Sequential and Opposing Activities of Wnt and BMP Coordinate Zebrafish Bone Regeneration

1Among mammalian organs, bones have unusually effective repair mechanisms, as demonstrated by continuous bone remodeling throughout life and the scarless healing of some fractures. Still, traumatic injuries or disease frequently exceed the innate regenerative capacities of human bone. Envisioned restorative bone therapies include, for example, engineered cell scaffolds combined with bone grafts and the delivery of manipulated stem cells including patient-specific induced pluripotent stem cells. An appealing alternative is to recapitulate mechanisms observed in animals with remarkable capacities for self-repair, including fish and salamanders.

2Osteoblasts are specialized bone producing cells that deposit a unique extracellular matrix, the osteoid, that mineralizes to form mature bone. Two transcription factors, Runx2 and sp7/Osterix (Osx), are key determinants of the osteoblast lineage. Runx2-null mice lack all ossification. Similarly, Sp7-deficient mice do not form bone due to a failure of osteoblasts to differentiate. Runx2 functions upstream of Sp7; Runx2-deficient mice fail to express Sp7, whereas Sp7-null mice retain Runx2 expression. Transcriptional control of Runx2 and Sp7 is mediated via cell signaling including by the Wnt and bone morphogenetic protein (BMP) pathways, both central components of bone developmental regulatory networks.

3Wnts are secreted proteins that function in many biological processes, including development and cancer. Wnts bind to Frizzled (Fz) and Lrp5/6 coreceptors culminating in the stabilization and nuclear translocation of β-catenin. Nuclear β-catenin functions with Tcf/Lef transcription factors to affect gene expression. Given Wnt’s signaling potency in influencing cell behavior and lineage decisions, the pathway is under tight control. For instance, the Dickkopf family of secreted Wnt antagonists moderate Wnt activity by binding to and blocking Lrp5/6. Deletion of mouse β-catenin in early mesenchymal precursors results in loss of Runx2 and Sp7 expression and a corresponding failure of bone formation, suggesting osteogenesis requires Wnt/β-catenin signaling.

4The BMP pathway is an additional key regulator of osteogenesis. Secreted BMPs bind and activate specific BMP serine/threonine kinase receptors (BMPRs). BMPRs phosphorylate Smad1/5/8 transcription factors, inducing their translocation to the nucleus where they activate downstream genes. BMPs are required for osteoblast-lineage commitment of cranial mesenchyme, and loss of both Bmp2 and Bmp4 results in a severe defect in osteoblast differentiation likely due to a failure of Runx2-expressing cells to upregulate Sp7. These and earlier studies led to the development of recombinant BMPs as therapies to augment bone healing, albeit with debated clinical benefits.

5Mammalian bone fracture repair is mediated by bone-marrow-derived mesenchymal stem cells that differentiate into osteoblasts and subsequently produce remineralized bone. Several signaling pathways, including Wnt and BMP, have been implicated in these repair processes, but their modes of action at the cellular and molecular level are largely unknown. For example, it is unclear whether Wnts and BMPs influence discrete osteoprogenitor populations or how expression of Runx2 and Sp7 the master regulators of osteogenesis, are coordinated during the repair process. Even more confounding are observations that, whereas both Wnts and BMPs positively influence bone repair, the pathways appear to function in opposition to one another.

6Adult zebrafish fins comprise multiple organized cell types, including osteoblasts, fibroblasts, endothelial cells, neurons, and epidermal cells. Each of these cell types regenerate in concert after resection to perfectly restore the entire fin within 2–3 weeks. The bony rays, or lepidotrichia, that characterize fish fins are ossified spokes lined with bone forming osteoblasts. Lineage-tracing studies demonstrate that replacement osteoblasts are derived from preexisting uni-potent osteoblasts, although an unknown population can substitute if needed. These observations indicate that mature osteoblasts respond to injury by acquiring progenitor cell properties, a process known as dedifferentiation. These fate restricted cells then populate lateral regions of the regenerative blastema, specialized mesenchymal tissue formed upon amputation. Subsequently, by poorly understood mechanisms, the osteoblast lineage cells undergo coordinated proliferation, differentiation, and morphogenesis to form replacement bone.

7Fin regeneration is accompanied by upregulation of transcripts associated with osteogenesis including runx2a, runx2b, and sp7. Transgenic lines utilizing fish sp7 promoter/enhancer elements indicate that fin osteoblasts distal to the amputation site express sp7, whereas cells below the amputation site downregulate sp7 expression. Similarly, human Runx2 regulatory elements drive reporter gene expression in cells distal to the amputation site. Although the Wnt/β-catenin pathway is required for fin regeneration and components of the Wnt pathway are expressed in the regenerating fin, it remains unknown which cell types participate in Wnt/β-catenin signaling or how the pathway promotes regeneration. Similarly, BMP gene expression patterns and overexpression of the BMP inhibitor Chordin implicate the BMP pathway in fin osteoblast differentiation. However, it is unresolved how BMP promotes bone differentiation or even when and where BMP signaling is active during regeneration.

