Variations in cell plasticity and proliferation underlie distinct modes of regeneration along the antero-posterior axis in the annelid Platynereis

Regeneration, the ability to reform a lost body part upon injury, is an essential process in animals. Its importance is illustrated by its wide deployment in metazoans, although the range of tissues that can be regenerated is highly variable from one species to another (Bely and Nyberg, 2010). Whereas mammals can, at best, regenerate an organ, many other species can perform ‘extensive’ regeneration, such as the reformation of a limb (e.g. salamanders), a large amputated part of their body axis (e.g. annelids) or even their whole body from a small fragment of tissue (e.g. cnidarians and planarians) (Bideau et al., 2021). Despite this diversity, all regeneration processes go through three common steps: the formation of a wound epithelium enclosing the area of the injury, followed by the recruitment of progenitors at the wound site, which often form a blastema (a mass of proliferative undifferentiated mesenchymal cells), and, finally, the growth of the blastema by cell proliferation and differentiation during a morphogenesis step (Galliot and Ghila, 2010; Tiozzo and Copley, 2015).

Uncovering the origin and fate of the cells contributing to blastema formation has been one of the greatest challenges in the field of regenerative biology for decades (Tanaka and Reddien, 2011). Studies of major regeneration models has established that the blastema can be formed by the progeny of activated progenitor or stem cells, as exemplified by the planarian Schmidtea mediterranea, the regeneration of which is sustained by adult stem cells called neoblasts (Wenemoser and Reddien, 2010). As pluripotent cells (at least part of them), neoblasts participate in the formation of all missing tissues (Wagner et al., 2011). Alternatively, the blastema can be formed by post-mitotic cells that dedifferentiate and re-enter cell cycle upon injury, as in the case in urodele limb regeneration (Stocum and Cameron, 2011). In this regenerative process, various local tissues close to the wound dedifferentiate into strictly lineage-restricted progenitors (Flowers et al., 2017; Kragl et al., 2009).

As such, two opposite models have been broadly defined: the first involves very highly plastic cells that migrate to the wound; the second involves local tissues that dedifferentiate to constitute a pool of diverse progenitors with low plasticity. However, the mechanisms of regeneration are often more complex and, in many species and contexts, both dedifferentiated cells as well as tissue-specific resident stem cells contribute to the blastema (e.g. axolotl limb regeneration; Lin et al., 2021; Sandoval-Guzmán et al., 2014).

Careful studies identifying the sources of blastema cells during regeneration are available for a few key regenerative model species but are lacking for the majority of regeneration-competent lineages. These models do not include representatives of the annelid phylum, which display substantial and diverse regenerative capacities. The majority of annelids can indeed regenerate their posterior and/or anterior parts upon amputation (Özpolat and Bely, 2016). Those regeneration processes have been studied for a long time and in various species (reviewed by Bely, 2006) but the precise cellular mechanisms involved in the formation of the blastema remain unclear.

The annelid Platynereis dumerilii is emerging as a useful and relevant model to address fundamental regeneration questions (Schenkelaars and Gazave, 2021; Özpolat et al., 2021). This marine worm has the ability, after embryonic development, to grow continuously by the addition of new segments in its posterior part, through a process of posterior elongation that relies on putative stem cells localized in a subterminal growth zone (Gazave et al., 2013). Importantly, Platynereis has the astounding ability to regenerate its complex posterior end after amputation via the formation of a blastema and through stereotyped steps that we have previously defined (Planques et al., 2019). Upon amputation, the wound heals in 24 h (stage 1 at 1 day post amputation or 1 dpa). One day later (stage 2, 2 dpa), a small blastema forms that rapidly grows. Between stages 2 and 3 (3 dpa), the blastema cells start to differentiate into various tissues (e.g. muscles, nervous system, etc.). At stage 3, the growth zone and the pygidium, which is the posterior-most part of the worm body, begin to reform. At stage 5 (5 dpa), a new fully functional growth zone is re-established and allows posterior elongation to resume (Planques et al., 2019).

The cellular and molecular mechanisms controlling this regeneration process are still largely unresolved. In our first study on this topic, we showed, through S-phase cell labelling coupled with proliferation inhibition experiments, that cell proliferation is absolutely necessary from stage 2 onwards for regeneration to be properly achieved (Planques et al., 2019). Additionally, pulse-chase experiments suggested that most of the blastema cells have a local origin, from the segment abutting the amputation plane (Planques et al., 2019). The origin(s), plasticity and fate of the cells contributing to the regeneration blastema remains to be precisely determined.

In this study, we developed tools to track proliferative cells as well as gut epithelial cells. We determined that some gut cells from differentiated tissues constitute a population of progenitors uniquely contributing to the regeneration of gut epithelial cells. Strikingly, we also showed that, upon posteriorization (i.e. by giving a posterior identity to previously anterior tissues), those lineage-restricted gut progenitors reveal a higher plasticity as they are also able to produce ecto/mesodermal derivatives through regeneration. However, they do not contribute to the regeneration of the growth zone stem cells, which are from local origin (i.e. from the segment abutting the amputation). These results argue for the existence of different degrees of plasticity, through regeneration, of intestinal progenitors along the antero-posterior axis of the animal.

Previous results have shown that cells proliferating in the context of normal continuous posterior growth may participate in the posterior regeneration in Platynereis (Planques et al., 2019). To further study the role of those cells during regeneration, we first examined the distribution of S-phase cells in non-amputated juvenile worms with EdU labelling. We exposed uninjured worms to EdU, either for 5 or 48 h. An exposure of 5 h labels a pool of rapidly cycling cells (hereafter ‘5 h EdU⁺ cells’). In contrast, the 48 h incubation labels a larger pool of cells including both rapidly and slowly cycling cells (hereafter ‘48 h EdU⁺ cells’). Such slowly cycling cells have been shown to contribute to regeneration in several species, including mammals (Karmakar et al., 2020; Koren et al., 2022) and planarians (Molinaro et al., 2021). The distributions of EdU⁺ cells were examined immediately after EdU exposure both in whole samples (Fig. 1A-B′) and in histological sections (Fig. 1C-D″). Three segments were examined, including the most posterior distinguishable segment (called segment 1 or S1) (Fig. 1A′,B′,C″,D″) as well as the sixth and seventh segments counted from the posterior region (called segments 6 and 7 or S6 and S7, respectively) (Fig. 1A,B,C-D′). We observed many EdU⁺ cells in S1 regardless of the EdU incubation time (Fig. 1A′,B′). S6 or S7 exhibit fewer labelled cells, especially the 5 h EdU⁺ cells (Fig. 1A,B). We next looked more carefully at the localization of EdU⁺ cells within the tissues by performing semi-thin sections of S6 and/or S7 (transversal and longitudinal sections, Fig. 1C-D′) and S1 (longitudinal sections, Fig. 1C″,D″). 5 h EdU⁺ cells in S6/S7 mainly delineate the central and circular epithelium of what appears to be the gut (Fig. 1C,C′) (Žídek et al., 2018; Dahlitz et al., 2023, Fig. S1). However, in S1 the 5 h EdU⁺ cells were located not only in the gut but also in ectoderm and mesoderm (Fig. 1C″). In contrast, 48 h EdU⁺ cells were found in all trunk tissue subtypes in S1, S6 and S7 (Fig. 1D-D″).

