Compartmentalization and synergy of osteoblasts drive bone formation in the regenerating fin

Summary Zebrafish regenerate their fins which involves a component of cell plasticity. It is currently unclear how regenerate cells divide labor to allow for appropriate growth and patterning. Here, we studied lineage relationships of fluorescence-activated cell sorting-enriched epidermal, bone-forming (osteoblast), and (non-osteoblast) blastemal fin regenerate cells by single-cell RNA sequencing, lineage tracing, targeted osteoblast ablation, and electron microscopy. Most osteoblasts in the outgrowing regenerate derive from osterix+ osteoblasts, while mmp9+ cells reside at segment joints. Distal blastema cells contribute to distal osteoblast progenitors, suggesting compartmentalization of the regenerating appendage. Ablation of osterix+ osteoblasts impairs segment joint and bone matrix formation and decreases regenerate length which is partially compensated for by distal regenerate cells. Our study characterizes expression patterns and lineage relationships of rare fin regenerate cell populations, indicates inherent detection and compensation of impaired regeneration, suggests variable dependence on growth factor signaling, and demonstrates zonation of the elongating fin regenerate.


INTRODUCTION
Zebrafish rapidly regenerate complex tissues after loss, including their appendages, the fins.Due to fast bone restoration and transparency, the fin serves as a valuable tool to study bone regeneration. 1After amputation, a multi-layered wound epidermis (WE) forms, which is followed by blastema formation within 2 days post amputation (dpa) and subsequent fin outgrowth. 2The blastema, a mass of proliferative cells accumulating at the amputation plane subdivides into different zones: a distal most blastema (DMB) of several cell diameters in size and more proximal and lateral regions (proximal blastema), in which proliferation, patterning, and osteoblast differentiation (lateral blastema domains in longitudinal section view) take place. 3he descendance of osteoblasts in the regenerate from stump cell populations has been intensely studied.Mature osterix+/osteocalcin+ osteoblasts in the fin stump dedifferentiate, proliferate, and migrate toward the forming blastema, where they contribute to restoration of bone matrices, [4][5][6] although osterix+ cells in the fin stump are dispensable for regeneration. 7Other, more progenitor cell-like stump cell populations add to the osteoblast cell pool in teleost fin regenerates as well; among them are mmp9+ and col10a1+ cells at the segment joints. 8,9hether and to what extent committed osteoblasts, osteoblast progenitors, and non-osteoblast cells assembled within the early regenerate contribute to ongoing bone formation is unclear.
In the regenerate, osteoblasts of different maturity levels reside in different locations.Osteoblast progenitors expressing Runx2 localize to a region close to the DMB while differentiating osterix+ osteoblasts reside in more proximal positions.Fully mature osteoblasts expressing osteocalcin are found in proximity to the amputation plane at later stages of regeneration. 4,10Thus, based on relative position within the regenerate and marker expression, a minimum of 3 osteoblast subtypes can be distinguished in the fin regenerate, although little is known about their respective expression profiles.The aforementioned Runx2/osterix hierarchy of osteoblasts in the distal regenerate appears to be supported by a pool of distal Runx2+ cells whose existence is maintained by Wnt/b-catenin signaling and opposed by BMP (bone morphogenetic protein) signaling. 11While this awaits further investigation, delineation of lineage relationships between different osteoblast subtypes within the regenerate and their potential descendance from non-osteoblast sources is crucial to understand the basis of successful regeneration (with bone providing the necessary structural support to the regenerate) and to acknowledge the extent of cellular plasticity that is required for it.

