Retinoic acid signaling spatially restricts osteoblasts and controls ray-interray organization during zebrafish fin regeneration

Zebrafish regenerate amputated fins by establishing lineage-restricted blastema cells (Gemberling et al., 2013; Tanaka and Reddien, 2011). The zebrafish caudal fin possesses 16-18 fin rays, each consisting of two segmented and opposing hemirays of acellular bone that surround a core of fibroblasts, osteoblasts, pigment cells, arterial blood vessels and nerves (Fig. 1A) (Akimenko et al., 2003). Fin rays are separated by boneless interray tissue, composed of fibroblasts, venous blood vessels, pigment cells and nerves. The principles that compel regenerating fin rays to respect ray-interray boundaries, therefore confining regenerating bone to extend the existing rays, are still unknown. Upon fin amputation, osteoblasts that cover the hemiray surfaces dedifferentiate into proliferating preosteoblasts and migrate into the nascent blastema, where they align at proximal lateral positions (Knopf et al., 2011; Sousa et al., 2011; Stewart and Stankunas, 2012). Thus, preosteoblasts form a layer between the basal epidermal layer and fibroblast-derived blastema cells. The distal blastema remains devoid of preosteoblasts. During subsequent regenerative outgrowth, proliferating preosteoblasts at the distal leading edge become differentiating osteoblasts in more proximal parts (Stewart et al., 2014). Notably, neither osteoblasts nor non-osteoblastic blastema cells mix with adjacent interray cells (Stewart and Stankunas, 2012). Thus, the regenerating fin consists of repeating blastema units dedicated to each fin ray that are separated by regenerating interrays (Fig. 1A).

During fin regeneration, fin rays respect ray-interray boundaries. An interesting exception occurs after amputation close to a bifurcation site (short cut, Fig. 1B). In such regenerates, sister rays ignore ray-interray boundaries and fuse (Fig. 1C) (Laforest et al., 1998), whereby the probability for ray fusion increases as the distance between them decreases (Fig. 1D). Using the osc:gfp line, which allows detection of preosteoblasts in the early blastema (Knopf et al., 2011; Sousa et al., 2011), we found ectopic preosteoblasts in the interray separating the two sister rays at 2 dpa (Fig. 1E). This finding reveals that sister ray fusion is due to preosteoblasts spreading into the interray. We were interested in the mechanisms that cause preosteoblasts to respect ray-interray boundaries. shha is expressed within the basal epidermal layer adjacent to pre- and differentiating osteoblasts (supplementary material Fig. S1) (Laforest et al., 1998; Lee et al., 2009; Quint et al., 2002; Zhang et al., 2012). We found a similar expression pattern for the RA-degrading enzyme cyp26a1 (Fig. 1F,G). By contrast, the RA-synthesizing enzyme aldh1a2 is expressed in fibroblast-derived blastema cells (Fig. 1G) (Blum and Begemann, 2012). Although proximal fibroblast-derived blastema cells express cyp26b1 (Blum and Begemann, 2015), it is unlikely that RA diffusion into adjacent epidermal cells is efficiently prevented. We thus suspected that cyp26a1-expressing cells provide niches of low RA levels that might facilitate expression of RA-sensitive genes.

It has been shown that immersion of fish in RA reduces shha expression (Laforest et al., 1998). However, RA administration via immersion induces epidermal cell death (Ferretti and Géraudie, 1995; Géraudie and Ferretti, 1997); therefore, it has remained unclear whether this was a specific effect. We used intraperitoneal (IP) injection of RA, which efficiently enhances RA levels in the adult fin without causing cell death (Blum and Begemann, 2012), and found decreased expression of shha and of the Hh receptor and target ptch2 6 h after injection at 4 dpa (Fig. 1H). Injection of R115866, a selective antagonist of Cyp26 enzymes (Hernandez et al., 2007; Stoppie et al., 2000), also caused downregulation of shha and ptch2 (Fig. 1H,I), indicating that shha expression requires RA degradation. Accordingly, overexpression of cyp26a1 in hsp70I:cyp26a1 fish (Blum and Begemann, 2012) at 4 dpa resulted in enhanced and laterally extended shha expression within the basal epidermal layer directly at the end of a single heat shock (Fig. 1J,K). We did not observe shha expression in interrays. We propose that this was due to the requirement for Fgfs (and putative other blastema-derived signals) for shha expression (Lee et al., 2009).