8Here, we examine mechanisms of regenerative osteogenesis in the zebrafish caudal fin. Upon fin amputation, Runx2+ preosteoblasts accumulate in the blastema following a Wnt/β-catenin dependent epithelial-to-mesenchymal transformation (EMT) of osteoblasts lining the damaged bone. Distal Wnt signaling continuously maintains Runx2+ preosteoblasts, whereas their differentiation into mature, sp7+ osteoblasts is promoted by autocrine BMP signaling. By inducing secreted Wnt antagonists, including dkk1b, BMP facilitates negative crosstalk between differentiating osteoblasts and preosteoblasts. This straightforward signaling network produces a hierarchically and spatially organized osteoblast lineage that balances a simultaneous need for bone growth and differentiation until fin regeneration is complete.

9We immunostained paraffin-sectioned fins with antibodies directed against the osteogenic transcription factors, Runx2 and sp7, at various times postamputation. At 24 hr postamputation (hpa), osteoblasts lining preexisting bone adjacent to the amputation site upregulated expression of Runx2. Further, Runx2, but not sp7, expressing cells were found distal to the amputation site distributed in the mesenchymal tissue of the forming blastema (Figures 1A–1C). At 32 hpa, sp7 expression was induced in select Runx2+ mesenchymal cells (Figures 1D–1F). By 48 hpa, osteoprogenitors became more linearly organized adjacent to the basal epidermis with Runx2−/sp7+ cells first becoming detected near the amputation site (Figures 1G–1I). At 72 hpa, the osteoblast lineage was highly organized along the proximal-distal axis (Figures 1J–1L). Runx2+ cells were always the most distally localized population and were neighbored by Runx2+/sp7+ cells. In contrast, sp7+ cells were concentrated near the amputation site. Similarly, runx2a and sp7 transcripts were enriched at the distal and proximal ends of the blastema, respectively (Figures 1M and 1N).

10We quantified osteoblast subtypes and their proliferation state by delivering 5-ethynyl deoxyuridine (EdU) to label cells in S phase 6 hr prior to harvesting fins at 72 hpa. In stained section fins, we observed robust EdU labeling in the distal blastema mesenchyme but only rarely in the epidermis (Figures 1O and 1P), in agreement with a previous report. Runx2+ and Runx2+/sp7+ cells were found in equal proportion (Figures 1O, 1P, and 1Q, ~45%) with significantly fewer sp7+ cells (Figure 1Q, ~8%, p < 10−5). More Runx2+ cells incorporated EdU compared to sp7+ cells (16% versus 7%) (Figure 1Q, p < 0.03). We also routinely saw clusters of several high-expressing Runx2+ cells that were EdU−, suggesting there are distinct proliferative and nonproliferative Runx2+ preosteoblasts (Figure 1P, red arrows). The temporal and spatial patterns of Runx2 and sp7 expression and osteoblast-subtype proliferation demonstrate a hierarchical organization of the osteoblast lineage in regenerating fins. Renewing Runx2+ preosteoblasts at the leading edge of the fin beget proliferative Runx2+/sp7+-differentiating cells. These cells subsequently mature into relatively quiescent sp7+ osteoblasts that append to progressively elongating bone.

11We performed lineage tracing to confirm the distal, leading edge Runx2+ cells were derived from preexisting osteoblasts. We generated adult zebrafish with caudal fins containing isolated fin rays lined with permanently labeled mCherry+ osteoblasts (Figure 2A). At 72 hpa, fins from the same animal displayed abundant mCherry+ osteoblasts in the regenerating tissue (Figure 2B). Immunostaining showed that Runx2+ cells extending from labeled hemirays coexpressed mCherry (Figures 2C–2H). Likewise, Runx2+ blastema cells observed at 24 or 48 hpa were also derived from mosaic-labeled osteoblasts (Figure S1). These results demonstrate that fin amputation triggers the migration of osteoblasts closely associated with the bone ray into the nascent blastema following their upregulation of Runx2. These Runx2+ blastema cells are the precursors to redifferentiated osteoblasts that will line and mineralize replacement bone.

12We hypothesized that fin amputation induces an epithelial-to-mesenchymal transformation (EMT) of osteoblasts, which display an underappreciated epithelial-like organization, to generate Runx2+ preosteoblasts. We immunostained nonregenerating fins with zns-5, α-catenin, and β-catenin antibodies to label osteoblasts (zns-5) and adherens junctions that interconnect epithelial sheets (catenins). Zns-5-positive osteoblasts expressed both α- (Figures 2I–2L) and β-catenin (Figure S2A), which were concentrated at membranes separating individual osteoblasts, demonstrating that fin ray osteoblasts have epithelial-like properties.