To better assess those different populations of proliferating cells, we performed cell cycle kinetics analyses. To determine whether the proliferative cells behave differently as a function of their location along the antero-posterior axis of the animal or depending on their specific tissue, we implemented a cumulative EdU labelling assay (Nowakowski et al., 1989). This method relies on the sequential administration of a thymidine analogue, EdU in this study, until all proliferative cells are labelled, allowing one to determine the total cell cycle length (Tc), S-phase length (Ts) and growth fraction (GF or proportion of cycling cells). We applied it to calculate such cell cycle parameters in different segments (S1, S6 and S7) and in different tissues (gut versus the other trunk tissues) (Fig. 1E-G; Tables S1 and S2). We found that, regardless of the tissue examined, many more cells are cycling in S1, compared with S6 and S7. Regarding the gut, the GF of 38% in S1 decreases to 29% and 28% in S6 and S7, respectively. For the rest of the trunk tissues, 63% of cells are cycling in S1, while this rate drops to 18% and 15% in S6 and S7, respectively (Fig. 1G; Tables S1 and S2). We also determined that the gut cells, whatever their locations, are cycling more quickly than the other trunk tissue cells. Indeed, the cell cycle length (Tc) for the S1 gut cells is 27 h whereas the other cells have a Tc of 54 h. Similarly, in S6 and S7, the gut cell Tc values are 50 h and 49 h, respectively, versus 62 h and 66 h for the other cells in the trunk (Fig. 1G; Tables S1 and S2). In addition, the S-phase length (Ts) is broadly similar between the gut cells and other cell types at positions S6 and S7 (15 h versus 14 h and 16 h versus 19 h). In contrast, for S1, Ts is drastically shortened for the gut cells (3 h versus 20 h) (Fig. 1G; Tables S1 and S2).

In summary, the proportion of cycling cells (or GF) varies along the antero-posterior (AP) body axis of the uninjured worms, with more proliferative cells in posterior segment that in anterior segments (Fig. 1A,A′,B,B′,G). In addition, the proportion of cycling cells varies according to the tissue type, with the anterior gut cells remaining highly proliferative with a faster cell cycle, compared with the other tissues.

Next, we aimed to determine the respective contribution of these populations of cycling cells to posterior regeneration. We incubated worms with EdU for 5 h or 48 h, as previously, and amputated them at two different positions (Fig. 2A), called amputation ‘0’ (if the cut was made immediately behind S1, removing only the pygidium and the growth zone) or amputation ‘−5’ (if the cut was made immediately behind S6, removing five recognizable segments in addition to the terminal part). After 5 days of regeneration, the patterns of EdU⁺ cells in the regenerating parts were determined (Fig. 2B-E′). After a 5h-long EdU pulse followed by an amputation ‘−5’, only 9% of the blastema cells were EdU⁺ (Fig. 2F; Table S1) and they appeared to be located mostly within the regenerating gut (Fig. 2B,B′). After a 5h-long EdU pulse followed by an amputation ‘0’, 19% of the blastema cells were EdU⁺ (Fig. 2F; Table S1) and they were located not only in the gut but also in mesodermal and ectodermal tissues (Fig. 2D,D′). In contrast, after a 48 h-long EdU pulse, more than 70% of the blastema cells were EdU⁺, regardless where the amputation was performed (amputation ‘−5’ or ‘0’) (Fig. 2F; Table S1) and located in all tissue types (Fig. 2C,C′,E,E′). These results indicate that the cells contributing to regeneration, whether from the anterior or posterior parts of the worms, are predominantly those that had been cycling before amputation.

Interestingly, almost all the blastema cells originate from 48 h EdU⁺ cells in both anterior (S6/S7) and posterior (S1) segments. 5 h EdU⁺ cells also contribute to a far lesser extent. In the posterior-most segment (S1), 5 h EdU⁺ cells produce derivatives in all cell compartments, whereas in anterior segments (S6/S7) they contribute mostly to the regeneration of gut cells. Those differences in patterns of EdU⁺ cells within the blastema reflect the initial patterns of cycling progenitors labelled with a short pulse of EdU in the segments abutting the amputation plane (Fig. 1A,A′,C-C″), i.e. distributed in all cell compartments in the posterior-most segment and mostly localized within the gut in anterior segments.

We aimed to better understand the specific contribution of these 5 h EdU⁺ cells in anterior segments to regeneration. First, to confirm their gut identity, we coupled EdU labelling (i.e. after a 5 h EdU pulse followed by an amputation ‘−5’) with in situ hybridization for the gut specification factor foxA (Fig. 2G) and the smooth muscle marker calponin (Fig. 2H) (Brunet et al., 2016). The majority of the EdU⁺ cells (60%) colocalizes with foxA inside the regenerating gut (Fig. 2G,G′; Table S1), whereas only very few EdU⁺ cells colocalize with calponin, most of which were inside the gut embedded in smooth muscles (Fig. 2H,H′), thus confirming their gut identity. To go further into the characterization of these potential gut progenitors, we performed in situ hybridization on sections (at the anterior position S6) for several markers of the germline multipotency program (or the GMP signature), a set of genes expressed in adult multi/pluripotent stem cells and progenitors, in many metazoan species (Juliano et al., 2010). We observed the expression of GMP genes Myc, PiwiB and Vasa in a large subset of gut cells (Fig. S1C-E), some of which were proliferative 5 h EdU⁺ cells (Fig. S1F).