ll OPEN ACCESS
In this study, we performed single-cell (sc) transcriptomics on fluorescence-activated cell sorted epidermal, "blastema" (blastema cells excluding osteoblasts), and osteoblast cell populations which may have been underrepresented in previous analyses, 12,13 due to the high abundance of outer epidermal and proximal mesenchymal populations in fin regenerates.We identified novel subpopulations and respective marker genes and generated and tested hypotheses regarding lineage relationships of regenerate cell populations by performing trajectory inference and Cre-loxP genetic fate mapping.Tracing osterix+ osteoblasts and mmp9+ progenitor cells within the regenerate showed that both populations contribute to bone regeneration to different degrees, and that their ablation affects proliferation, patterning, and matrix formation in different parts of the regenerate.Furthermore, trajectory analysis suggested that shha+ cells of the basal layer of the WE (BLWE) do not contribute to the osteoblast population, while Wnt-responsive 7xTCFsiam+ cells (here referred to as siam+ cells) in the distal blastema do so, as confirmed by label-retaining cell analysis.This progenitor cell population compensates for impaired regenerate elongation after impaired regeneration due to suppression of Fgf signaling or osterix+ cell ablation, a recovery process that is itself independent of Fgf signaling but involves enhanced Wnt signaling.These findings, together with the identification of novel fin regenerate markers, advance our understanding of complex tissue regeneration in zebrafish.

A robust regenerative response to repeated amputation
In order to enrich for DMB cells, osteoblasts, and cells of the lateral BLWE of the growing 3 dpa fin regenerate, we FACS (fluorescence-activated cell sorting)-isolated fluorescently labeled cells of quadruple transgenic reporter zebrafish, in which siam+, Runx2+, osterix+, and shha+ cells were labeled by either GFP or mCherry protein expression (Figures 1A, 1B, and S1A), and performed sc RNA sequencing.We repeated the procedure 4 times (Reg1-Reg4) using the same zebrafish in intervals of 4 weeks (Figures 1B and S1B).Single-cell transcriptomic analysis across samples 14 revealed proliferating cells in distinct populations of the fin regenerate.We inferred the presence and UMAP (uniform manifold approximation and projection) location of the DMB by the near-absence of proliferating cells 3 and confirmed the presence of proliferating cells in other populations (pcna, mki67, Figures 1B and S1C).Cluster composition revealed similar contributions of each amputation experiment (Figure 1B) and showed that gene expression between the first (Reg1) vs. the following (Reg2-4) samples was overall similar (Figure S1D).Significantly downregulated transcripts were only detected for 5 genes with a fold change (FC) of less than 0.5, while no significantly upregulated genes with an FC beyond 2 were detected (Figure S1E; Table S1).The overwhelming number of genes whose expression was similar between different samples (Figure S2) suggested that successive amputations do not influence gene expression levels in cells of the regenerate.

Separation of epidermal cell clusters from osteoblast and blastema cell clusters and their relative positions in the regenerate
In order to characterize the transcriptome of cell populations of interest, we used the combined dataset of Reg1-4 with 6,668 cells for cluster analysis and identified 3 main clusters (Figure 1C) encompassing 2, 4, and 5 subclusters (Figure 1D).With the help of this dataset, we then identified marker genes for the respective subclusters and used these to localize them via RNA in situ hybridization (ISH).On a UMAP representation of the data, the main BLWE cluster (Basal) was clearly separated from the other two main clusters, which were connected and represented the osteoblast (Osteo) and blastema (here referring to DMB and proximal blastema cells but excluding osteoblasts which are found in lateral parts of the blastema and excluding mesenchymal cells close to the amputation plane) clusters (Figure 1D).Trajectory inference using partition-based graph abstraction (PAGA) confirmed the Osteo-Blastema cluster connection (Figure 1E), 15 although RNA velocities visualized on the UMAP did not show a clear pattern (Figure S3A). 16We assessed the biological identity of the main clusters by inspecting marker genes (Figures 1F and S3B-S3D; Table S2), which included epcam, cldni, phlda2, fn1b, krt5, and lef1 in the BLWE, 12,17,18 and1/2, lepb, her6, and wnt5b in the blastema, [19][20][21][22] and twist2, crip2, and cdh11 in osteoblasts 11,12 (Figure 1F; Table S2).Gene set enrichment analysis uncovered overrepresentation of the gene ontology (GO) terms peptide metabolic process, translation, and peptide biosynthetic process in blastema cells (Table S3); ossification, skeletal system development, and extracellular matrix (ECM) organization in osteoblasts (Table S4); and cell adhesion, tight junction, and cytoskeleton in cells of the BLWE (Table S5).
Blastema2 cells encompass DMB cells.Underneath, proliferative peripheral (touching the BLWE) Blastema1 cells enriched for LOX (1 of many) (Figure 2C) and more central (devoid of contact with BLWE) postna+ Blastema0 cells (Figure 2B) can be found.mfap5+ Blastema3 cells (Figure 2E) reside in a position directly adjacent (distal) to Osteo1 osteoblasts (Figure 2J).Both populations share expression of abi3bpb and mxra8b (Figure S4D) and might thus represent a transition zone of blastema cells and osteoblasts.The diversity of osteoblast, blastemal, and BLWE clusters in our dataset illustrates the advantage of FACS enriching for rare cell populations of interest.