13We examined α-catenin localization in 96 hpa fins from transgenic Tg(sp7:EGFP)b1212 animals, which express EGFP in mature and differentiating osteoblasts. We saw membrane-localized α-catenin staining in both Runx2+/sp7:EGFP+ and sp7:EGFP+ cells that were in close proximity to new bone. In contrast, distally located Runx2+ cells did not express α-catenin, suggesting they lacked adherens junctions (Figures 2M–2R). N-cadherin, another component of adherens junctions, also was expressed in maturing osteoblasts (Figures S2B–S2E). At 24 hpa, we observed α-catenin staining only in the fin epidermis (Figures 2S and 2T), indicating that osteoblasts rapidly lose α-catenin expression as they populate the blastema and acquire a progenitor state. Further indicative of EMT, osteoblasts changed cell shape from strikingly long and thin epithelial cells lining nonre-generating bone to a more compact, polygonal morphology when they became Runx2+ and Runx2+/sp7+ blastema-populating cells (compare Figures S3E and S3F).

14EMTs are directed by transcription factors including Snail/Slug and Twist. In developing bone, Twist1 and Twist2 are expressed in Runx2+ preosteoblasts where Twist interacts with Runx2. RNA in situ hybridization showed that both twist2 and runx2a were induced at 24 hpa in tissue immediately adjacent to the amputation site (Figures 2U and 2V). Combining fluorescent in situ hybridization with Runx2 immunostaining on 72 hpa sections succinctly demonstrated that distal Runx2+ cells coexpressed twist2 mRNA (Figures 2W–2Y) and are thus bona fide mesenchymal cells. Similar to twist2, twist3 was expressed at 24 hpa adjacent to the amputated bone (Figure S2G) and by 72 hpa was distally localized in a pattern reminiscent of twist2 (Figures S2F and S2H). We conclude that Runx2+ cells originate from an EMT of epithelial-organized differentiated osteoblasts present near the amputation site, and, throughout regeneration, distal Runx2+ preosteoblasts are maintained in a mesenchymal twist2/3-expressing state. However, the initial induction of Runx2 expression does not appear to depend on EMT, because epithelial osteoblasts lining the bone proximal to the amputation site also upregulate Runx2+ (Figure 1).

15Wnts promote EMT, osteogenesis, and fin regeneration. We monitored Wnt signaling in regenerating fins by examining β-catenin localization by antibody staining. At 24 and 32 hpa, we saw Runx2+ cells with diffuse nuclear β-catenin localization dispersed in the mesenchyme distal to the amputation site (Figures 3A–3F). Nuclear β-catenin was still apparent at 48 hpa when Runx2+ cells became organized along the anterior-posterior body axis (Figures 3G–3I). At 72 hpa, strong nuclear β-catenin staining persisted in distally residing Runx2+ preosteoblasts, with less nuclear β-catenin in maturing osteoblasts located closer to the amputation site (Figures 3J–3L), suggesting that downregulation of Wnt signaling accompanies osteoblast differentiation. At 96 hpa, Tg(sp7:EGFP) animals showed robust nuclear β-catenin primarily in distal Runx2+ cells and not in proximal osteoblasts that expressed only sp7:EGFP (Figures 3M–3R and S3A–S3D). High-magnification single optical sections confirmed overlapping nuclear β-catenin and Runx2 in distal osteoblasts at 96 hpa (Figures 3S–3U). Expression of the Wnt transcriptional effector tcf7 was also enriched in Runx2+ osteoblasts (Figures 3V–SX, S3G, and S3H), supporting the potential of Runx2+ cells to respond to Wnt signals.

16We examined axin2 expression, a well-accepted target gene of canonical Wnt signaling, to monitor Wnt-responsive cell types in the regenerating fin. At 72 hpa, axin2 was strongly expressed in distal Runx2+ preosteoblasts (Figures 3Y–3AA, S3I, and S3J). Axin2 transcripts additionally were detected in Runx2− cells directly medial to the preosteoblasts and extending to the extreme tip of the blastema. These nonosteoblasts also showed low levels of nuclear β-catenin (Figures S3A–S3D). We further examined Wnt activity using Tg(TOP:GFP)w25 transgenic fish that express destabilized GFP under the control of a synthetic Wnt-responsive promoter. Due to the rapid turnover of the destabilized GFP, we monitored GFP expression by in situ hybridization. At 72 hpa, TOP:GFP was found in cells of the distal blastema, including Runx2+ preosteoblasts, in a pattern indistinguishable from axin2 (Figures 3BB–3DD). We did not observe nuclear β-catenin, axin2 or TOP:GFP expression in the epidermis, suggesting that Wnt/β-catenin is not a major regulator of fin epidermis regeneration even though the basal epidermis expresses the Wnt effector lef1. Collectively, the striking nuclear β-catenin, axin2 expression, and TOP:GFP activity in leading edge Runx2+ cells suggest ongoing roles for canonical Wnt signaling in regulating Runx2+ preosteoblasts from their earliest EMT-driven emergence following amputation through later stages of regenerative outgrowth. Our data further indicate that nonosteoblast lineage cells in the distal blastema represent an additional node of Wnt/β-catenin signaling.