In summary, this study identifies an important pool of cycling gut progenitors cells in anterior segments, present before amputation, that express a pool of markers found in stem/progenitor cells, and that contribute to posterior regeneration and give rise to functional differentiated gut cells.

To further explore the contribution of anterior gut progenitors to regeneration, we developed a means to specifically label gut epithelial cells. Worms incubated with 1 µm diameter fluorescent beads would ingest them and specifically incorporate them within nearly all the gut epithelial cells (Fig. 2I-I″) except for the last few posterior segments (Fig. 2I). With this technique, we could then precisely determine the contribution of gut epithelial cells over the course of regeneration (five stages) after anterior amputation (Fig. 2J1-J5′). At stage 1, when the wound epithelium has formed, bead-labelled cells could be found within the segment abutting the amputation where the gut has retracted (Fig. 2J1,J1′). As soon as the blastema has formed at stage 2 (Fig. 2J2,J2′), the fluorescent beads were found inside the regenerating gut consistent with anus reformation. During the later growth and differentiation phases (stages 3 to 5) the fluorescent bead signal remained restricted to the gut epithelium (Fig. 2J3-J5′). At stage 5, by combining the previous EdU labelling experiment (EdU⁺ cells in the regenerative part coming from the cycling gut progenitors, compare with Fig. 2B) with the fluorescent beads (Fig. 2K,K′), we found overall colocalization of the two labels, confirming the congruence of the two methods for labelling the gut cells as well as confirming their specific contribution to blastema gut cells. We conclude that gut progenitors in anterior segments of unamputated worms contribute exclusively to posterior regeneration of the gut epithelium, and are therefore lineage restricted.

To assay the lineage restriction, if any, of the gut from the posterior-most segment, we first had to find a way to label them, as the posterior-most gut epithelial cells are unable to incorporate fluorescent beads (Fig. 2I) and because a 5 h EdU pulse labels many progenitors outside the gut in posterior segments (Fig. 1A′,C″). We overcame this problem by performing two sequential rounds of amputation on worms whose anterior gut progenitors were labelled with a 5 h EdU pulse or with fluorescent beads. After the first round, the regenerated gut of the posterior segments had incorporated progeny of the bead/EdU-labelled cells from the anterior gut. These worms could then be subjected to a second amputation (see Fig. 3A; see Materials and Methods).

Using this procedure, we observed differences in the distribution of labelled beads (Fig. 3B) or EdU⁺ cells (Fig. 3C) along the antero-posterior axis of the ‘blastema-like’ structure [i.e. a structure composed of a mix of a 10 dpa (anteriorly) and a 5 dpa (posteriorly) regenerated structure]. In the anterior part, which corresponds to the remaining tissues of the 1st regeneration, the fluorescent beads and the EdU⁺ cells were still specifically or mainly restricted to the gut, respectively (Fig. 3B′,C′). Strikingly, however, in the posterior part of the blastema-like structure, which corresponds specifically to the 2nd regeneration event, fluorescent beads and EdU⁺ cells were detected both inside and outside the gut, mainly in the dorsal side of the regenerating part (Fig. 3B″,C″, white arrows). Moreover, there were significantly more EdU⁺ cells in the posterior part (23.8%, Fig. 3D; Table S1) than in the anterior part (17.4%, Fig. 3D; Table S1). Similarly, the numbers of BrdU⁺ and EdU⁺/BrdU⁺ cells were higher in the posterior part of the structure than in the anterior part (Fig. 3D; Table S1). This means that the gut progenitors, previously labelled with EdU, are actively mobilized during the 2nd regeneration and seem to contribute to the formation of tissues other than the gut, and more than one third of them are still proliferative. In addition, highly proliferative cells participating in the 1st regeneration labelled with BrdU also contribute significantly to this 2nd regeneration event (Fig. 3C-D; Table S1).

We then determined the molecular identity of the cells originating from the gut progenitors that we observed outside the gut after the 2nd regeneration (Fig. 3B″,C″) by coupling EdU labelling and in situ hybridization for tissue-specific genes (Fig. 3E-I‴), including: the gut specification factor foxA (Fig. 3E-E″), the smooth muscle marker calponin (Fig. 3F-F″), the pygidial marker caudal (Fig. 3G-G″), the neural progenitor factor neurogenin (Fig. 3H-H″) and the ectodermal growth zone stem cell marker hox3 (Fig. 3I-I‴). In the anterior part of the ‘blastema-like’ structure, EdU⁺ cells mostly colocalize with foxA endodermal expression and are encompassed by the mesodermal expression domain of calponin, similar to the patterns obtained earlier for the 1st regeneration (Fig. 3E′,F′; compare with Fig. 2G-H′). In contrast, in the posterior part of the ‘blastema-like’ structure, many EdU⁺ cells did not colocalize with foxA but did with calponin (Fig. 3E″,F″, white arrows). Some of the EdU⁺ cells also expressed caudal within the mesoderm and the ectoderm of the regenerating pygidium (Fig. 3G″, white arrow). In contrast, we never observed any colocalization with the neurogenin expression domain (Fig. 3H″) in the ventral neurectoderm. Similarly, the EdU⁺ cells did not colocalize with the ventral part of hox3 expression domain in the ectodermal growth zone (Fig. 3I,I‴). In the dorsal part of hox3 expression domain, the presence of a very few EdU⁺ cells cannot be ruled out (Fig. 3I″).

Taken together, these results argue that the gut cells have acquired plasticity upon posteriorization. These results support the idea that the gut progenitors within the posterior-most segments may display such plasticity compared with their anterior lineage-restricted counter parts. It suggests the presence along the AP axis of different degrees of cellular plasticity positively associated with the amount of proliferation of the tissues we documented above.

