Fluctuation of cellular differentiation in limb regeneration is regulated by Pde4b in urodele amphibians

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. Interspecies comparative analysis clarified that Pde4b upregulation by B2FF specifically took place in the newt cells. Blocking PDE4B signaling by Rolipram 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.

6 cartilage properties. Cell surface-located and spotted red signals were observable, consisting of the 136 fluorescent pattern observed in Fig. 1F (Fig. 2F). Furthermore, it was suggested that dermis-derived cells 137 can redifferentiate into various connective cell types 10,12,13 . To clarify whether mCherry + cells were 138 observed in other connective tissues, we focused on the tendons. Tenascin is a marker of tendons and 139 ligaments and is expressed in the connecting region between skeletal structures and muscles in 140 amphibians 16 . Tenascin expression was visualized by immunofluorescence (Sup. Fig.1). Tenascin 141 expression was observed in the peripheral region of the epiphysis and some fibroblast near the epiphysis 142 (Sup. Fig.1A). Confocal observation confirmed that tenascin was located by the mCherry + cells, 143 suggesting that the mCherry + cell became the tendon cell (Sup. Fig.1B, C). Those data suggest that the 144 grafted B2FF treated newt fibroblasts participated in some connective tissue lineage. 145 Next, the comparative RNA-Seq analysis was conducted in order to identify the genes related 146 to the acquisition of multipotency. To investigate gene dynamics prior to the formation of cell aggregate, 147 cells 48 hours after B2FF treatment were harvested for comparison. Furthermore, to focus on the specific 148 gene dynamics in the regeneration-competent animals, mouse cells were used as a comparison. Mouse 149 dermal fibroblasts were collected from neonates and 5 passages were undergone before the experiment. 150 B2FF treatment was performed for 48 hours. RNA-seq was performed by CAGE-seq. To compare gene 151 expression between newts and mice, gene symbols were used in this comparison. This was because of 152 the lack of fixed genome information of Pleurodeles waltl. Thus, a precise orthologue determination 153 could not be fixed. We found 4185 genes were commonly expressed in the cultured cells of both species 154 Among those genes, we selected the genes showing FC<-0.5 and FC>0.6. We found that there were 42 157 genes affected by B2FF treatment (Fig. 3B, Supplemental Data 1). We found out that Col1A2, which is a 158 major component of the dermis, was down-regulated by B2FF treatment in both species (Supplemental 159 Data 1). This is consistent with the report, in which an axolotl limb blastema is a less collagenous 160 structure 17 . We further focused on the genes which react oppositely to B2FF in the two species (Fig. 3B). 161 Fblim, Loxl2, Pde4b, Spry1, Timp3 were upregulated in newt cells and down-regulated in mouse cells by 162 B2FF treatment. Adra2a showed opposite dynamics. To confirm the RNA-seq results, quantitative 163 RT-PCR (qPCR) analysis was performed (Fig. 3C). The gene expression profiles were consistent with 164 the results from RNA-seq. (Fig. 3C). 165 These 6 genes dynamics were also investigated in axolotl blastemas (Fig. 4). As mentioned, 166 axolotls have many advantages in vivo experiments. Thus, we investigated the gene expression patterns 167 in axolotl blastemas in order to perform further experiments in axolotls. qPCR analysis revealed that 168 quick upregulation after limb amputation (Fig. 4). Only Pde4 expression was settled down at 10 days 170 post-amputation (dpa). Spry1 and Timp3 have a late-activation profile in axolotl limb blastemas. Adra2a 171 expression was not detectable throughout the period we tested. These gene expression patterns were 172 confirmed by in situ hybridization. (Fig. 5). Consistently, Adra2a was not detectable at 2, 5, and 10 dpa 173 (Fig. 5A, A', G, G', M, M'). Fblim1 expression could be recognized in the amputation region from 2 dpa 174 to 10 dpa (Fig. 5B, H, N). The signal of Fblim1 could be observed at the border of the amputation plane 175 (Fig, 5B'), but the signal was weakened in the later time points in the stump regions ( Fig. 5H', N'). 176 Loxl2 expression could also be detected from 2 dpa ( which showed early upregulation and late down-regulation, was suitable for this criteria. 199 We next attempted to inhibit Pde4b functions in limb regeneration using a chemical inhibitor, 200 Rolipram. Rolipram binds to the catalytic sites of PDE4B at several amino acids, where are 100% 201 8 conserved between human PDE4B and axolotl PDE4B (Sup. Fig. 2). It is well known that PDE4B has a 202 function to hydrolyze cAMP to 5'AMP 19 . To confirm the inhibitory effects to PDE4B in axolotl tissues, 203 cAMP concentration in axolotl limbs was measured by ELISA (Fig. 6A). Consistently, cAMP 204 concentration in limb tissues was upregulated by the 7 day-Rolipram treatment ( Fig. 6A). This suggests 205 that Rolipram can effectively inhibit hydrolysis of cAMP by PDE4B in axolotl limbs. Next, we 206 investigated the dermal fibroblast's transdifferentiation into cartilaginous cells in the presence of 207 Rolipram ( Fig. 6B-E). The limb skin from a GFP animal was transgrafted onto a normal animal, and the 208 grafted limb was kept a week for the recovery from the grafting damages (Fig. 6B). Then, the limb was 209 amputated, and the limb-amputated animals were kept in Rolipram-containing water until digits were 210 identifiable ( Fig. 6B, C1-C3). The GFP positive domain was expanded from the amputation stump to the 211 digit tips of the regenerate ( Fig. 6C1-3). The regenerates were fixed and sectioned. Col2A1 + cartilage 212 cells were revealed by in situ hybridization and the location of the GFP + cells was revealed by 213 immunofluorescence (Fig. 