An approach for elucidating dermal fibroblast dedifferentiation in amphibian limb regeneration

Urodele amphibians, Pleurodeles waltl and Ambystoma mexicanum, have organ-level regeneration capability, such as limb regeneration. Multipotent cells are induced by an endogenous mechanism in amphibian limb regeneration. It is well known that dermal fibroblasts receive regenerative signals and turn into multipotent cells, called blastema cells. However, the induction mechanism of the blastema cells from matured dermal cells was unknown. We previously found that BMP2, FGF2, and FGF8 (B2FF) could play sufficient roles in blastema induction in urodele amphibians. Here, we show that B2FF treatment can induce dermis-derived cells that can participate in multiple cell lineage in limb regeneration. We first established a newt dermis-derived cell line and confirmed that B2FF treatment on the newt cells provided plasticity in cellular differentiation in limb regeneration. To clarify the factors that can provide the plasticity in differentiation, we performed the interspecies comparative analysis between newt cells and mouse cells and found the Pde4b gene was upregulated by B2FF treatment only in the newt cells. Blocking PDE4B signaling by a chemical PDE4 inhibitor suppressed dermis-to-cartilage transformation and the mosaic knockout animals showed consistent results. Our results are a valuable insight into how dermal fibroblasts acquire multipotency during the early phase of limb regeneration via an endogenous program in amphibian limb regeneration.


Background
Urodele amphibians can regenerate various organs including limbs. In their appendage regeneration, a structure called a blastema is induced after organ damage, and the induction of blastema is responsible for successful appendage regeneration. In other words, no blastema formation essentially results in no appendage regeneration. Thus, understandings of blastema formation would be the closest way to elucidate the ability of organ-level regeneration in urodele amphibians. And limb regeneration is an ideal model to study blastema formation and the organlevel regeneration ability in urodele amphibians.
Limb regeneration necessarily accompanies blastema formation, which has been considered to be a similar structure to a developing limb bud [1]. A regeneration blastema actually showed a similar gene profile to a developing limb bud [2]. Thus, a mechanism to induce a re-developmental field in an adult body has been investigated with great curiosity. From this angle, two issues are embossed: 1) how is the re-developmental field induced? 2) how do the differentiated cells turn into cells with an embryonic gene profile? Regarding blastema induction, classic studies clearly demonstrated that the nerves play a central role in the process [3,4]. Denervation from limbs results in no regeneration after limb amputation. Wound Open Access *Correspondence: satoha@cc.okayama-u.ac.jp 1 Research Core for Interdisciplinary Sciences (RCIS), Okayama University, 3-1-1, Tsushima-naka, Kitaku, Okayama 700-8530, Japan Full list of author information is available at the end of the article healing takes place instead. The amputation plane is covered by the migrating epidermis immediately after limb amputation. After the epidermal covering of the exposed amputation surface, the nerves secret some factors to create a regenerative environment. Molecules, which are secreted from the nerve ends and contribute to creating the regenerative environment in an amputated limb, had been investigated for a long time. Some factors have been identified as nerve factors [5][6][7]. Among them, FGFs and BMPs can induce a blastema in multiple species and organs [8,9]. Fgf2, Fgf8, and Bmp2 are expressed axolotl neurons in the dorsal root ganglion (DRG). Application of FGF2 + FGF8 + BMP2 (B2FF) to the wounded skin results in blastema formation instead of skin wound healing in urodeles, axolotls and newts [9]. FGF2 + FGF8 without BMP2 (FF) can induce a blastema. However, the FF-induced blastema does not have the ability to keep growing up to a patterned limb. BMP2 induced structure lacks proper blastema gene expressions and does not have the ability to grow a limb [9]. Downregulation of those genes in DRG neurons resulted in decreasing in the limb regeneration ability [10]. Thus, FGF2, FGF8, and BMP2 can work as nerve-secreting molecules and regeneration inducers in Ambystoma mexicanum (axolotl) and Pleurodeles waltl (newt). Thanks to the determination of the inductive molecules, an approach to the other issue of how the differentiated cells turn into an embryonic state can be possible.