This intriguing plasticity harboured by posterior gut progenitors is de facto spatially limited. Indeed, we showed that anteriorly located gut progenitors are lineage restricted. Their plasticity is thus apparently lost upon growth and differentiation of the tissues. We aimed here to describe the cellular mechanisms underlying this plasticity, and also to determine at which point the posteriorized gut progenitors would lose it. As cell proliferation and migration are often required in similar processes (Friedl and Gilmour, 2009), we hypothesized it would be the case here as well. To this end, we performed the same re-amputation procedure, but experimentally inhibited cell proliferation with hydroxy-urea (HU) or cell migration with the actin inhibitor Latrunculin B (LatB; Spector et al., 1983), which is used, for example, to inhibit leukocyte migration (Lerchenberger et al., 2013; Yan et al., 2019) during the second phase of regeneration (Fig. 4A). Previous results have shown that the inhibition of cell proliferation with HU does not prevent the formation of the blastema but hinders its growth and differentiation (Planques et al., 2019). The inhibition of cell migration (and modifications of cell shape) with LatB slows down regeneration and causes mild morphological defects (thicker anal cirri, Fig. S2A,B). In contrast to the controls in which EdU⁺ cells are found, as expected, outside the gut in the posterior region of the blastema-like structure (Fig. 4B-B″, white arrow), when cell migration is inhibited by LatB, most EdU⁺ cells remain mainly confined to the gut in both anterior and posterior parts of the ‘blastema-like’ structure (Fig. 4C-C″). To ensure that the regeneration slowdown observed after treatment with confined to LatB was not due to an indirect alteration of cell proliferation (Fig. S2B,C), we made a short EdU pulse in LatB-treated and DMSO-control samples at 5 dpa, and found a similar percentage of EdU⁺ cells in both conditions (Fig. S2B). A pulse of EdU at 2 dpa, followed by a chase until 5dpa, demonstrated that cells do divide under both conditions (Fig. S2C). We concluded that LatB does not affect cell proliferation, as cells are neither blocked in S-phase nor in M-phase, and that the restriction of EdU⁺ cells to the gut is due to cell migration defects. Similarly, most EdU⁺ cells are restricted inside the gut in the ‘blastema-like’ structure when inhibiting cell proliferation with HU (Fig. 4D-D″).

To determine at which point the posteriorized gut progenitors lose their enhanced plasticity, we performed the re-amputation procedure, but at later stages of initial regeneration (i.e. at 9 dpa or 12 dpa, instead of 5 dpa) when newly produced tissues are differentiating (Fig. 4A). We observed that, as expected, the distribution of EdU+ cells is mostly restricted inside the gut in the anterior part of the regenerating structure (1st regeneration, Fig. 4E,E′,F,F′) but also in its posterior part (2nd regeneration, Fig. 4E,E″,F,F″), showing that posteriorized gut progenitors have lost plasticity at those stages. We conclude that cell proliferation and the cellular functions dependent on actin are crucial for the posteriorized gut progenitors to acquire their plasticity, and that this plasticity is lost very quickly upon maturation of the tissues.

The above experiments showed that, whatever their position along the antero-posterior (AP) axis of the worms, the gut progenitors comprise only part of the blastema, even the plastic posteriorized gut progenitors. The anterior gut progenitors provide only blastema gut cells, and the posterior ones, although producing several derivatives, probably do not regenerate neural tissues (Fig. 3H) and can only, at best, regenerate a few stem cells of the growth zone (Fig. 3I). This raises questions about the origin of the cells that will give rise to the rest of the tissues, and, in particular, the stem cells of the regenerated growth zone. Additionally, as the position of the gut progenitors along the AP axis impacts their fate, could this also be the case for the other cells participating in the blastema formation? Our previous work showed that most of the blastema cells are of local origin, from the segment abutting the amputation plane, in the context of an anterior amputation (i.e. amputation ‘−5’; Planques et al., 2019). We thus sought to determine whether the source of the regenerative cells would remain local along the AP axis and also whether the stem cells of the growth zone would also be of local origin. To distinguish local versus more distant tissue contributions to the regeneration along the body axis, we used an experimental set-up consisting in two serial amputations performed either in anterior (‘−5’) or posterior (‘0’) locations (Fig. 5 and Fig. S4; and Fig. S3, respectively) (see Planques et al., 2019). First, to have a proxy for the cells activated by anterior or posterior amputation, we performed a 5 h EdU pulse at stage 1, when regeneration has been initiated and the wound epithelium formed (Fig. 5A), chased the EdU for 2 days (i.e. until 3 dpa) and performed a 1 h BrdU pulse before collecting the samples. We determined that 53% of the blastema cells (at 3 dpa for an anterior amputation) arose from those EdU⁺ cells proliferating after amputation (Fig. 5B). Moreover, they contribute massively to the pool of highly proliferative cells at 3 dpa (among the 32% of BrdU⁺ cells, 80% of them are also EdU⁺), which means that those EdU⁺ cells participate not only in the formation of the blastema but also in the subsequent phases of regeneration. Thus, they harbour a high regenerative potential. We wished to quantify the precise difference in the regenerative potential of the EdU⁺ cells in the segment abutting the amputation plane with those in the segment directly anterior to it, for the two positions of amputations (anterior versus posterior). We amputated worms anteriorly (see Fig. 5) or posteriorly (see Fig. S3), and let them regenerate for 24 h, performed a 5 h EdU pulse to label the cells activated by the amputation and let the worms regenerate for two more days until the blastema had formed. We then either amputated the blastema and the first abutting segment (Amputation A) or the blastema only (Amputation B). After this second amputation, we let the worms regenerate for 3 more days until another blastema had formed, and performed a 1h BrdU pulse before collecting the samples to quantify how many EdU⁺ are still proliferative after both amputations (Fig. 5A).

We found that the number and localization of BrdU⁺ cells are similar between condition A (Fig. 5C; Fig. S3A) and B (Fig. 5D; Fig. S3B; ∼33%, see individual values in Fig. 5G and Fig. S3C), whatever the position of the amputation (anterior versus posterior). In contrast, we obtained very different EdU⁺ cell patterns in the blastema depending on the second amputation condition (A versus B). For an anterior amputation, in condition A, only around 7% of internal cells are EdU⁺ (Fig. 5C,C′,G; Table S1), whereas in condition B, around 69% of the blastema cells are EdU⁺, in both superficial and internal tissues (Fig. 5D,D′,G; Table S1). The situation is similar for a posterior amputation: whereas in condition A, around 27% are EdU⁺ (Fig. S3A,A′,E), around 75% of the blastema cells are EdU⁺ in condition B (Fig. S3B,B′,E). Interestingly, ∼40% of the EdU⁺ cells are BrdU⁺ at 3 days post-2nd amputation in all conditions (Fig. 5C,D,G; Fig. S3A,B,E; Table S1).