6D, E). In the control samples, GFP + cells were observable in the Col2A1 + 214 region (Fig. 6D, Table 1). Five limbs were obtained and sectioned. Sectioning was performed on the 215 entire limb along the dorsoventral axis. Only sections with GFP + cells in the mesenchymal region on the 216 prepared sections were extracted, and the GFP + cells were counted. We found 3321 GFP positive cells in 217 104 extracted sections, of which 149 GFP positive cells were Col2A1 + . In the Rolipram-treated animals, 218 GFP + cells were little observed in the Col2A1 + region (Fig. 6E, Table 1). We obtained 36 sections from 7 219 Rolipram-treated samples. Although the number of sections obtained from a single limb sample was 220 about the same as the control, the number of sections to be extracted in the Rolipram-treated samples 221 was lower as compared to the control. This was because of the poor participation of GFP+ cells in the 222 regenerates compared to controls. Using the same method of cell counting as the control, we observed 223 only one Col2A1 + cell out of 462 GFP + cells (Table 1). It is also noteworthy that cartilage formation in 224 regenerates was not influenced by the Rolipram treatment. These suggest that Rolipram treatment 225 increases cAMP concertation in tissues resulting in suppression of the transdifferentiation from dermal 226 fibroblasts to cartilaginous cells. 227 We further investigated increasing in cAMP concentration impaired cartilage 228 transdifferentiation from dermal fibroblasts. Dibutyryl-cAMP is a cell-permeable cAMP analog that 229 activates cAMP-dependent protein kinases 20 . Similarly, GFP + skin was transgrafted onto a normal 230 animal and the GFP skin-grafted limb was amputated to trace the lineage of GFP + cells in the absence or 231 presence of dibutyryl-cAMP (Fig. 7A). The amputated limbs were kept until the regenerates reached the 232 digit stage (Fig. 7B, F). To visualize GFP and Col2A1, we performed immunofluorescence on the 233 identical sections (Fig. 7C-I). In the control sample, GFP + cells in the cartilaginous region could be 9 detected as well as epidermis and dermis (Fig. 7C-E). On the other hand, the Dibutyryl-cAMP treated 235 limbs showed a little number of GFP + cells in the cartilaginous region ( Fig. 7G-I). We plotted the rate of 236 the GFP + /Col2A1 + in the regenerates (Fig. 7J). We counted 13 sections from 8 independent animals in 237 the control and 21 sections from 17 independent animals in the Dibutyryl cAMP-treated animals. It is 238 noteworthy that the two exceptional plots in the Dibutyryl cAMP-treated samples were derived from an 239 identical animal. These results consistently suggest that cAMP concentration influences a fluctuation of 240 differentiation of dermal fibroblasts. 241 Next, we attempted to inhibit PDE4B in cultured newt cells by Rolipram (Fig.8). The newt 242 cells were cultured as above. The control (no B2FF) and the B2FF treated cells gave rise to the sheet and 243 the aggregate formation, respectively ( Fig. 8B-D). Rolipram application into B2FF culture media 244 resulted in no aggregate formation ( Fig. 8E; n=6/6). We grafted the + Rolipram/ + B2FF cell as a sheet 245 since no aggregate formation could be obtained (Fig. 8A). The participation of the newt cells into 246 cartilage was assessed at the digit stage. The grafted cells could survive and expand in the regenerate 247 (Fig. 8F, G). The section revealed that a large number of mCherry + newt cells could be observed outside 248 of the cartilages (Fig. 8H, I). Even though the mCherry + newt cells were located just by the regenerated 249 cartilage, no participation of the grafted newt cells in the cartilage could be observed (Fig. 8J). All cell 250 counts were shown in Table 2 and Fig. 8K. These results strongly suggest that Rolipram treatment 251 inhibits re-differentiation from dermal fibroblasts to cartilaginous cells. 252 We next directly manipulated the Pde4b gene in axolotls. CRISPR/Cas9 systems allowed to 253 generate mosaic Pde4b knockout animals (Pde4b crispants). We had not succeeded in generating 254 homogenous Pde4b crispants. Five crispants were used and the knockout rate was assumed by 255 ICE-analysis (30-58%, Fig. 9G). We labeled the dermal fibroblasts by GFP electroporation and traced 256 the lineage during limb regeneration (Fig. 9A-E). The electroporation was performed prior to limb 257 amputation. The electroporated limb was amputated 3 days after the electroporation, and the animals 258 were kept until digits were apparent (Fig. 9A, D). In both the control limbs (n=4) and the Pde4b crispant 259 limbs (n=8), GFP + cells could be seen in the regenerated limbs (Fig. 9A, B, D, E). The section revealed 260 that GFP + cells could be seen in the Col2a1 + cartilage region and other connective tissues in the 261 regenerated autopodial region in both the control animals and crispants (Fig. 9C, F). We counted GFP + 262 cells in the regenerated autopodial region (Fig. 9G). Longitudinal sections were made throughout the 263 regenerate. The GFP + and the GFP + Col2a1 + cells were counted on all sections. The control limbs, in 264 which the knockout score was 0, showed that many GFP + cells differentiated into Col2a1 + cartilaginous 265 cells. On the other hand, Pde4b crispants showed that much fewer GFP + cells participated into Col2a1 + 266 cartilage. The alignment of the GFP + Col2a1 + /GFP + ratio with the knockout score calculated from ICE 267 analysis revealed a strong correlation (R 2 =0.912). Limbs from an animal having a higher knockout score 268 showed a low integration rate of GFP + into a cartilaginous region. In contrast, limbs with a lower 269 knockout score showed a relatively higher integration rate. This suggests that the Pde4b function relates 270 to the conversion from dermal fibroblasts to cartilage cells in axolotl limb regeneration.