17Wnt is one secreted factor that can initiate EMT during embryonic development, and Wnts induce twist expression during mouse craniofacial bone development. We performed loss-of-function studies of Wnt signaling during fin regeneration using IWP-2, a small molecule inhibitor of Porcupine, an acyltransferase that covalently modifies Wnts and is required for their secretion. We treated Tg(sp7:EGFP) animals with IWP-2 and monitored regeneration in individual animals from 0 to 8 dpa. At 2 and 4 dpa, IWP-2-treated fish arrested regeneration after forming a wound epidermis and lacked sp7:EGFP+ osteoblasts beyond the amputation site (Figure S4). By 8 dpa, control animals had substantially regenerated and sp7:EGFP+ osteoblasts were seen throughout the regenerate (Figures 4A–4D and S4). In contrast, IWP-2-treated fish displayed no regenerative outgrowth, although bone proximal to the amputation site continuously maintained sp7:EGFP+ expression (Figures 4E–4H and S4). Low doses (100 nM) of Wnt-C59, a chemically distinct Porcupine inhibitor, also completely inhibited bone regeneration (Figures S5A–S5D).

18Given the complete lack of bone regeneration in both IWP-2-and Wnt-C59-treated fish, we speculated that an initiating step(s) requires Wnt activity. Sections of fins from fish treated with IWP-2 from 0 to 24 hpa lacked both nuclear-localized β-catenin and Runx2-expressing blastema cells (Figures 4I–4P). However, Runx2 expression was still evident in bone-lining osteoblasts near the amputation site, suggesting its initial activation is Wnt-independent (Figures 4I–4P). A later treatment with IWP-2 from 48 to 72 hpa abolished twist2 expression (Figures 4Q and 4R). These results demonstrate that Wnt initiates osteoblast EMT and then maintains the mesenchymal state of preosteoblasts by promoting twist2 expression.

19We tested if Wnt has an ongoing role maintaining preosteoblasts during bone regrowth by blocking Wnt production from 48 to 72 hpa using IWP-2. This IWP-2 regimen stunted overall regeneration and depleted osteoblast-lineage cells distal to the amputation site (Figures 5A–5D). The few remaining cells displayed membrane-localized β-catenin and reduced levels of Runx2 (Figures 5A and 5B), whereas sp7 expression was largely unchanged (Figures 5C and 5D). Along with fewer osteoblasts, fins from IWP-2-treated fish had a pronounced decrease in collagen deposition near amputated bones (Figures S5M–S5P). Although the small number of osteoblasts remaining after a 24 hr IWP-2 treatment precluded a detailed analysis, we were able to rule out apoptosis as a cause of the phenotype by TUNEL staining (Figures S5Q–S5V).

20At least two populations of distally located cells transmit Wnt signals in the regenerating fin: Runx2+ preosteoblasts and adjacent nonosteoblast distal mesenchymal cells. Therefore, the loss of osteoblasts after 24 hr IWP-2 treatment could represent direct and/or indirect Wnt signaling roles. To distinguish between these possibilities, we inhibited Wnt for brief periods (8 hr) during regeneration and monitored β-catenin, Runx2, and sp7 by immunostaining. Administration of IWP-2 or Wnt-C59 from 64 to 72 hpa was sufficient to inhibit nuclear localization of β-catenin (Figures 5E and 5F and S5E–S5L). An examination of osteoblast subtypes (Figures 5G–5J) indicated both IWP-2- and Wnt-C59-treated fish contained fewer Runx2+ cells (Figure 5I, p < 0.02) and increased numbers of Runx2+/sp7+ cells (Figure 5I, p < 0.02). However, no significant difference in the number of sp7+ cells was observed (Figure 5I). EdU incorporation in Runx2+ cells, but not sp7-expressing osteoblasts, was also reduced in fish treated with either IWP-2 or Wnt-C59 (Figure 5J, p < 0.003), indicating that Wnt-dependent proliferation specifically maintains Runx2+ preosteoblasts. We also quantified Runx2 and sp7 expression on sections from multiple animals and produced scatterplots depicting normalized expression levels in individual cells. Confirming our qualitative observations, there was a 3.1-fold decrease in the fraction of Runx2 single positive cells relative to all mesenchymal cells in animals treated with IWP-2 from 64 to 72 hpa (Figures S5W and S5X).