A regeneration blastema fulfilled by undifferentiated cells is induced on the amputation plane, below which differentiated tissues exist. As mentioned above, a similar morphogenetical program to that in a developing limb bud regulates after blastema formation. However, a regeneration-specific mechanism would be involved in the early phase before blastema formation since it is necessary to induce undifferentiated blastema cells. The induction phase of blastema cells would be regeneration-specific. Undifferentiated blastema cells, having an embryonic profile, emerged from the differentiated tissues [2]. The dermis is the major source of blastema cells among the limb tissues [11,12]. Dermal fibroblasts have been considered to turn into blastema cells. Dermisderived blastema cells can change their cell type into varied connective tissue lineage, such as cartilage. Such transdifferentiation ability is restricted within the connective tissue lineage [11,13,14]. This is consistent that a blastema has a similar gene expression profile to a developing limb bud. Limb bud cells derived from the lateral plate mesoderm can participate in varied connective tissue lineages, but not other non-connective tissues, such as the muscle [15]. Considering these, dermal fibroblasts may get reprogrammed and become limb bud-like cells. On the other hand, another possibility can explain the emergence of blastema cells from the differentiated dermis. It is well known that multipotent stem cells exist throughout the body. The dermis involves many cells. Thus, it is feasible that the axolotl dermis contains stem cells, and that the stem cells participate in blastema formation. Whether the induction mechanism of blastema cells is cellular reprogramming or stem cell recruiting has not been clarified The determination of the regeneration induction molecules makes us possible to investigate molecular regulations, which dermis-derived cells receive just after limb amputation. Most blastema cells are derived from the dermis [11]. And the dermis-derived cells show connective tissues-restricted multipotency in limb regeneration [11,13,14]. The multipotent blastema cells can be induced by B2FF application in axolotls. Hence, investigation of a regulation downstream of B2FF in dermal fibroblasts leads to understandings of cellular reprogramming in amphibian organ regeneration.
Here, we explored the downstream gene network of B2FF by comparing gene expressions in vitro. We used mouse dermal fibroblasts as a representative of nonregenerative animals and newt skin fibroblasts as a representative of regenerative animals for the comparative analysis. The following in vivo experiments were performed using axolotls because of the benefits of fluorescent observation. The function of Pde4b in conferring multipotency was investigated by the usage of a chemical compound and the mosaic Pde4b crispants. Our findings provide important insights into the generation of multipotent cells during limb regeneration in caudate amphibians.

Results
We first developed an in vitro system to investigate the blastemal transformation of dermal fibroblasts. We cannot keep axolotl cells in the satisfyingly proliferative state for a long time. Cultured axolotl cells were inevitable that they decrease in cell division and undergo senescence and/or quiescence state after a couple of passages. Dermal fibroblasts derived from Pleurodels waltl (newts), however, were highly mitotic and could be stably cultured for a long time. Newt cell line was established from an animal that stably expresses red fluorescent proteins on the cell membrane (Fig. 1A, B, E, F). Due to the property of the membrane-anchored mCherry, the fluorescent signal was observed on the cell membrane and shows the spotted pattern (Fig. 1E, F). The growth rate was approximately 1 doubling per week (Fig. 1G). To investigate the blastemal transformation of the cells, we treated the cultured cells with the regeneration inductive molecules (B2FF: Fig. 1C, D). No significant change was observed after 48 h by B2FF treatment. EdU incorporation assay at the 48 h time point revealed the upregulation of cell proliferation (Fig. 1H). The B2FF treated cells showed large cell aggregates 2 weeks after the treatment (Fig. 1J, L). No cellular aggregates were observed in the control (Fig. 1I, K). In the control, no cell aggregation was formed even after 3 weeks even though the cultured cells reached over-confluent (n = 4/4). Sections of the cell aggregates were prepared and gene expression of blastemal marker genes was investigated by in situ hybridization. Msx1 expression was confirmable throughout the cell aggregates (Fig. 1M), and Msx2 expression was much intense in the surface area (Fig. 1N). The signal of Prrx1 was weak but recognizable ( Fig. 1O) as compared to the control (Fig. 1P). The results from the in situ hybridization were confirmed by the quantitative RT-PCR (Fig. 1Q). These results suggest that newt skin fibroblasts can react with B2FF under in vitro environment.