Given that the only difference between conditions A and B is the absence or the presence of the segment abutting the first amputation plane, we conclude from these experiments that the cells activated by the amputation and located close to the wound have a higher plasticity than those located more anteriorly. They thus contribute to most of the blastema cells, regardless the site of the initial amputation (anterior versus posterior).

We next sought to determine the origin of the stem cells of the regenerated growth zone and, more precisely, whether they would also be of local origin all along the AP axis. To track the regenerated stem cells, we relied on the fact that they constitute a population of label retaining cells (LRCs; de Rosa et al., 2005). As putative stem cells, their proliferation rate is rather low and, consequently, after EdU incorporation the signal should be retained longer than in other cell types. We used the same experimental set-up as described above, but instead of collecting the samples after the 1h BrdU pulse, we chased both the EdU and BrdU by letting the worms regenerate for 4 more days, until the very end of the regeneration process (Fig. 5A). For an anterior amputation, as expected, the pattern of BrdU⁺ cells is similar between conditions A and B (Fig. 5E,E′,F,F′,G; Table S1). In contrast, the patterns of EdU⁺ cells are very distinct between both conditions. In condition A, there are still very few EdU⁺ cells (6%, Fig. 5G; Table S1) scattered in internal tissues (Fig. 5E,E′). As for condition B, there is a dramatic fourfold reduction of the proportion of EdU⁺ cells in the samples (from around 69% to 17%, Fig. 5F,G; Table S1). This reflects the fact that most of the blastema cells underwent enough cell divisions to dilute the EdU signal. The remaining 17% of EdU⁺ cells appear localized primarily at the interface between the pygidium (the terminal part of the worm body) and the new growing segments, which presumably corresponds to the regenerated growth zone stem cells (Fig. 5F,F′). This EdU pattern is what could be expected for an LRC. A similar experimental approach has been followed after an initial posterior amputation (Fig. 5A). In this case, there are higher proportions of EdU⁺ cells in both conditions (17.9% for condition A and 36.2% for condition B, Fig. S3C,D,E) but, similarly, EdU⁺ cells do not appear to be located in the growth zone in condition A, whereas they are in condition B (Fig. S3C-D′). The higher proportions of EdU⁺ cells for both conditions (A and B) for an amputation ‘0’, compared with an amputation ‘−5’ may reflect more proliferation in the tissues abutting a first initial posterior amputation than in more anterior tissues.

To confirm the localization and identity of those LRCs, we coupled EdU labelling with whole-mount in situ hybridization for markers of different cell populations of the growth zone (Gazave et al., 2013). We selected the genes hox3 and evx, which are expressed in the ectodermal cells of the growth zone, as well as piwiB, which is expressed in both the ectodermal and mesodermal cells of the growth zone (Fig. 5; Fig. S4). After an initial anterior amputation, in condition A, the few remaining internal EdU⁺ cells do not colocalize with hox3 signal (Fig. S5H,H′), whereas they do colocalize in condition B (Fig. 5I,I′, white arrow). Similar results were obtained for evx (Fig. S4A-B′) and piwi (Fig. S4C-D′).

These results demonstrate that the stem cells of the regenerated growth zone originate from local cells (i.e. from the segment abutting the amputation plane) activated by the amputation, whether this amputation has been performed anteriorly or posteriorly along the animal body axis. In addition, those cells have a higher plasticity compared with those located more distantly from the amputation plane (i.e. one segment upstream).

Our proliferation labelling experiments revealed the contribution of both gut and non-gut progenitors from non-amputated tissues in Platynereis posterior regeneration. Importantly, the progenitors from the non-gut anterior trunk tissues, which are cycling more slowly than the gut progenitors, give rise to the vast majority of blastema cells. Slowly cycling stem/progenitor cells are the main source of regenerative cells in other contexts of regeneration in many species. In mammalian skin (Koren et al., 2022) and gut (Karmakar et al., 2020) regeneration, they are crucial for wound repair. Similarly, they are involved in whole-body regeneration in planarians (Molinaro et al., 2021) or in the ctenophore Mnemiopsis leidyi (Ramon-Mateu et al., 2019), as well as in head regeneration in the cnidarian Hydra (Govindasamy et al., 2014). Usually the cell cycle of those slowly cycling progenitors has to accelerate to produce the cells required for regeneration. This can be achieved through either the acceleration of their S phase (as in Drosophila imaginal disc regeneration; Crucianelli et al., 2022) or the shortening of their G1 phase (as in Drosophila gut regeneration; Cohen et al., 2021).

In Platynereis, the cell cycle length in progenitors from the anterior non-gut tissues does not correlate with the timing of tissue reformation during posterior regeneration. In fact, it requires 62 h for the non-gut progenitors to cycle, during which time, a large highly proliferative blastema has already formed (Planques et al., 2019). Thus, only a cell-cycle acceleration can explain their massive contribution to the blastema. Although we now have data about cell cycle kinetics in non-amputated tissues, precise measurement of the cell cycle parameters during regeneration in Platynereis remains to be performed. Nonetheless, the identification of these two types of progenitors was decisive for a better understanding of the cellular sources of posterior regeneration in Platynereis.

In this study, we aimed to establish a model of regeneration for the annelid Platynereis at the cellular level. We first obtained new insights on the fate of the cells participating in the regeneration. We highlighted different degrees of plasticity of gut progenitors during regeneration by showing that, in contrast to their lineage-restricted anterior counterparts, posterior gut progenitors can widen their range of cell fate during regeneration and produce ecto/mesodermal derivatives. These different degrees of plasticity were shown to be positively associated with the levels of proliferation of the tissues, with the most proliferative tissues being localized posteriorly. As those posterior-most tissues close to the growth zone are also the most-recently formed, we can presume that anterior tissues are generally more differentiated than posterior tissues. This intuitive assumption for a continuously growing animal has some experimental support. For example, posterior-most segments never possess chaetae (extracellular structures produced by mature parapodia) nor do they express the chaetae-associated gene marker (chitin synthase) (Gazave et al., 2017). Similarly, they do not contribute to gas exchanges, as they never express globin markers (Song et al., 2020), nor to efficient digestion, as revealed by their inability to incorporate beads (Fig. 2I). Thus, the different levels of plasticity and proliferation our study uncovered may be negatively associated with the degree of tissue differentiation.