21We developed a method to isolate and culture primary fin osteoblasts where 78.5% of the cells stained positively for Runx2 and/or sp7 (SD = 8.3%, n = 5 independent cell preparations). To test if Wnt directly promotes fin osteoblast proliferation, we treated these cultures with Wnt3A and monitored EdU incorporation in Runx2+ cells. Wnt3A significantly increased the fraction of EdU+/Runx2+ cells (Figures S5Y–S5DD, p = 0.0021). We conclude that, in regenerating fins, continuously secreted Wnt activates canonical Wnt signaling in Runx2+ preosteoblasts to support, directly or indirectly, their renewal.

22BMPs have been implicated in osteoblast differentiation and are expressed in fin osteoblasts during regeneration. We examined the pattern of BMP activity by staining fin sections from 96 hpa Tg(sp7:EGFP) fish with antibodies against phosphorylated, active Smad1, 5, and 8 (pSmad1/5/8), Runx2, and EGFP. We observed pSmad1/5/8 in differentiating sp7:EGFP+ cells but not Runx2+/sp7:EGFP− preosteoblasts or other cell types (Figure 6A). pSmad1/5/8 staining was evident by 48 hpa, coincident with the appearance of sp7+ maturing osteoblasts (Figures S7A–S7D). The discrete localization of pSmad1/5/8 suggests that differentiating osteoblasts are the primary targets of BMP signals during fin regeneration and that the pathway promotes osteoblast redifferentiation.

23We blocked BMP signaling in Tg(sp7:EGFP) animals using the BMPR inhibitor (BMPRi) LDN193189 from 0 to 8 dpa (Figures 6B–6I and S6). At 2 dpa, we could not distinguish BMPRi-treated from control fish, but by 4 dpa BMP inhibition produced a pronounced decrease in the extent and levels of sp7:EGFP expression (Figure S6). Paradoxically, by 8 dpa of BMPRi treatment, sp7:EGFP expression became relatively high, possibly indicating the activation of feedback mechanisms (Figure S6). Regardless, long-term inhibition of the BMP pathway (0–8 dpa) reduced bone formation (Figures 6B–6I) with BMPRi-treated fish lacking mature lepidotrichia including the joints between ray segments but maintaining regeneration of epidermis and blood vessels at 4 dpa (Figures S7E–S7H). Von Kossa staining revealed that fins from BMPRi-treated fish still formed a blastema but failed to produce mineralized bone (Figures S7I–S7L), consistent with a reduced expression domain of col10a1, a marker of differentiating osteoblasts (Figures S7M and S7N). Importantly, pSmad1/5/8 levels were greatly diminished by chronic BMPRi treatment (Figures S7O and S7P), demonstrating drug efficacy. The bone-specific phenotype induced by BMPRi exposure contrasts those observed with IWP-2 or Wnt-C59, which largely blocked blastema outgrowth, and is consistent with the observation that BMP activity is primarily located in differentiating osteoblasts.

24To determine BMP roles during ongoing regeneration, we treated animals with BMPRi from 48 to 72 hpa, which blocked Smad1/5/8 phosphorylation, increased Runx2+ cells near the amputation site (Figures 6J–6O), and decreased sp7 expression (Figures 6P and 6R). We scored osteoblast subtypes in DMSO versus BMPRi-treated animals (Figure 6T), confirming BMPRi treatment increased Runx2+ cells (p < 0.002) and decreased the Runx2+/sp7+ and the sp7+ populations (p < 0.007 for the Runx2+/sp7+ cells and p < 0.02 for the sp7+ population). EdU incorporation (Figures 6Q and 6S, quantified in Figure S7Q) and TUNEL staining (Figures S5Q–S5V) revealed osteoblast proliferation and cell death, respectively, were unaffected by BMPRi treatment. These results argue against a role for BMP in maintaining Runx2 expression and rather demonstrate that BMP specifically drives osteoblast differentiation by promoting the transition from Runx2+/sp7− cells to sp7+ cells.

25Because inhibiting BMP activity expanded the Runx2+ population, we hypothesized that BMP signaling in proximally located osteoblasts quenches Wnt/β-catenin activity. Indeed, stained fin sections from 48 to 72 hpa BMPRi-treated fish exhibited a dramatic increase of nuclear-localized β-catenin throughout an expanded population of Runx2+ preosteoblasts (Figures 7A–7E and 7G–7K). The domain of axin2 expression also was expanded proximally in BMPRi-exposed fish, further indicating that BMP functions in proximal osteoblasts to restrain Wnt activity to the distal-most progenitor pool (Figures 7F and 7L).