Short-cut amputations resulted in shared cyp26a1 expression domains between sister rays at 1.5 dpa (Fig. 1L), indicating that fusion of the RA-degrading niches precedes spreading of preosteoblasts into interrays. To examine how preosteoblasts behave if efficient lowering of RA levels in the cyp26a1-niches fails, we injected fish with RA or R115866, starting with the first IP directly after amputation (normal cut). We observed spreading of preosteoblasts into interrays in RA- and R115866-injected osc:gfp fish at 2.5 dpa (Fig. 2A; data not shown), indicating that preosteoblasts failed to align at proximolateral parts of the blastema and disregarded ray-interray boundaries. Of note, formation of a distinct basal epidermal layer was not inhibited (supplementary material Fig. S2A). As R115866 treatment also reduced Cyp26b1 activity in fibroblast-derived blastema cells, we cannot exclude that misguidance of preosteoblasts in R115866- or RA-treated fish was at least partially due to enhanced RA levels in the mesenchyme. However, the expression pattern of cyp26a1 strongly supports a model in which alignment, and thereby spatial restriction of preosteoblasts, is controlled by signals from cyp26a1-expressing epidermal cells.

Interestingly, treatment with RA or R115866 also caused expansion of the blastema marker msxb (Akimenko et al., 1995) into interrays (Fig. 2B; supplementary material Fig. S2B). 5-ethynyl-2′-deoxyuridine (EdU) labeling further suggested that these were cycling blastema cells (Fig. 2C). We did not observe an enlarged preosteoblast or msxb⁺ cell population in RA- or R115866-treated fish (Fig. 2A,B; supplementary material Fig. S2B), thus making it unlikely that spreading into interrays was simply due to increased cell numbers.

Blood vessels regenerate into the blastema accompanied by anastomosis between the injured arteries and veins of the same ray (Huang et al., 2003). Intriguingly, wounded arteries and veins had formed connections with vessels of neighboring rays in RA- and R115866-treated fli:gfp fish (Lawson and Weinstein, 2002) at 2.5 dpa (Fig. 2D; data not shown). Together, these data show that also non-osteoblastic cell types ignore ray-interray boundaries if efficient RA degradation in the proximal basal epidermal layer fails.

Expansion of msxb expression into interrays was also observed in regenerates that lacked preosteoblasts, but showed expression of cyp26a1 (supplementary material Fig. S3). This was achieved by treating fish with the Hh inhibitor cyclopamine from 24 hpa onwards, a treatment condition that is expected to inhibit preosteoblast proliferation in the stump. We thus conclude that, rather than signals from cyp26a1-expressing cells, aligning preosteoblasts themselves provide spatial orientation for non-osteoblastic blastema cells.

To test how disrespect of ray-interray boundaries by blastema cells affects the overall organization of the regenerating fin, we gave the last RA or R115866 injection at 1.5 dpa and examined the long-term consequences for regeneration. Of note, owing to autoregulatory feedback mechanisms of RA signaling, cyp26a1 expression became upregulated in the entire regenerate during RA or R115866 treatment (data not shown). This upregulation can only counterbalance small fluctuations in RA levels and can therefore be neglected during the treatment period. However, ectopic expression persisted for some days after termination of treatment (supplementary material Fig. S4A), and was therefore expected to allow for shha expression in the entire regenerate. Consistently, we detected shha expression in both rays and interrays in R115866-treated shh:gfp fish (Ertzer et al., 2007) 2 days after treatment was stopped at 4 dpa (supplementary data Fig. S4B). As heat shock-induced overexpression of cyp26a1 at 4 dpa is insufficient to induce shha expression in interrays (Fig. 1J,K), this result suggests that shha expression in interrays in R115866-treated fish was due to ectopic blastema cells in the interray region. Thus, this finding supports a model in which shha expression is spatially confined by RA-degrading epidermal cells and signals from underlying blastema cells.