We transgrafted the cell aggregates into an axolotl blastema to investigate the differentiation capability of cells involved in the aggregate. Xenografting of newt cells into axolotl tissues had been described previously [16]. Newt cells can survive in axolotl tissues for a month. Because of the transparency of an axolotl limb as compared to a newt limb, axolotl limbs are much useful to trace grafted cells in limb regeneration. The B2FF induced aggregate could be easily removed from the plastic dish and transgrafted into an axolotl mid-bud blastema ( Fig. 2A). Regarding the control (non-treated), the cultured cells never form aggregates. Hence, cell sheets were collected using a scraper and transgrafted into an axolotl blastema ( Fig. 2A). Axolotl limbs are relatively transparent, which allows the trace of the grafted cells by the red-fluorescent (Fig. 2B, C). The grafted newt fibroblasts could survive in the axolotl tissues and participate in a regenerate (Fig. 2B, C). To explore cellular contribution, regenerated limbs were sectioned ( Fig. 2D-F). Immunofluorescent analysis revealed the location of the grafted cells (red) and cartilage formation (Col2a; green) in the regenerate ( Fig. 2D-F). Little mCherry + was found in the cartilage when the control cells (non-treated) were transgrafted (Fig. 2D1-3). In contrast, the mCherry + cells were identifiable within the cartilage region when the B2FF treated cells were grafted ( Fig. 2E1-3). Considering the overlap of the Col2a signal and the mCherry signal, it is very likely that the dermis-derived newt cells became cartilaginous cells. Cell surface-located and spotted red signals were observable, consisting of the fluorescent pattern observed in Fig. 1F (Fig. 2F). Furthermore, it was suggested that dermis-derived cells can redifferentiate into various connective cell types [11,13,14]. To clarify whether mCherry + cells were observed in other connective tissues, we focused on the tendons. Tenascin is a marker of tendons and ligaments and is expressed in the connecting region between skeletal structures and muscles in amphibians [17]. Tenascin expression was visualized by immunofluorescence (Sup. Fig. 1). Tenascin expression was observed in the peripheral region of the epiphysis and some fibroblast near the epiphysis (Sup. Fig. 1). Confocal observation confirmed that tenascin was located by the mCherry + cells, suggesting that the mCherry + cell became the tendon cell (Sup. Fig. 1B, C). Those data suggest that the grafted B2FF treated newt fibroblasts participated in some connective tissue lineage.
Next, the comparative RNA-Seq analysis was conducted in order to identify the genes related to the acquisition of multipotency. To investigate gene dynamics prior to the formation of cell aggregate, cells 48 h after B2FF treatment were harvested for comparison. Furthermore, to focus on the specific gene dynamics in the regeneration-competent animals, mouse cells were used as a comparison. Mouse dermal fibroblasts were collected from neonates and 5 passages were undergone before the experiment. B2FF treatment was performed for 48 h. RNA-seq was performed by CAGE-seq. To compare gene expression between newts and mice, gene symbols were used in this comparison. This was because of the lack of fixed genome information of Pleurodeles waltl. Thus, a precise orthologue determination could not be fixed. We found 4185 genes were commonly expressed in the cultured cells of both species (Fig. 3A, Supplemental Data 1). Apparently, no huge different profiles could be found as both species were compared (Fig. 3A). B2FF treatment did not change the gene profile greatly as shown in Fig. 3A. Among those genes, we selected the genes showing FC < -0.5 and FC > 0.6. We found that there were 42 genes affected by B2FF treatment (Fig. 3B, Supplemental Data 1). We found out that Col1A2, which is a major component of the dermis, was down-regulated by B2FF treatment in both species (Supplemental Data 1). This is consistent with the report, in which an axolotl limb blastema is a less collagenous structure [18]. We further focused on the genes which react oppositely to B2FF in the two species  These 6 genes dynamics were also investigated in axolotl blastemas (Fig. 4). As mentioned, axolotls have many advantages in vivo experiments. Thus, we investigated the gene expression patterns in axolotl blastemas in order to perform further experiments in axolotls. qPCR analysis revealed that Fblim, Loxl2, Pde4b, Spry1, Timp3 were upregulated in the blastema. Fblim1, Loxl2, and Pde4b showed quick upregulation after limb amputation (Fig. 4). Only Pde4 expression was settled down at 10 days post-amputation (dpa). Spry1 and Timp3 have a late-activation profile in axolotl limb blastemas. Adra2a expression was not detectable throughout the period we tested. These gene expression patterns were confirmed by in situ hybridization. (Fig. 5). Consistently, Adra2a was not detectable at 2, 5, and 10 dpa (Fig. 5A, A' , G, G' , M, M'). Fblim1 expression could be recognized in the amputation region from 2 dpa to 10 dpa (Fig. 5B, H, N). The signal of Fblim1 could be observed at the border of the amputation plane (Fig. 5B'), but the signal was weakened in the later time points in the stump regions ( Fig. 5H' , N′). Loxl2 expression could also be detected from 2 dpa (Fig. 5C, I, O). The Loxl2 expression was confirmed in the part of the basal layer of the blastema epithelium ( Fig. 5C, I). Loxl2 activation was relatively specific in the distal (blastemal) region and not apparent in the stump region ( Fig. 5C' , I′, O′). Upregulation of Loxl2 likely takes place in the specific cells, rather broader blastema mesenchymal cells. Pde4b was upregulated in the early stages (2 and 5 dpa; Fig. 5D, J). The upregulation of Pde4b could also be observed around the amputation sites 2 and 5 dpa (Fig. 5D' , J'). But Pde4b expression was down-regulated in the later stages (10 dpa) (Fig. 5P, P'). Spry1 signals were not obvious 2 dpa in both the distal and the proximal region (Fig. 5E, E'). But the signal became confirmable in the blastema 5 and 10 dpa in the blastema (Fig. 5K, Q). Spry1 expression was also observed around the amputation site 5 dpa (Fig. 5K'). No signal was confirmed in the stump region 10 dpa (Fig. 5Q') Upregulation of Timp3 could be detected 2 and 5 dpa (Fig. 5F, L). However, the Timp3 signal was weakened in the proximal region of the blastema although the distal blastema maintains a relatively high level of Timp3 expression at 10 dpa (Fig. 5R). The faint signal could also be observed around the amputation site 2 dpa (Fig. 5F'). But no signal could be detected in the proximal region 5 and 10 dpa (Fig. 5L' , R') Those Fig. 3 Selection of genes responding to B2FF in cell culture. A Z-score obtained via CAGE-seq is visualized as the heatmap. The samples are mouse dermal fibroblasts with B2FF application, mouse dermal fibroblasts without B2FF application (CTRL), newt dermal fibroblasts with B2FF application, and newt dermal fibroblasts without B2FF application (CTRL) from the left. B the heatmap of the selected genes, which were up/down-regulated by B2FF application. C the gene expression pattern was confirmed by the quantitative RT-PCR. Genes, which were oppositely responded to B2FF in mouse and newt cells, were selected and confirmed. *p < 0.05. ***p < 0.01 data suggest that the selected 5 genes from the RNA-seq in mouse and newt cells were consistently upregulated in the axolotl blastema.