26Cultured primary caudal fin osteoblasts became uniformly pSmad1/5/8+ without the addition of exogenous BMP, suggesting cell autonomous BMP production and signaling (Figures S7R–S7U). Among BMPs reported to be expressed in the regenerating fin, we detected robust expression of bmp2b in cultured osteoblasts by quantitative RT-PCR (qRT-PCR) (Figure S7V). We reconstituted BMP’s negative regulation of Wnt/β-catenin by combined addition of BMPRi and recombinant Wnt3A to primary fin osteoblasts. Wnt3A induced partial nuclear accumulation of β-catenin (Figures 7M–7S), which was substantially enhanced by inhibiting BMP signaling using BMPRi (Figures 7P–7S, p < 0.01). These results are congruent with our in vivo observations that BMP signaling is active in osteoblasts where its roles include negatively regulating Runx2+ preosteoblasts by countering Wnt signals.

27Dkk proteins compete with Wnt ligands for binding to Fz/Lrp5/6 receptor complexes. We therefore hypothesized that BMPs could negatively regulate Wnt/β-catenin signaling by activating Dkk expression. Primary fin osteoblasts expressed dkk1a, dkk1b, dkk2, and dkk3b, among which dkk1a, dkk1b, and dkk3b expression were reduced upon BMPRi treatment, as was sp7 (Figure 8A, p < 0.05). We next examined BMP-dependent expression of dkk genes in regenerating fins from fish treated with DMSO or BMPRi from 48 to 96 hpa. Dkk1b, dkk2, and dkk3b all were significantly downregulated in BMPRi-treated animals (Figure 8B, p < 0.05). Consistent with antibody staining and the in vitro results, sp7 transcript levels were also reduced upon BMPRi treatment. Dkk2 was downregulated by BMPRi in vivo but not in cultured cells, possibly reflecting nonosteoblast expression of dkk2 in intact fins. Unique among Dkk proteins, Dkk3 is not thought to directly antagonize Wnt/receptor interactions. Nevertheless, dkk3b was enriched in regenerating osteoblasts, and its expression was reduced by a 24 hr BMPRi treatment from 48 to 72 hpa (Figures S8A and S8B).

28The BMP-dependent expression of dkk1b and its ability to inhibit Wnt signaling in zebrafish and attenuate fin regeneration suggested that Dkk1b may be sufficient to inhibit the Wnt-dependent Runx2+ preosteoblast pool. We amputated fins from control and Tg(hsp70l:dkk1b-GFP)w32 fins, subjected them to a heat shock regimen from 48 to 72 hpa, and examined β-catenin localization and Runx2 expression. Dkk1b induction decreased nuclear-localized β-catenin and produced a pronounced deficiency in Runx2+ preosteoblasts (Figures 8C–8H), consistent with Dkk1b being a potent BMP-dependent inhibitor of osteoblast progenitor renewal.

29The division of regenerating fin osteoblasts into distal Wnt active and proximal BMP active populations suggests the pathway’s respective ligands would be similarly distributed. Guided by earlier studies, we determined the cell specific expression patterns of bmp2b, wnt5a, wnt5b, and wnt10a transcripts on 72 hpa fin sections. In situ hybridization revealed that bmp2b mRNA was produced by differentiating proximal osteoblasts (Figure S8C). In contrast, expression of wnt5a (Figure S8D) and wnt5b (Figure S8E) was concentrated toward the distal end of the regenerating fin, in both blastema mesenchymal cells and basal epidermis. Although more weakly expressed, wnt10a was also present in distal and lateral regions of the blastema in or adjacent to osteoblasts (Figure S8F). To conclusively identify the cellular sources of wnt5a and bmp2b relative to osteoblast lineage cells, we performed double immunostaining/in situ hybridization studies. Wnt5a expression was expressed in mesenchymal cells occupying the extreme distal blastema bordered laterally by Runx2+ preosteoblasts. Minimal wnt5a was detected in Runx2+ preosteoblasts or blastema cells adjacent to maturing sp7-expressing osteoblasts (Figures 8I–8L). Bmp2b was expressed in maturing sp7+ osteoblasts, but not Runx2+ distal preosteoblasts (Figures 8M–8P). Based on their spatial separation, Wnt5a/5b and Bmp2b are attractive candidate ligands to simultaneously promote the opposing activities of Wnt-dependent preosteoblast maintenance and BMP-dependent osteoblast differentiation, respectively.