In fins of R115866-treated fish, regeneration was subsequently irreversibly blocked and ectopic bone had formed in interrays at 11 dpa (Fig. 2E). However, bone matrix did not seal the wound (supplementary material Fig. S5), indicating that inhibition of regeneration was not due to a mechanical block but rather due to mispatterning of the early regenerate. In RA-treated fins, neighboring rays were occasionally connected by bony bridges at the amputation site (Fig. 2F), but otherwise, regeneration proceeded normally. This weaker phenotype was probably due to rapid clearance of excess RA after treatment had stopped. When fins were amputated within the third segment distal to the first branching point (long cut), RA treatment caused fusions of sister rays (supplementary material Fig. S6A). This is consistent with results obtained previously by RA treatment via immersion (Géraudie and Ferretti, 1997; Géraudie et al., 1995; White et al., 1994). We next determined the distance range at which formation of ectopic bone occurs, and found a much greater range in RA-treated fish (supplementary material Fig. S6B). Fusion of sister rays had sometimes been interpreted as a proximalization of the regenerate (White et al., 1994). However, our data show that sister rays fuse if preosteoblasts spread into interrays, which can be caused either by insufficient lowering of RA levels in the proximal basal epidermal layer or by induction of two RA-degrading domains in close proximity. Thus, our findings provide a more parsimonious explanation for sister ray fusion that is not based on putative changes in proximodistal information.

Proliferation of preosteoblasts has been suggested to be controlled by Shha (Laforest et al., 1998; Lee et al., 2009; Quint et al., 2002; Zhang et al., 2012). We found reduced osteoblast proliferation within 12 h of cyclopamine treatment (Fig. 3A). Osteoblast differentiation was unaffected (supplementary data Fig. S7). Moreover, we did not detect TUNEL⁺ osteoblasts in control or cyclopamine-treated fish (data not shown). Hence, even though RA signaling promotes osteoblast proliferation (Blum and Begemann, 2015), prolonged experimental RA exposure should cause downregulation of shha and consequently lead to a reduction of osteoblast proliferation. Indeed, osteoblast proliferation was reduced 12 h after RA injection at 3 dpa (Fig. 3B). Neither osteoblasts nor other cells undergo cell death after RA treatment via IP injections (Blum and Begemann, 2012). Concomitant activation of Hh signaling, using the Smoothened agonist SAG, rescued osteoblast proliferation (Fig. 3C), thus confirming that decreased proliferation upon RA treatment was due to impaired Hh signaling. Together, these findings indicate that Shha from cyp26a1-expressing epidermal cells promotes proliferation of adjacent osteoblasts. Interestingly, cyclopamine treatment has been reported to block proliferation of fibroblast-derived blastema cells (Lee et al., 2009), for which ptch expression has not been demonstrated. Accordingly, prolonged RA treatment downregulated proliferation of fibroblast-derived blastema cells, and concomitant SAG treatment could rescue this effect (supplementary material Fig. S8). Although we cannot exclude a direct effect, Hh signaling might indirectly promote proliferation of other cell types via osteoblasts.