Instability of cellular differentiation should be induced in the early phases prior to blastema formation. Blastemas at 5 dpa have been reported to express blastema marker genes and blastemas at the time point sometimes take a dome shape containing blastema cells [19]. Moreover, the induction mechanism of cellular dedifferentiation and maintenance mechanism of an undifferentiated state would be different. Considering those, genes related to the induction of instability of cellular differentiation would be upregulated in early, and down-regulated in later. From this point of view, we thought Pde4b, which showed early upregulation and late down-regulation, was suitable for this criteria.
We next attempted to inhibit Pde4b functions in limb regeneration using a chemical inhibitor, Rolipram. Rolipram binds to the catalytic sites of PDE4B at several amino acids, where are 100% conserved between human PDE4B and axolotl PDE4B (Sup. Fig. 2). It is well known that PDE4B has a function to hydrolyze cAMP to 5'AMP [20]. To confirm the inhibitory effects to PDE4B in axolotl tissues, cAMP concentration in axolotl limbs was measured by ELISA (Fig. 6A). Consistently, cAMP concentration in limb tissues was upregulated by the 7 day-Rolipram treatment (Fig. 6A). This suggests that Rolipram can effectively inhibit hydrolysis of cAMP by PDE4B in axolotl limbs. Next, we investigated the dermal fibroblast's transdifferentiation into cartilaginous cells in the presence of Rolipram (Fig. 6B-E). The limb skin from a GFP animal was transgrafted onto a normal animal, and the grafted limb was kept a week for the recovery from the grafting damages (Fig. 6B). Then, the limb was amputated, and the limb-amputated animals were kept in Rolipram-containing water until digits were identifiable (Fig. 6B, C1-C3). The GFP positive domain was expanded from the amputation stump to the digit tips of the regenerate (Fig. 6C1-3). The regenerates were fixed and sectioned. Col2A1 + cartilage cells were revealed by in situ hybridization and the location of the GFP + cells was revealed by immunofluorescence (Fig. 6D, E). In the control samples, GFP + cells were observable in the Col2A1 + region (Fig. 6D, Table 1). Five limbs were obtained and sectioned. Sectioning was performed on the entire limb along the dorsoventral axis. Only sections with GFP + cells in the mesenchymal region on the prepared sections were extracted, and the GFP + cells were counted. We found 3321 GFP positive cells in 104 extracted sections, of which 149 GFP positive cells were Col2A1 + . In the Rolipram-treated animals, GFP + cells were little observed in the Col2A1 + region (Fig. 6E, Table 1). We obtained 36 sections from 7 Rolipram-treated samples. Although the number of sections obtained from a single limb sample was about the same as the control, the number of sections to be extracted in the Rolipram-treated samples was lower as compared to the control. This was because of the poor participation of GFP + cells in the regenerates compared to controls. Using the same method of cell counting as the control, we observed only one Col2A1 + cell out of 462 GFP + cells (Table 1). One reason why relatively few GFP+ cells participate in a regenerate in Rolipramtreated samples is that cAMP may affect cell migration. Actually, Rolipram treatment resulted in a slight decrease in cell migration performance, compared to the control (Sup. Fig. 3). This raises the possibility that decreasing the participation of GFP+ cells in cartilage by Rolipram treatment is a result of impairing the cell mobility of a certain cell population. However, if we focus only on the cells that participated in the regenerates, the difference in the frequency of GFP+ cells in cartilage is apparent. Therefore, it is suggested that cAMP is involved in the appearance of cells with fluctuating differentiation in a blastema.