It has been shown previously in Platynereis that most of the blastema cells arise locally, from the segment directly abutting the amputation plane (Planques et al., 2019). We confirmed and extended this finding by determining that most blastema cells, including the stem cells of the regenerated growth zone, are of local origin whatever the level of differentiation of the segment abutting the amputation plane (i.e. anterior or posterior tissues).

Two broad strategies have been proposed for tissue regeneration in various contexts (Bideau et al., 2021). In one, there are pre-existing resident pluripotent stem cells that can migrate to the wound site and initiate formation of the missing tissue as in the planarian Schmidtea mediterrannea or the cnidarian Hydractinia symbiolongicarpus (Wagner et al., 2011; Varley et al., 2023). The existence of adult stem cells (pluripotent or with less potency) in regeneration has been proposed (with little evidence) in many annelids, including earthworms (Lumbriculus sp. and Enchytraeus japonensis; Randolph, 1892; Myohara et al., 1999; Sugio et al., 2012) and Capitella teleta (de Jong and Seaver, 2017). In addition, a recent single cell atlas for another species, Pristina leidyi, suggests the existence of such stem cells with a pluripotency signature, but their role during regeneration remains to be established (Álvarez-Campos et al., 2024). Although we cannot definitely rule out this possibility, our data do not support the idea that adult multi/pluripotent stem cells are involved during posterior regeneration in Platynereis.

In the second strategy, differentiated cells close to the wound undergo a partial dedifferentiation, forming lineage-restricted cells that can repair the injury, e.g. in the salamander limb regeneration (Kragl et al., 2009). Dedifferentiation has also been proposed for a couple of annelid species [i.e. Syllis malaquini (Ribeiro et al., 2021) or Alitta virens (Shalaeva and Kozin, 2023)] during posterior regeneration. So far, the cellular trajectories supporting this idea remain unclear but such data already highlight the likely diversity of cellular mechanisms of regeneration within annelids (Bely, 2014).

Our data led us to define a hybrid cellular model (Fig. 6) for Platynereis posterior regeneration: most of its regenerative cells have a local origin, but the massive contribution of slowly-cycling progenitors rules out the possibility of a ‘complete’ dedifferentiation process in which post-mitotic cells re-enter cell cycle. Rather, our results support the idea of pools of specialized progenitors with different replication rates, maintained throughout juvenile stage, that are mobilized upon an amputation trigger. Among them, posterior gut progenitors can become plastic during regeneration.

Seminal studies of embryonic development have led to the formalization of a broad model in which cell differentiation is an irreversible process, and development as a whole constitutes a gradual loss of potency from the totipotent zygote to fully differentiated adult cells (Caplan and Ordahl, 1978; Merrell and Stanger, 2016). This inflexibility of cell identity is currently being reconsidered in the light of recent genetic lineage tracing experiments, which have uncovered various developmental processes showing metaplasia: the acquisition of cell identities that are unusual for a given tissue (Virchow, 1886; Mills et al., 2019). Metaplasia is often observed during various regeneration contexts in which progenitor/stem cells can acquire a higher plasticity compared with tissue homeostasis, even though it is always limited in terms of potency and temporality. For example, in the zebrafish fin, specific subpopulations of fibroblasts can restore more types of fibroblasts during regeneration than they do during homeostasis, but they can only produce fibroblasts (Tornini et al., 2017). Likewise, in mammals, a skin injury can transiently widen the progeny of specific skin stem cells that can then produce all the diversity of skin cells but not cells outside the skin (Blanpain and Fuchs, 2014).

In this study, we uncovered the intriguing ability of posterior less differentiated gut progenitors to produce other types of derivatives (e.g. epidermis or muscle) during regeneration, which can also be considered as an example of metaplasia. Interestingly, the metaplasia of the gut in Platynereis is also limited in different ways. It is rather transitory; metaplasia probably stops as soon as the gut starts to differentiate. Related to this, it is also limited spatially in the body of the animal: only newly produced (and therefore posterior) gut progenitors can switch lineage during regeneration. Moreover, this metaplasia is only partial, as posterior gut progenitors likely do not produce nervous system or putative posterior stem cells.

Why are posterior gut progenitors with enhanced plasticity unable to produce nervous system derivatives upon amputation? In many annelids, the ventral nerve chord (VNC) from non-amputated tissues plays a major role in the formation of the nervous system in the regenerated structure (Sinigaglia and Averof, 2019). In Platynereis, nerves from the VNC will rapidly innervate the blastema (Planques et al., 2019) and may serve as both direct and indirect source of signals (Boilly et al., 2017). Perhaps such posterior gut progenitors cannot replace this important signal as well as the physical support provided by the VNC, in contrast to muscle cells, which appear de novo in the blastema and do not seem to rely on the muscle system in non-amputated tissues for their initial differentiation (Planques et al., 2019).

Upon amputation, posterior gut progenitors also never produce the putative posterior stem cells responsible for the constant growth of the worms (Gazave et al., 2013). We can imagine that ‘enhanced’ posterior gut progenitors are unable to produce stem cells as the latter may possess too high a potency, which is unreachable through their metaplasia. However, our results indicate that other cells (the identity of which remains to be determined) located in the segment abutting the amputation plane can reach such a potency that they produce posterior stem cells. As such, the gut is not the only tissue capable of metaplasia after the amputation signal. It would be interesting to determine whether this metaplasia is specific to given types of tissue or whether it depends solely on the degree of differentiation of any tissue.

This phenomenon of metaplasia of posterior gut cells highlights an important point regarding the cellular processes involved during regeneration. Indeed, different sources of cells along the antero-posterior axis of the animal contribute to the blastema and eventually to a morphologically identical structure. This notion of different cellular ‘paths’ for the same regeneration process is intriguing and rather uncommon. One key example is the retinal regeneration of Xenopus, in which the involvement of three different cellular sources has been reported, depending on the extent of the injury (Parain et al., 2024). So far, we have no idea about the molecular mechanisms underlying such diversity of cellular mechanisms for regeneration.