The cultured newt cells derived from the dermis 275
We cultured fibroblasts from a newts' limb skin (dermis), in which many types of cells exist. It 276 is reasonably assumed that dermal fibroblasts are not homogenous, rather heterogeneous. Moreover, the 277 determination of fibroblasts is still ambiguous. Thus, it is still difficult to determine the cultured cells we 278 used precisely. In the present study, we obtained the constantly dividing fibroblasts (Fig. 1). During the 279 process of establishing the cell line, it is very likely that certain cell populations were selected and 280 survived. The RNA-Seq data revealed that the cultured fibroblasts express Col1a2, Vimentin, and Twist1 281 (Supplemental data 1), which are well-known marker genes as fibroblast marker genes. The expression 282 profile reasonably suggests dermal fibroblasts were dominantly cultured in our experiment. However, 283 culturing cells are dividing. Differentiated dermal fibroblasts in vivo are assumedly not actively dividing. 284 Thus, our procedures in cell preparation might somehow transform cells. On the other hand, the cultured 285 newt cells derived from the dermis did not show cartilage differentiation when the cells were grafted in 286 the axolotl blastema (Fig. 2). This suggests that the cultured cells were not multipotent and that the 287 culture condition did not provide multipotency. Further characterization of the cultured cells should be 288 necessary to determine for precisely describing the cells we used. 289 290