We further investigated increasing in cAMP concentration impaired cartilage transdifferentiation from dermal fibroblasts. Dibutyryl-cAMP is a cell-permeable cAMP analog that activates cAMP-dependent protein kinases [21]. Similarly, GFP + skin was transgrafted onto a normal animal and the GFP skin-grafted limb was amputated to trace the lineage of GFP + cells in the absence or presence of dibutyryl-cAMP (Fig. 7A). The amputated limbs were kept until the regenerates reached the digit stage (Fig. 7B, F). To visualize GFP and Col2A1, we performed immunofluorescence on the identical sections (Fig. 7C-I). In the control sample, GFP + cells in the cartilaginous region could be detected as well as epidermis and dermis (Fig. 7C-E). On the other hand, the Dibutyryl-cAMP treated limbs showed a little number of GFP + cells in the cartilaginous region ( Fig. 7G-I). We plotted the rate of the GFP + /Col2A1 + in the regenerates (Fig. 7J). We counted 13 sections from 8 independent animals in the control and 21 sections from 17 independent animals in the Dibutyryl cAMP-treated animals. It is noteworthy that the two exceptional plots in the Dibutyryl cAMP-treated samples were derived from an identical animal. These results consistently suggest that cAMP concentration influences a fluctuation of differentiation of dermal fibroblasts. Next, we attempted to inhibit PDE4B in cultured newt cells by Rolipram (Fig. 8). The newt cells were cultured as above. The control (no B2FF) and the B2FF treated cells gave rise to the sheet and the aggregate formation, respectively ( Fig. 8B-D). Rolipram application into B2FF culture media resulted in no aggregate formation ( Fig. 8E; n = 6/6). We grafted the + Rolipram/ + B2FF cell as a sheet since no aggregate formation could be obtained (Fig. 8A). The participation of the newt cells into cartilage was assessed at the digit stage. The grafted cells could survive and expand in the regenerate (Fig. 8F, G). The section revealed that a large number of mCherry + newt cells could be observed outside of the cartilages (Fig. 8H, I). Even though the mCherry + newt cells were located just by the regenerated cartilage, no participation of the grafted newt cells in the cartilage could be observed (Fig. 8J). All cell counts were shown in Table 2 and Fig. 8K. These results strongly suggest that Rolipram treatment inhibits re-differentiation from dermal fibroblasts to cartilaginous cells.
We next directly manipulated the Pde4b gene in axolotls. CRISPR/Cas9 systems allowed to generate mosaic Pde4b knockout animals (Pde4b crispants). We had not succeeded in generating homogenous Pde4b crispants. Five crispants were used and the knockout rate was assumed by ICE-analysis (30-58%, Fig. 9G). We labeled the dermal fibroblasts by GFP electroporation and traced the lineage during limb regeneration (Fig. 9A-E). The electroporation was performed prior to limb amputation. The electroporated limb was amputated 3 days after the electroporation, and the animals were kept until digits were apparent (Fig. 9A, D). In both the control limbs (n = 4) and the Pde4b crispant limbs (n = 8), GFP + cells could be seen in the regenerated limbs (Fig. 9A, B, D, E). The section revealed that GFP + cells could be seen in the Col2a1 + cartilage region and other connective tissues in the regenerated autopodial region in both the control animals and crispants (Fig. 9C, F). We counted GFP + cells in the regenerated autopodial region (Fig. 9G). Longitudinal sections were made throughout the regenerate. The GFP + and the GFP + Col2a1 + cells were counted on all sections. The control limbs, in which the knockout score was 0, showed that many GFP + cells differentiated into Col2a1 + cartilaginous cells. On the other hand, Pde4b crispants showed that much fewer GFP + cells participated into Col2a1 + cartilage. The alignment of the GFP + Col2a1 + / GFP + ratio with the knockout score calculated from ICE analysis revealed a strong correlation (R 2 = 0.912). Limbs from an animal having a higher knockout score showed a low integration rate of GFP + into a cartilaginous region. In contrast, limbs with a lower knockout score showed a relatively higher integration rate. This suggests that the Pde4b function relates to the conversion from dermal fibroblasts to cartilage cells in axolotl limb regeneration.

