 25dpa (treated worms had about 4 to 6 segments compared to about 10 to 12 segments for controls), and showed morphological abnormalities (Fig. 8b, c). A reduced number of newly added segments was due not only to a marked delay during regeneration, but also to a reduced rate of segment addition after the drug had been washed out (Additional file 13: Fig. S11A). A high variability was observed among Decitabine-treated animals compared to controls (Additional file 13: Fig. S11B-F). We defined three classes of animals based on their morphology and the number of newly added segments at 25dpa (Fig. 8c; Additional file 14: Fig. S12). Class 1 animals (30.6% of Decitabine-treated worms) show a characteristic bottleneck-like shape with a marked constriction between non-regenerated and regenerated regions, no or few newly added segments, and no or abnormal anal cirri. These worms are prone to undergo autotomy. Class 2 animals (61.6%) have an abnormal body shape, a reduced number of newly added segments, an absence of well-differentiated parapodia on newly added segments, and no or abnormal anal cirri. Class 3 worms (7.9%) have a morphology and number of newly added segments similar to control animals.
Taken together, these observations indicate that Decitabine treatments during regeneration have long-term effects and affect subsequent post-regenerative posterior growth, possibly by affecting growth zone regeneration. Some Decitabine-treated worms were however able to add new segments in an almost normal manner, which led us to hypothesize that the growth zone was not impacted similarly in all animals. To point out a potential link between regeneration of the growth zone and ability to later add segments, we performed a multiple correlation analysis (Additional file 15: Fig. S13). In control worms, as expected, only positive correlations were observed, which means that, for example, worms with numerous segments at 20dpa already had a high number of segments at 11dpa. In contrast, in Decitabine-treated worms, while there were positive correlations for closely related days of scoring (for example: 2 to 3dpa, 3 to 4dpa, …), negative correlations were also found and suggested that treated worms that regenerated faster eventually produced less segments. It has been shown that the growth zone is regenerated and becomes functional, producing news segments, at about 3dpa [37]. Our interpretation is therefore that worms with high regeneration scores (scored at stage 3 or more) at 5dpa regenerated a dysfunctional growth zone in the presence of Decitabine, which later led to a null or reduced production of segments, the few segments produced additionally displaying morphological abnormalities. Worms with low scores (less than 3) probably did not regenerate their growth zone during the Decitabine treatment period and did it after 5dpa in the absence of the drug, which led to the formation of a functional growth zone and therefore to normal segment addition. Our data therefore suggest that Decitabine affects the functionality of the growth zone. Consistently, expression of growth zone, stem cell, and segment markers (Pdum-hox3, Pdum-piwiB and Pdum-engrailed, respectively) is affected in some Decitabine-treated worms (Fig. 7c).
As described above, worms treated with Decitabine from 0 to 5dpa show morphological abnormalities at 25dpa. To better understand these alterations, we performed WMISH on Decitabine-treated worms at 25dpa for a set of previously studied marker genes [37]. A wide range of abnormal expression patterns were found in Decitabine-treated worms (Fig. 9). This includes a reduced number of segmental stripes of Pdum-engrailed in worms with few morphologically visible segments (Fig. 9a1–a3), reduced expression of Pdum-dlx (which is normally expressed at the base of anal cirri and in parapodia), on one side of the worm (Fig. 9b1-b3), as well as ectopic expression of Pdum-hox3, Pdum-cdx, and Pdum-piwiB in developing segments (Fig. 9c1–e3), which are consistent with persistent defects in growth zone functionality.
Finally, we investigated whether Decitabine might have effects over an even longer time period. We treated worms with Decitabine from 0 to 5dpa, then put them in normal sea water until 25dpa, performed a second amputation one segment anterior to the first amputation plan (meaning that we eliminated the regenerated region plus one segment), and scored these worms at several time points until 18 days post-second amputation (18dpSa; Additional file 16: Fig. S14A). Control and Decitabine-treated worms regenerated properly and similarly after this second amputation and were able to add new segments at a similar rate (Additional file 16: Fig. S14B). A slight but significant delay, however, was observed for worms treated with 10 μM Decitabine at 18dpSa (Additional file 16: Fig. S14B) and about 10–15% of Decitabine-treated worms showed minor defects at the level of their parapodia (Additional file 16: Fig. S14C). A same proportion of worms that were class 1 or 2 at 25dpa showed abnormalities after the second amputation at 18dpSa (Additional file 16: Fig. S14C). Multiple correlation analysis showed that, for both control and Decitabine-treated worms, only positive correlations were found (Additional file 16: Fig. S14D, E). Therefore, Decitabine treatments after a first amputation have only very minor effects on regeneration and post-regenerative posterior growth occurring after a second amputation.
On the whole, our data show that Decitabine decreases methylation levels in P. dumerilii and affects larval development and regeneration. During regeneration, it impairs post-regenerative posterior growth occurring in the absence of the drug, a long-term effect that could be due to defects in the regeneration of the stem cell-containing growth zone.