In addition, those results led us to consider the robustness of the posterior regeneration in Platynereis. In a previous study, we reported that serial amputations (up to 10) do not impair regeneration efficiency, at least at the morphological level (Planques et al., 2019). Similar robust regeneration events after successive injuries exist in some species, such as the zebrafish fin (Azevedo et al., 2011) and the newt's lens, that can properly regenerate even after 18 repeated lens removal spanning over 30 years (Eguchi et al., 2011), without any modification of the associated transcriptional program (Sousounis et al., 2015). In contrast, in other species, the succession of injuries does reduce regeneration effectiveness (Eming et al., 2014), either by exhausting the tissue-specific stem cell pool [e.g. in the mouse lung epithelium (Ghosh et al., 2021) or in the Drosophila gut epithelium (Haller et al., 2017)] or through the misexpression of regeneration initiation genes [e.g. in the axolotl limb (Bryant et al., 2017)]. We can hypothesize that the different cellular paths for regeneration, which depend on the localization, proliferation and differentiation levels of the tissues, are important for the morphological robustness of the process.

P. dumerilii juvenile worms were obtained from a husbandry established in the Institut Jacques Monod (for breeding conditions, see Dorresteijn et al., 1993; Vervoort and Gazave, 2022). Standard worms used in experiments were 3-4 months old with 30-40 segments, and were amputated according to the procedure detailed by Planques et al., 2019; Vervoort and Gazave, 2022. In some experiments, regenerative parts were also re-amputated (either blastemas were totally removed or regenerative parts were amputated in their middle, see Results section). At given time points (see Results section), regenerative parts and the two posterior-most segments were collected and fixed in 4% paraformaldehyde (PFA) diluted in PBS Tween20 0.1% (PBT) for 2 h at room temperature, then washed in PBT and gradually transferred in 100% methanol, at which point they can be stored at −20°C. For fluorescent beads experiments, the regenerative parts were similarly collected and fixed but were not put in methanol. Instead, they were stored for up to 24 h in PBT at 4°C.

To perform histological sections, the samples were fixed in a solution of 4% PFA diluted in 1×PBS (without Tween 20) for 90 min at room temperature and washed in 1×PBS. They were then cryoprotected in a solution of sucrose diluted in PBS (300 g/l) and stored for 4-5 days at 4°C. After the samples had settled in the sucrose solution, they were transferred into OCT embedding medium (Tissue Freezing Medium, Leica). The remaining sucrose was removed and the samples were put into moulds and positioned according to the desired type of section (transverse or longitudinal). The samples were subsequently frozen with dry ice and stored inside the moulds at −80°C. The sectioning was performed with a microtome (Leica CM3050S). Sections of 12-14 µm were collected on SuperFrost glass slides and stored at −80°C. For in situ hybridization on slices experiments, samples were fixed in a solution of 3.7% formaldehyde, 0.2% glutaraldehyde diluted in 1×PBT, then similarly cryoprotected, embedded, frozen and cut into 10 µm slices. The cross-sections were then processed for immunostaining (Briscoe et al., 2000) with mouse anti-acetylated tubulin (Demilly et al., 2013, Sigma T7451, 1:500), for phalloidin labelling (Cytoskeleton, PHDR1; 1/500), or for in situ hybridization or EdU labelling (see below).

Proliferating cells in S phase were labelled by incubating the worms with the thymidine analogues EdU (5-ethynyl-2′-deoxy-uridine) and/or BrdU (bromo-deoxy-uridine), at respective concentrations of 50 µM and 1 mM in natural fresh sea water (NFSW). Various incubation conditions (duration and biological stage) and pulse-and-chase experiments were performed as described in the Results section and related figures. The samples were then fixed and sections produced as described above. Briefly for EdU and/or BrdU labelling, after rehydration, the samples were digested with Proteinase K (40 µg/ml for 10 min), then the enzyme was inactivated with 2 mg/ml glycine in PBT (1 min), and the samples post-fixed with 4% PFA in PBT (20 min) and washed with PBT. EdU-labelled cells were fluorescently marked by click-it chemistry with the specific addition of a fluorescent azide on EdU molecules (Click-iT EdU Cell Proliferation Kit, 488 or 555 dye, ThermoFisher, C10337 and C10638), following the procedures of Demilly et al. (2013) and Vervoort and Gazave (2022). For EdU labelling on sections, samples were shortly permeabilized in PBS Triton 0.1% before the click-it reaction. BrdU-labelled cells were marked by immunohistochemistry (primary antibody MoBU-1, mouse, 1:250, ThermoFisher, B35128; secondary antibody anti-mouse IgG Alexa Fluor 555 conjugate, 1:500, goat, Cell Signaling, 4409). First, they underwent an antigen-retrieval treatment with hydrochloric acid (final concentration of 2 M diluted in distilled water for 1 h at room temperature) and were incubated with the antibodies after a blocking step in a solution of sheep serum diluted in PBT. Dual labelling with EdU and BrdU were performed according to a method previously described by Liboska et al. (2012). The EdU sites were first marked as described above, and the remaining unlabelled EdU sites were then saturated by click-it chemistry with an excess of non-fluorescent azide (azido-methyl-phenyl-sulfide 95%, Sigma, 244546). The reaction was carried out according to the manufacturer's recommendations except the standard fluorescent azide was replaced by the non-fluorescent azide at a final concentration of 2 mM. The samples were then incubated in hydrochloric acid and the BrdU sites were bound by immunohistochemistry as described previously.

Total cell cycle length (Tc), S-phase length (Ts) and growth fraction (GF or proportion of proliferative cells) were determined by a cumulative labelling experiment with EdU as defined by Nowakowski et al. (1989) and Locker and Perron (2019), and using an Excel spreadsheet provided by Dr R. Nowakowski (Florida State University College of Medicine, Tallahassee, FL, USA). Non-amputated worms (from n=9 to 13) were exposed to 5 µM EdU for increasing lengths of times, until all proliferative cells were labelled (1, 5, 10, 16, 24, 48 and 72 h). The proportion of EdU⁺ cells was determined at three different positions, segments 1, 6 and 7 for the gut and other trunk tissues (see Cell counting section).