The cultured newt cells derived from the dermis
We cultured fibroblasts from a newts' limb skin (dermis), in which many types of cells exist. It is reasonably assumed that dermal fibroblasts are not homogenous, rather heterogeneous. Moreover, the determination of fibroblasts is still ambiguous. Thus, it is still difficult to determine the cultured cells we used precisely. In the present study, we obtained the constantly dividing fibroblasts (Fig. 1). During the process of establishing the cell line, it is very likely that certain cell populations were selected and survived. The RNA-Seq data revealed that the cultured fibroblasts express Col1a2, Vimentin, and Twist1 (Supplemental data 1), which are well-known marker genes as fibroblast marker genes. The expression profile reasonably suggests dermal fibroblasts were dominantly cultured in our experiment. However, culturing cells are dividing. Differentiated dermal fibroblasts in vivo are assumedly not actively dividing. Thus, our procedures in cell preparation might somehow transform cells. On the other hand, the cultured newt cells derived from the dermis did not show cartilage differentiation when the cells were grafted in the axolotl blastema (Fig. 2). This suggests that the cultured cells were not multipotent and that the culture condition did not provide multipotency. Further characterization of the cultured cells should be necessary to determine for precisely describing the cells we used.
We also admit that we could not maintain axolotl cells in the present study. However, there is one axolotl cell line, AL-1(ACT-1) [22]. AL1 cells are immortalized cells derived from mature limb fibroblasts. However, just a couple of cell lines have been successfully established so far, suggesting culturing axolotl cells are not easy. Developing much easier conditions for axolotl cells should be a great contribution to proceed investigation further.

The xenografting between axolotls and newts
The grafted newts cell could survive and differentiate into a couple of cell types in the axolotl limb (Fig. 2).  Such xenografting between an axolotl and a newt could be found in the histology of the amphibian regeneration study [23,24]. The grafted tissues and cells functioned physiologically in the xenografted environments. However, it is still unknown that cells from one species precisely behave normally in the other. On the other hand, limb regeneration can be induced by B2FF in both species [9]. B2FF, which we used in order to induce regeneration responses in urodele amphibians, are the recombinant proteins, whose amino acid sequences are derived from a mouse or a human. Of course, the axolotl B2FF genes can induce limb regeneration reactions when axolotl B2FF are electroporated [25]. Thus, mouse B2FF has been considered to activate the same or quite similar gene cascades in both species.  Even though the initial activation mechanism is identical, it is not known that the following mechanisms are identical or similar. For instance, the time to progress the regeneration stages is different, implying that the grafted newt cells receive inputs from outside at different timing in an axolotl blastema. We are not sure how the differences influence the grafted newt cells and their differentiation. The reason why we had to use xenografting in the present study is that we cannot find any good way to culture axolotl cells for a long time. There are ways to culture the axolotl cells [26,27]. However, it is still tough to have cells that can keep a proliferative state for a long time. To investigate a more focused and detailed mechanism of dermal fibroblasts' dedifferentiation, finding a way to culture axolotl cells for a long time is needed.

Gene selection
In this paper, we used a unique method for gene selection. We compared mouse fibroblasts with newt fibroblasts using cultured cells. The comparison would be controversial because mouse cells and newt cells are physiologically different. While acknowledging the differences in various physiological properties, we were able to find out candidate genes by comparing the downstream factors regulated by a common factor, B2FF. In this gene selection, we focused on genes that are inversely regulated by B2FF treatment in mouse and newt cells. This is based on the finding that the dynamics of dermal fibroblasts in the axolotl (limb) and mouse (fingertip) regeneration are different [28]. Urodeles can induce multipotent cells from the dermis, while mice cannot induce multipotent cells from the dermis. Therefore, we thought that there might be differences in gene expression during the generation of multipotent cells in the early stages of regeneration. However, it is possible that other systems, such as epigenetic regulation, are involved in the process of generating pluripotent cells from the dermis, and further research is needed to determine whether this is reflected in simple differences in transcriptome expression. Although a Fig. 9 Transdifferentiation from the dermis to cartilage in the mosaic Pde4b crispants. A-C the control animals (wt: wild-type). D-E The mosaic Pde4b crispants. A, D limb morphology. B, E The distribution of the GFP + cells. GFP plasmids were electroporated in the dermal region prior to amputation. C, F the merged images of the Col2a1, GFP, and the bright field view. Col2a1 and GFP were visualized by in situ hybridization and immunofluorescence, respectively. The scale bars in A and C are 1 mm and 300 μm, respectively. G the plot of the GFP + Col2a1+/GFP+ ratio along with the KO score calculated by ICE analysis multidimensional study is definitely necessary, the fact that we were able to find a functional molecule in the comparative analysis between the two species in this study provides a certain amount of positive endorsement for the gene selection method used in this study.