The xenografting between axolotls and newts 291
The grafted newts cell could survive and differentiated into a couple of cell types in the axolotl 292 limb (Fig. 2). Such xenografting between an axolotl and a newt could be found in the histology of the 293 amphibian regeneration study 21,22 . The grafted tissues and cells functioned physiologically in the 294 xenografted environments. However, it is still unknown that cells from one species precisely behave 295 normally in the other. On the other hand, limb regeneration can be induced by B2FF in both species 8 . 296 B2FF, which we used in order to induce regeneration responses in urodele amphibians, are the 297 recombinant proteins, whose amino acid sequences are derived from a mouse or a human. Of course, the 298 axolotl B2FF genes can induce limb regeneration reactions when axolotl B2FF are electroporated 23 . 299 Thus, mouse B2FF has been considered to activate the same or quite similar gene cascades in both 300 species. Even though the initial activation mechanism is identical, it is not known that the following 301 mechanisms are identical or similar. For instance, the time to progress the regeneration stages is different, 302 implying that the grafted newt cells receive inputs from outside at different timing in an axolotl blastema. 303 We are not sure how the differences influence the grafted newt cells and their differentiation. 304 The reason why we had to use xenografting in the present study is that we cannot find any 305 good way to culture axolotl cells for a long time. There are ways to culture the axolotl cells 24,25 . 306 However, it is still tough to have cells that can keep a proliferative state for a long time. To investigate a 307 more focused and detailed mechanism of dermal fibroblasts' dedifferentiation, finding a way to culture 308 axolotl cells for a long time is needed. 309 310

Gene selection 311
In this paper, we used a unique method for gene selection. We compared mouse fibroblasts 312 with newt fibroblasts using cultured cells. The comparison would be controversial because mouse cells 313 and newt cells are physiologically different. While acknowledging the differences in various 314 physiological properties, we were able to find out candidate genes by comparing the downstream factors 315 regulated by a common factor, B2FF. In this gene selection, we focused on genes that are inversely 316 regulated by B2FF treatment in mouse and newt cells. This is based on the finding that the dynamics of 317 dermal fibroblasts in the axolotl (limb) and mouse (fingertip) regeneration are different 26 . Urodeles can 318 induce multipotent cells from the dermis, while mice cannot induce multipotent cells from the dermis. 319 Therefore, we thought that there might be differences in gene expression during the generation of 320 multipotent cells in the early stages of regeneration. However, it is possible that other systems, such as 321 epigenetic regulation, are involved in the process of generating pluripotent cells from the dermis, and 322 further research is needed to determine whether this is reflected in simple differences in transcriptome 323 expression. Although a multidimensional study is definitely necessary, the fact that we were able to find 324 a functional molecule in the comparative analysis between the two species in this study provides a 325 certain amount of positive endorsement for the gene selection method used in this study. 326 327

Pde4b function in transformation of dermal fibroblasts 328
We found the Pde4b gene from our gene selection. Pde4b was upregulated in the very early phase in 329 axolotl limb regeneration (Fig. 4, Fig. 5D, J). Pde4b was downregulated after blastema cell emergence 330 (Fig. 5P). This expression pattern is reasonable that Pde4b has a function in the dedifferentiation stage. 331 Generally, limb regeneration and development share the same or similar gene cascades to form a 332 patterned limb after blastema formation. Considering this, the cellular dedifferentiation process is unique 333 and takes place before blastema formation. Moreover, the functions of cellular dedifferentiation should 334 be down-regulated after blastema formation because blastema cells are going to be re-differentiated. 335 Thus, the Pde4b gene expression pattern would be suitable as a factor involved in cellular 336

dedifferentiation. 337
Pde4b encodes an enzyme to hydrolyze cAMP to 5'AMP. Thus, PDE4B functions in cAMP 338 regulation in limb regeneration. It is well known that there are cAMP-dependent pathways, such as the 339 PKA-pathway 27 . Thus, disturbing PDE4B leads to influencing many intracellular signaling cascades. 340 Downregulation of cAMP in the very early phase of limb regeneration is likely important to 341 fluctuate cellular differentiation in dermal fibroblasts. This is consistent with the previous report, in 342 which a low level of cAMP in limb blastemas within 7 days after amputation was described 28,29 . 343 Functions of the low level of cAMP at the beginning of limb regeneration have not been investigated.