30Bone regeneration in zebrafish fins requires regulatory mechanisms to (1) generate dedifferentiated progenitors from mature osteoblasts, (2) maintain this preosteoblast population until regeneration is complete, and (3) spatially and temporally restrict redifferentiation to a subset of preosteoblasts that progressively reform lost bone. The latter two processes are opposing activities; a fine balance between progenitor expansion and terminal differentiation must be maintained to sustain bone regeneration over 2 or more weeks. Our studies suggest a model for bone regeneration in the fin (Figure 8Q) whereby a Wnt-dependent EMT of epithelial osteoblasts populates the newly formed blastema with dedifferentiated Runx2+ preosteoblasts. Spatial segregation of two opposing pathways, Wnt/β-catenin and BMP, then balances growth and differentiation until regeneration is complete. Sustained high levels of Wnt in the distal blastema directly and continuously maintain a leading edge pool of Runx2+ mesenchymal preosteoblasts. Conversely, as more proximally located Runx2+ cells become physically distanced from the distal Wnt source, they upregulate bmp2b and activate autocrine BMP signaling. BMP promotes osteoblast differentiation by inducing sp7 and dkk1b expression to feedback inhibit Wnt activity and prevent unproductive overexpansion of the progenitor pool. In agreement, dkk1b promoter elements drive gene expression exclusively in differentiating osteoblasts near the amputation site. As osteoblast maturation ensues, sp7-expressing cells downregulate Runx2, re-epithelialize by reforming adherens junctions, and progressively extend new mineralized bone (Figure 8Q).

31Our examination of canonical Wnt activity shows that osteoblasts are a major Wnt-responsive cell type in the regenerating fin. Rapid changes in preosteoblast subtypes and their proliferation with short-term Wnt inhibitor exposure and the sufficiency of recombinant Wnt to promote osteoblast proliferation demonstrate canonical Wnt directly affects preosteoblasts. The requirement for Wnt in Runx2+ preosteoblast proliferation during fin regeneration is consistent with known functions for the Wnt/β-catenin pathway in maintaining progenitor cells in diverse biological settings and regulating bone development in mice. Precisely how Wnt/β-catenin signaling supports the preosteoblast state requires further study, but it is likely that Wnt/β-catenin’s effects on transcription favor the silencing of genes required for differentiation and activation of genes associated with “stemness”.

32We also observed axin2 and TOP:GFP expression, as well as modest nuclear β-catenin in nonosteoblast distal blastema cells, indicating they, like preosteoblasts, are Wnt responsive. A role for Wnt signaling outside of osteoblasts is supported by our long-term Wnt inhibitor and Dkk1b overexpression experiments that produce a near complete arrest of fin regeneration. Therefore, akin to its role in preosteoblasts, canonical Wnt signaling may maintain these distal-most blastema cells, which then act as a signaling niche/center that orchestrates the overall fin regeneration process. This concept is described in detail in the accompanying paper from in this issue of Cell Reports. Wnt5a, wnt5b, and wnt10a are expressed in distal regenerating fin tissue, results we confirmed and expanded upon by demonstrating that wnt5a is expressed in distal blastema cells adjacent to Runx2+ preosteoblasts. Therefore, distal blastema cells may be a self-sustaining source of a Wnt morphogen gradient that both directly and indirectly regulates osteoblasts and other cell types of the regenerating fin.

33Our observations show that BMP signaling likely is dedicated to differentiating osteoblasts in regenerating fins. Antibody staining localizes pSmad1/5/8 to proximal maturing osteoblasts and not distal Runx2+ preosteoblasts. Further, BMP receptor function is required for normal sp7 expression and bone formation, but not to maintain distal Runx2+ cells. Intriguingly, although distal Runx2+ cells with robust nuclear β-catenin reside near a field of bmp expression, they and other distal cells are pSmad1/5/8 negative and therefore likely refractory to BMP signals. One possible explanation is feedback inhibition of BMP in the distal blastema by Wnt-induced BMP antagonists. Ectopic expression of Chordin, a secreted BMP inhibitor, blocks fin regeneration, although it is unknown whether endogenous Chordin, or other BMP inhibitors, quench BMP signals in the context of fin regeneration. We find that cultured fin osteoblasts strongly express bmp2b and accordingly become pSmad1/5/8+. Further, bmp2b expression is concentrated in differentiating sp7+ osteoblasts in vivo, as suggested by an earlier study. Direct downstream targets of BMP/Smad during bone regeneration could include both sp7 and dkk genes, such as dkk1b. Our model does not resolve if other signaling pathways required for fin ray regeneration, including fibroblast growth factor (FGF) and Hedgehog (reviewed in ), interface with Wnt/β-catenin and BMPs in osteoblasts or have secondary effects on regenerating bone.