Besides a requirement for shha expression, Fgf signaling has been shown to exclude shha from distal regions (Lee et al., 2009), suggesting that Fgf signaling restricts shha to the proximal basal epidermal layer by repressing cyp26a1. We manipulated Fgf signaling at 3 dpa by either overexpressing a dominant negative Fgfr1 (hsp70I:dn-fgfr1; Lee et al., 2005) or a constitutively active Ras (hsp70I:v-ras; Lee et al., 2009) and quantified cyp26a1 expression. cyp26a1 was downregulated in hsp70I:v-ras fish 2 h after a single heat shock and upregulated in hsp70I:dn-fgfr1 fish at the end of a single heat shock (Fig. 3D), demonstrating that Fgf signaling inhibits cyp26a1 expression. Thus, proximal expansion of Fgf signaling should result in proximal regression of cyp26a1 expression. We took advantage of the finding that Fgf activity expands more proximally in fins that had been amputated at a more proximal position and retracts distally as regeneration proceeds (Lee et al., 2005, 2009). We found that amputation at a proximal position results in a proximal shift of the cyp26a1 expression domain (Fig. 3F,G). In return, cyp26a1 expression shifted distally during the course of regeneration (Fig. 3E). Notably, also the distal limit of aligned preosteoblasts was always adjacent to the distal limit of cyp26a1 expression (Fig. 3E-H), suggesting that cyp26a1-expressing cells spatially confine preosteoblasts also along the proximodistal axis.

During fin regeneration the ray-interray organization has to be faithfully re-established in order to rebuild an exact copy of the lost fin parts and to ensure proper function of the regenerated fin. Here, we show that disrespect of the ray-interray boundaries by preosteoblasts and other blastema cells in the nascent blastema has adverse consequences for subsequent fin patterning and may disrupt the whole regeneration process. Our findings support a model in which signals from RA-degrading niches established by Cyp26a1 in the basal epidermal layer ensure the appropriate initial alignment of preosteoblasts in the nascent blastema (Fig. 4) and compel blastema cells to respect ray-interray boundaries. Furthermore, during regenerative outgrowth, Cyp26a1 activity remains important to facilitate Shha-promoted proliferation in adjacent preosteoblasts (Fig. 4).

Zebrafish were raised under standard conditions at 27-28°C. Caudal fins of 3- to 18-month-old fish were amputated along the dorsoventral axis, intersecting the median rays halfway for normal cuts, at ∼30% ray length for proximal cuts, at ∼70% for distal cuts, within 1-2 segments distal to the first branching point for short cuts and within the third segment away for long cuts. Heat shocks were performed at 38°C for 1 h. Approximately 20 µl RA (all-trans RA, Sigma) or R115866 (Janssen Pharmaceutica) were intraperitoneally injected into size-matched siblings every 12 h. RA: 1 mM in 1% DMSO/PBS. R115866: 0.67 mM 10% DMSO/PBS. For cyclopamine treatment, fish were incubated in 5 μM cyclopamine (Sigma), 0.1% EtOH, 0.1% DMSO in E3-medium (HEPES-buffered at pH 7.4). Cyclopamine was exchanged daily. SAG (Calbiochem) treatment was performed using 5 μM SAG in E3-medium. SAG stock solution was prepared in water. Control fish were always treated with an equivalent concentration of the drug solvent. All animal experiments were approved by the state of Baden-Württemberg, Germany.

Osteoblast differentiation was examined by measuring the GFP-free distal region in osx:gfp fish (Olsp7:nlsgfp; Spoorendonk et al., 2008). Gene expression levels were analyzed by qPCR (for primers see supplementary material Table S1), cell death by TUNEL staining and cell proliferation by EdU labeling. Cryosectioning was used to produce longitudinal and transverse fin sections. Further information concerning these methods, as well as descriptions of length measurements and cell number quantifications, imaging, immunohistochemistry, in situ hybridization and Alizarin Red staining can be found in the supplementary material methods.

Student's t-test was used to test significance of differences. The numbers of specimens used are given in supplementary material Table S2.

We thank S. Schulte-Merker, K. Poss and U. Strähle for transgenic fish lines, Janssen Pharmaceutica for the R115866 compound, and A. Pfeifer, I. Zerenner-Fritsche, S. Leuschner and K.-H. Pöhner for fish care.