Probe synthesis and colorimetric whole-mount in situ hybridization were performed as described by Demilly et al. (2013) and Vervoort and Gazave (2022). As detailed above for EdU and/or BrdU labelling, after rehydration, the samples were first digested with Proteinase K and post-fixed with PFA. Following that, the samples were pre-hybridized in hybridization buffer at 65°C (1.5 h), and then incubated with the DIG-labelled probes overnight at 65°C. DIG was then bound to specific antibodies bearing alkaline phosphatase and the tissues were stained by cleavage of NBT (Nitro Blue Tetrazolium chloride) by alkaline phosphatase, catalysed by BCIP (5-bromo-4-chloro-3-indolyl phosphate). Whole-mount in situ hybridization samples intended for observation under bright-field microscopy were mounted in glycerol. Whole-mount in situ hybridization samples intended for observation under confocal microscopy using the reflection procedure (see below) were counterstained with 0.1% Hoechst overnight at 4°C to stain nuclei and mounted in glycerol/DABCO (2.5 mg/ml DABCO in glycerol). In situ hybridization on slices was carried out similarly, with the following specificities: samples were digested using PBT+0.2% Triton for 20 min. During hybridization step, samples were covered using parafilm to avoid evaporation of the probes. When combined with EdU labelling, the in situ hybridization was first performed as described (whole mount or slice). After the coloration with NBT/BCIP, the EdU-labelled cells were fluorescently marked by click-it chemistry, as described previously, and then all the nuclei were counterstained with Hoechst and finally mounted in glycerol/DABCO.

To perform in vivo labelling of worm gut cells, we used 1 µM diameter fluorescent beads (Fluoresbrite PolyFluor 570 Microspheres, Polysciences, 24061-10) that were ingested by the worms. Worms were incubated in a solution of beads diluted in NFSW (1:100) for 1 week in 24-well plate, one worm per well, in the absence of food. After incubation, worms were rinsed with NFSW to remove non-ingested beads and individually monitored with a fluorescent binocular microscope to determine at which point the gut was marked with fluorescent beads, as the gut is never entirely labelled (the terminal part is always beads free). Posterior amputations were then performed immediately anterior to the bead labelling limit, and the worms were allowed to regenerate in NFSW complemented with food for specific times, depending on the following experiments (see Results section and associated figures). Samples were then collected, counterstained with Hoechst and mounted in glycerol/DABCO. When combined with in vivo gut cell labelling, EdU labelling was performed by click-it chemistry, as seen previously, but for a shorter period of time (5 min instead of 1 h).

We established a procedure to label posterior gut cells (which do not incorporate fluorescent beads or for which EdU labelling is not specific) by giving a posterior identity to anterior tissues. We took advantage of the fact that after an amputation in an anterior segment, gut progenitors, labelled either with a 5 h EdU pulse or with fluorescent beads, participate in the regeneration of gut cells that retain EdU and bead labelling when regeneration has finished. Unamputated worms are incubated either with fluorescent beads for 1 week or with EdU for 5 h, and amputated anteriorly shortly after. After 5 days of regeneration, a second amputation in the middle of the regenerated structure, removing the pygidium and growth zone that had regenerated after the first amputation, was carried out. Worms then regenerate a second time for 5 more days, before collection of the samples for further experiments. This regeneration procedure establishes a proxy of posteriorized gut epithelial cells that are labelled. A schematic representation of the posteriorization procedure is available in Fig. 3A.

Cell proliferation was blocked using hydroxy-urea (HU) at 20 mM, as previously described (Planques et al., 2019). Cell migration was blocked using Latrunculin B (LatB) at 20 nM (as used by Tweeten and Anderson, 2008). HU and LatB were dissolved in sea water and DMSO, respectively. Both solutions were changed every 24 h to maintain their activities for the duration of the experiment (5 days). Briefly, individual worms were incubated in 2 ml of each solution (or control) on 12-well plates. LatB-treated worms were scored every day for the regeneration stages that had been reached, as previously described (Planques et al., 2019; Vervoort and Gazave, 2022).

Bright-field images of colorimetric whole-mount in situ hybridization samples were acquired with a Leica CTR 5000 microscope. Fluorescent confocal images of whole-mount in situ hybridization samples were acquired with a Zeiss LSM780 microscope using a 633 nm laser in reflection mode as described previously (Jékely and Arendt, 2007). Other fluorescent confocal images were acquired with either a Zeiss LSM780 or LSM980 confocal microscope. Image processing (contrast and brightness, z-projection, auto-blend layers and transverse views) was performed with FIJI and Abode Photoshop. Figures were assembled using Adobe Illustrator.

An automatic cell counting procedure was established and performed with the Imaris software by BitPlane (version 9.5). First, for each sample, all nuclei positions were identified and modelled using the Hoechst signal and the function ‘Spots’ with a standardized nucleus diameter of 5 µm. A region of interest (ROI) corresponding specifically to the regenerative part, a body segment or the gut was then manually delineated with the surface tool, using the Hoechst signal and the general morphology of the structure. Then, the spots inside the ROI were sorted along the fluorescent signals of the EdU or/and BrdU with a filter ‘Intensity Mean’ to discriminate true positive nuclei from background. This procedure allowed us to determine the absolute number of nuclei inside the ROI and, among them, the number of positive nuclei for each signal; hence, we could extract the proportions of EdU⁺, BrdU⁺ and EdU⁺/BrdU⁺ cells for each sample.

All statistical tests and subsequent graphic representations were performed with GraphPad Prism 7. Mann–Whitney U-tests were used to compare whole-blastema proportions of EdU⁺, BrdU⁺ and EdU⁺/BrdU⁺ cells between different experiments. Wilcoxon signed-rank tests were used to compare those proportions between different parts of the same sample. Multiple Mann–Whitney U-tests were used to compare the labelling index between the gut and other trunk tissues for all durations of EdU Pulse. Similarly, multiple Mann–Whitney U-tests were used to compare LatB-treated and control worms for each day of treatment.

This study is dedicated to the memory of our friend, mentor and supervisor, the late Professor Michel Vervoort. We are grateful to all members of the Gazave & Vervoort lab for their support and suggestions on this study. We deeply thank Dr Quentin Schenkelaars for suggesting the use of fluorescent beads, as well as Dr Gabriel Krasovec and Dr Lucie Laplane for helpful comments on the manuscript. We thank Dr Yves Clément for his help with statistical analyses. We are grateful to Dr Roger Karess for his diligent proofreading of the manuscript. We acknowledge the ImagoSeine core facility of Institut Jacques Monod, a member of France-BioImaging (ANR-10-INBS-04) and IBiSA, with the support of Labex ‘Who Am I’, Inserm Plan Cancer, Region Ile-de-France and Fondation Bettencourt Schueller. We are grateful to the staff of the animal facility for their help with the Platynereis culture.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202452.reviewer-comments.pdf