34Self-renewing tissues, including intestinal epithelium, stomach lining, interfollicular epidermis, sebaceous glands, and hair follicles of the skin, contain stem cells that produce cells whose differentiation states are both temporally and spatially arranged. The logic of how these hierarchical progenitor cell systems are maintained is remarkably similar to what we describe during fin bone regeneration with Wnt/BMP mutual antagonism being a common theme. Mathematical models predict an obligatory role for negative feedback loops to balance stem cell renewal and differentiation. Therefore, an equilibrium between Wnt and BMP output may be a common logic component of regulatory networks that establish a precise balance between cell plasticity and differentiation.

35Our interpretation of bone regenration in the zebrafish fin suggests that the signaling networks that control regenerative osteogenesis are evolutionarily conserved, relatively simple, and largely lineage intrinsic. Human bone may retain the competency to robustly regenerate if inherent networks could be somehow activated and/or augmented. Currently, recombinant BMPs are the only biologic factors approved for use in bone repair therapies; however, whereas BMPs can promote ossification, their clinical efficacy is controversial. A potential explanation for this can be inferred from our results and those of others that show that BMPs downregulate Wnt/β-catenin signaling in bone. New therapeutic approaches can be envisioned that exploit the features of regenerative osteogenesis that we have uncovered. For instance, first enhancing Wnt/β-catenin activity using small molecules or Wnt proteins soon after injury may expand endogenous preosteoblasts, an idea supported by enhanced healing of damaged bone in mice treated with recombinant Wnts. Then, later stages of bone healing may benefit from BMP treatment to enhance osteoblast redifferentiation and mineralization. More broadly, our observations encourage approaches aimed at coaxing human cells to mimic those of animals, like zebrafish, that have remarkably robust regenerative abilities.

36Wild-type AB, Tg(sp7:EGFP)b1212, Tg(hsp70l:dkk1b-GFP)w32, Tg(dusp6:Cre-ERT2,myl7:ECFP) b1230, Tg(TOP:GFP)w25, and Tg(Xla.Eef1a1-actb2:LOXP-LOX5171-FRT-F3-EGFP,mCherry) vu295a lines were maintained according to University of Oregon institutional guidelines. Regeneration studies were performed at 28°C –29°C as described previously.

37Sections were rehydrated and antigen retrieval was performed for 10 min in a pressure cooker in retrieval buffer (1 mM EDTA [pH 8], 0.1% Tween 20). Antibodies were diluted in PBST containing 10% nonfat dry milk and applied to slides overnight at 4°C, followed by staining with Alexa-conjugated secondary antibodies (Invitrogen) and Hoechst nuclear staining. Antibody staining details are provided in the Supplemental Experimental Procedures.

38Wnt pathway inhibitor IWP-2 and the BMPR inhibitor, LDN193189, referred to in the text and figures as BMPRi, were purchased from StemRD; Wnt-C59 was from Biovision; all were dissolved in DMSO. At the indicated times after fin amputation, IWP-2, Wnt-C59, BMPRi, or DMSO was added to fish water (one animal per 200 ml of water) at 10 μM, 100 nM, or 5 μM, respectively. For each drug treatment (n = 3), cohorts of three or four animals were used, each was analyzed at the completion of the study, and images shown are representative examples of each cohort.

39For combination fluorescent in situ hybridization immunostaining, sections on slides were sequentially antigen retrieval treated, hybridized with DIG-labeled probes, stained with α-DIG peroxidase-conjugated antibody (Roche), developed using the TSA system (PerkinElmer), and finally immunostained. Protocol details and probe synthesis are described in the Supplemental Experimental Procedures.

40These methods are presented in the Supplemental Experimental Procedures.

41Statistically significant differences between osteoblast subtypes were determined by scoring their fractional representation on comparable immuno-stained sections (for Figures 1 and 5, >600 osteoblasts from more than six rays compiled from at least three fish; for Figure 6, >400 osteoblasts from more than four rays collected from at least three animals). One-tailed Student’s t tests compared the means of each population’s percentage of total osteoblasts across individual rays. Fisher’s exact tests were used to determine significant differences in the proportion of EdU+ cells between osteoblast subtypes, combining osteoblasts scored in 12 rays from at least four fish (>1,100 cells total). To assess changes in EdU incorporation between drug-treated fish, comparable stained sections were scored for the fraction of EdU incorporating cells of each osteoblast subtype. Student’s t tests assessed differences between drug-treated groups using the means of each subtype’s percentage of EdU+ incorporating cells across individual rays. For nuclear β-catenin quantitation in cultured osteoblasts, one-way ANOVA and post hoc Tukey’s tests were used to determine significant differences in the ratio of nuclear to total β-catenin in individual cells across and between treatments. To determine differentially expressed gene in qRT-PCR studies, two-tailed Student’s t tests used raw ΔCts from three or four independent control and experimental samples.