There is great significance in using cultured cells for gene selection. Most gene selections, which target to identify functional genes in limb regeneration, by RNA-Seq has been performed using regenerating blastemas. However, the use of regenerating blastemas means that a variety of cell types is included in it. Thus, the conventional method would not be preferable for a gene selection targeting a specific cell type, as in this study. In particular, dermal fibroblasts are known that only acquire multipotency in the process of limb regeneration. So, to identify such a reprogramming mechanism, a dermal fibroblast-specific analysis is necessary. Therefore, the development of the dermal fibroblast culture system will bring great progress in elucidating the process of acquisition of multipotency of dermal fibroblasts.

Pde4b function in transformation of dermal fibroblasts
We found the Pde4b gene from our gene selection. Pde4b was upregulated in the very early phase in axolotl limb regeneration (Fig. 4, Fig. 5D, J). Pde4b was downregulated after blastema cell emergence (Fig. 5P). This expression pattern is reasonable that Pde4b has a function in the dedifferentiation stage. Generally, limb regeneration and development share the same or similar gene cascades to form a patterned limb after blastema formation. Considering this, the cellular dedifferentiation process is unique and takes place before blastema formation. Moreover, the functions of cellular dedifferentiation should be down-regulated after blastema formation because blastema cells are going to be re-differentiated. Thus, the Pde4b gene expression pattern would be suitable as a factor involved in cellular dedifferentiation. Pde4b encodes an enzyme to hydrolyze cAMP to 5' AMP. Thus, PDE4B functions in cAMP regulation in limb regeneration. It is well known that there are cAMP-dependent pathways, such as the PKA-pathway [29]. Thus, disturbing PDE4B leads to influencing many intracellular signaling cascades. However, the detail mechanism downstream of PDE4B function needs further investigation.
Downregulation of cAMP in the very early phase of limb regeneration is likely important to fluctuate cellular differentiation in dermal fibroblasts. This is consistent with the previous report, in which a low level of cAMP in limb blastemas within 7 days after amputation was described [30,31]. Functions of the low level of cAMP at the beginning of limb regeneration have not been investigated. Our results suggest that a low level of cAMP contributes to fluctuating cellular differentiation in limb regeneration. Since Rolipram treatment affected cell mobility (Sup. Fig. 3), it is possible that the intracellular concentration of cAMP has a certain effect on cell mobility, which in turn affects the participation of cells in a blastema and subsequent differentiation. However, considering the result of Fig. 7, a system regulated by cAMP would affect cellular differentiation at a different hierarchy of cell mobility. the cAMP has a variety of functions in cells. Therefore, it will be important to examine the effects of cAMP on regeneration more comprehensively in the future. On the other hand, cAMP-dependent activities are unlikely to have severe influences on limb regeneration. PDE4B inhibition by Rolipram resulted in increasing in cAMP concentration in tissues and impaired the rate of transdifferentiation from dermal fibroblasts to cartilaginous cells. However, no skeletal pattern defects were observed. Previous reports strongly suggest that organ-level regeneration can be achieved by lineage-restricted cells, with no transdifferentiation [14]. Considering this, it is reasonable that a fluctuation of cellular differentiation has little influence on the outcome of limb regeneration. Although the fluctuation of cellular differentiation does have little influence on limb regeneration, the mechanism of induction of multipotent cells from differentiated tissues has been valuable. Blastema cells raised from dermal fibroblasts have been considered to be similar to limb bud cells. The connective tissue lineage-restricted multipotency is just like that of lateral place mesoderm derived limb bud cells. Thus, a mechanism of acquisition of multipotency in dermal fibroblasts might be closely related to the reactivation of developmental programs. We believe that reprogramming of cellular differentiation and/or rewinding developmental programs can be learned from amphibian organ regeneration.

Technical limitations
To approach the PDE4B function in limb regeneration, we used Rolipram. Rolipram treatment modified cAMP concentration in amphibian cells (Fig. 6A). Besides, interactive points between Rolipram and PDE4B are conserved among the species (Sup. Fig. 2). These suggest that Rolipram targets PDE4B properly in amphibian tissues. However, Rolipram might have other uncovered effects in axolotl tissues. To support the specific Rolipram effect on PDE4B in amphibian tissues, we designed a CRISPR/ Cas9 transgenic. We found a reasonable correlation between the phenotype and the knockout score (Fig. 9G). However, we could not generate homogenous Pde4b knockout animals in the study. This causes apparent technical limitations. The PDE4B function likely is involved in fluctuating cellular differentiation. On the other hand, it is still possible that Rolipram inhibits other uncovered signaling cascades leading to the fluctuation of cellular differentiation. We admit that further studies should be followed on the point.

Conclusion
Although it is known that fully differentiated cells undergo a de-differentiation process to form limb budlike cells during limb regeneration in urodele amphibians, the mechanism of the dedifferentiation has not been elucidated. The identification of the inducers of limb regeneration has made it possible to analyze the subsequent dedifferentiation phase. In this study, we investigated downstream factors of the regeneration-inducing factors, FGF2 + FGF8 + BMP2, and identified Pde4b, which may be involved in the dedifferentiation of fibroblasts during early regeneration.