Lineage tracing of osterix+ osteoblasts and mmp9+ cells suggests variable contribution to bone formation and the presence of different regenerate domains
Next, we made use of CreERT2-loxP-mediated lineage tracing (Figure 3A) to determine the contribution of differentiated osterix+ osteoblasts (including Osteo4 osteoblasts, Figures 3B-3D) 4 and mmp9+ cells (including Osteo2 and likely Osteo1 osteoblasts, Figures 3E-3H, but also epidermal cells 8 and some non-osteoblast blastema cells; Figure 3F) as the fate and extent of progeny formation of these osteoblast populations within the elongating regenerate were unexplored.We activated CreERT2 in osterix:CreERT2-p2a-mCherry x hsp70L:R2nlsGFP, 35 mmp9:CreERT2 x hsp70L:R2nlsGFP, and osterix:CreERT2-p2a-mCherry x mmp9:CreERT2 x hsp70L:R2nlsGFP x Actb:dsRed2GFP 36 zebrafish fin regenerates by injection of 4-hydroxytamoxifen (4-OHT) at 2.5 dpa, when CreERT2 is expressed in osteoblasts/osteoblast progenitors of the regenerate 4,8 (Figure 3A).The hsp70L:R2nlsGFP responder line drives expression of nuclear GFP (nGFP) in cells after CreERT2-induced deletion of a floxed dsRed2 cassette (R2: loxP dsRed2 loxP) reliably after heat shock (Figure S6A) 4 ; the Actb:dsRed2GFP responder line drives expression of GFP. 36osterix+ recombined, nGFP+ cells in osterix:CreERT2-p2a-mCherry x hsp70L:R2nlsGFP zebrafish were dispersed in the 3 dpa regenerate, their location spanning the region of Osteo4 cells (Figure 3B) and more proximal osteoblasts, in which CreERT2 and mCherry proteins are produced in the transgenic Cre-driver zebrafish (Figure S6B).At 4 and 5 dpa the regenerate had considerably and progressively enlarged, as did the nGFP+ domain (62.45%G 5% of the regenerate at 3 dpa, 79.46% G 7.8% at 4 dpa, and 84.18% G 3.5% at 5 dpa, Figure 3C).At this time, the distalmost nGFP-labeled cells in the regenerate were detected $250 mm apart from the regenerate tip (compared to $350 mm at 3 dpa), with the density of nGFP-labeled regenerate cells remaining constantly high (Figure 3D).This suggests profound contribution of osterix+ osteoblasts and their progeny to bone-forming cells in proximal regions of the regenerate, sparing the most distal 250 mm of the regenerate.In mmp9:CreERT2 x hsp70L:R2nlsGFP zebrafish, the progeny of recombined mmp9+ cells labeled by nGFP was found close to the joints but also represented scattered cells within the fin rays and sometimes interrays 8 (Figure 3E).While recombined cells at the joints neither drastically changed their position (arrowheads in Figure 3E) nor changed their number (average clone size 24 G 10 cells, Figure 3G) until 5 dpa, distal scattered cell clones amplified in size (Figure 3E), varying considerably in number (average clone size 54 G 42 cells, Figure 3H).This indicates a limited contribution of mmp9+ cell progeny at the prospective joints to the osteoblast regenerate populations, while distal mmp9+ cells may contribute to considerably more (pre-) osteoblasts, but also non-osteoblast tissue in the regenerate, likely derived from mmp9+ cells in the epidermis 8 as well as other mmp9+ cell populations.In osterix:CreERT2-p2a-mCherry x mmp9:CreERT2 x hsp70L:R2nlsGFP x Actb:dsRed2GFP zebrafish, in which both osterix+ and mmp9+ cells had been recombined, more (distal) cells were labeled by GFP than in the individual Cre-driver line experiments.This was evident by a larger domain of GFP+ cells in individual fin rays of zebrafish in which both Credrivers were present (Figure 3I) and a concomitant shorter distance of labeled cells to the regenerate tip (brackets in Figure 3J).These data indicate that osterix+ cells including Osteo4 cells and mmp9+ cells including Osteo2 and Osteo1 cells contribute to different sets of fin regenerate osteoblasts in different compartments and at different quantities.