Animals and cell culture
Axolotls (Ambystoma mexicanum) with a nose-to-tail length of 4-8 cm were used. Small axolotls (3-4 cm) were used for chemical treatment. The GFP axolotls were from the Ambystoma Genetic Stock Center (AGSC, Univ. of Kentucky). The axolotls were housed in aerated water at 22 °C. Photoshop CS5 software (Adobe, San Jose, CA, US). Newts (Pleurodeles waltl) were obtained from Hiroshima University (Amphibian research center) [32]. All animal experiments were conducted following the guidelines of Okayama University.
The cells from a TgSceI (CAG::TVA-mCherry-p2A-Cas9) were isolated to obtain mCherry positive cells. Mouse cells were from the skin of C3H newborn mice (1 day old). Skin fibroblasts were isolated as follows. The limb skin was removed from a mCherry + animal by forceps and micro scissors. The skin was minced well and incubated in 0.5% collagenase solution (0.5% collagenase (Wako-Fuji film, Osaka, Japan) in 70% phosphate-buffered saline (PBS)) for 6 h. Then, the equal amount of trypsin solution (0.5% Trypsin (Wako-Fuji film, Osaka, Japan) in PBS) was added and incubated for 30 min. The solution was filtered by 45 μm mesh and enzymatic activity were neutralized by adding the culture medium (10% FBS, 40% Glutamax DMEM (ThermoFisher, MA, USA), 50% water, 300 μg/ml Gentamycin (Nakarai tesque, Kyoto, Japan), and 10 mM HEPES, pH 7.5). A couple of centrifugation was taken for wash, and then the cells were cultured on the plate. FGF2 (#3139), FGF8 (#423-F8), and BMP2 (#355-BM) (R&D Systems, MN, USA) were used and added to the medium at 0.1ug/ml concentration. For the Rolipram treatment, we used 13 μM concentration. Cell count for calculating the growth rate was performed by Countess II (Thermo Fisher). For the RNA preparation, we collected the RNA samples from the 14-days cultured cells in each condition. The RNA preparation was performed using the TriPure reagent (Roche). For the analysis of cell mobility, we cultured 3 x10^5 cells on a plastic culture dish (Falcon #3001). The cultured cells were incubated for a day and then the record was started. The images were recorded by BZ-X800 (Keyence). Total 25 cells were traced from the 5 independent datasets. The cell mobility was traced for 1890 min with the 15 min interval. To obtain data of cellular mobility, Fiji/ImageJ combined with MTrackJ plugin was used. The statistical analysis was performed using Wilcoxon rank-sum test. The fundamental procedures were described previously [33].

CAGE-seq and RT-PCR
CAGE-seq was performed by DNAFORM (Kanagawa, Japan). Total RNA for CAGE-seq was extracted from the cells, which were treated with PBS or B2FF for 48 h. The samples were sent to DNAFORM (Kanagawa, Japan) in order to perform CAGE-seq. High-quality total RNA was prepared by Trizol (Invitorgen). Mapping of CAGE-tag sequences to the transcriptome assembly of Pleurodeles waltl (Trinity_Pwal_v2.fasta.gz, from iNewt website) was performed by Bowtie2. The read count was through the program featureCounts. The differential analysis was using DESeq2.Heatmap was described by Python. The quantitative RT-PCR was performed using primers listed in the Supplemental Table 1. The RNA samples were prepared from 2 independent samples. The quantitative RT-PCR analysis was performed by StepOne ™ (TermoFisher) and KAPA DNA polymerase (#KK4600, NIPPON genetics, Tokyo, Japan).

Pde4b knockout by CRISPR/Cas9 and ICE analysis
The animals, in which the Pde4b gene was heterogeneously edited, were generated as described [32]. The two guide RNAs were prepared as follows; gPde4b ver1: GGA GGA GCT GGA CTG GTG CC, gPde4b ver2: GTC CGT GTG CTT GTT TGC AG. The guide RNA and sgRNA were purified using a CUGA gRNA synthesis kit. The synthesized gRNA was incubated with Cas9 proteins (IDT) for 30 min at room temperature before injection.
Gene knockout score was calculated by ICE analysis (https:// ice. synth ego. com/#/). The axolotl genome for the ICE analysis was extracted from the limb skin. The small piece of the forelimb skin was removed from an upper arm and the genome was purified using the QIAGEN genome extraction kit (#69506). The standard Sanger sequence was performed to the amplified genomic PCR fragment using the following primers; for: ATG ATG AAG GAG CAC TGC CCC ACC; rev: CTT GTT CGA TGC CAT CTC GCT GAC G.