Mixing of blastema and osteoblast cell populations in the distal regenerate suggests contribution of distal blastema cells to bone formation
The fact that the progeny of osterix+ lineage-traced cells spared a $250 mm region at the regenerate tip indicated that other cells might contribute to distal osteoblast populations.In order to explore this possibility, we performed transmission electron microscopy (TEM) of 3 dpa regenerating fins and compared cell phenotypes of WE-underlying blastema and osteoblast cells in different proximodistal positions of the regenerate (Figures 4A, 4B, and S7).While DMB cells were small with few and flat endoplasmatic reticulum (ER) cisternae and few mitochondria and produced a thin ECM layer, putative pre-osteoblasts and osteoblasts produced more ECM and showed a more secretory phenotype with more dilated ER cisternae and more Golgi complexes and mitochondria, especially close to the amputation plane (Figures 4B and S7).We did not discern any abrupt phenotype changes between cells but rather a gradual change in morphology and tested for cells transitioning  between the distal and proximal blastema.A prime candidate for a transitioning population was the runx2:GFP+ cell population (Figure 1A), 10,11 in addition to distal mmp9+ cells in the blastema.Of note, runx2a transcripts were detected particularly in proliferating osteoblasts (Osteo0, Osteo3) (Figure 4C), while its expression was low in Osteo4 cells.We tested whether distal runx2:GFP+ pre-osteoblasts are generated by blastema cells using transgenic siam:mCherry x runx2:GFP zebrafish, in which DMB cells and pre-osteoblasts are labeled by mCherry and gfp transcripts and proteins, respectively (Figures 4D-4F).Live imaging, FACS, and combined RNA ISH and immunohistochemistry (ISH-IHC) were used to follow the respective cell progeny.Double RNA ISH showed that gfp and mCherry reporter transcript domains were sometimes clearly separated but often continuous with a slight overlap (arrowhead in Figure 4E), as previously described. 27ISH-IHC revealed that siam:m-Cherry protein distribution extended much more proximally than mCherry transcripts, and that there was only a minimum overlap between mCherry mRNA and GFP protein (arrowheads in Figure S8A).We hypothesized that some of these siam:mCherry protein+ cells had lost siam promoter activity and accordingly mCherry transcription, exited the distal Wnt-active domain, and relocated proximally to contribute to the (pre-) osteoblast cell pool.In order to test this, we performed FACS analysis in runx2:GFP x siam:mCherry fin regenerates and detected a considerable number of cells with simultaneous red and green fluorescence, i.e., mCherry and GFP protein overlap (Figure 4D).Confocal imaging of 3 dpa regenerates of the same zebrafish line revealed that GFP and mCherry double+ cells were detectable beyond 150 mm proximal to the distalmost DMB mCherry+ cells (protein level, Figures 4F and 4G).We never detected mCherry transcripts in such proximal locations and concluded that siam:mCherry+ blastema cells (potentially reflecting and1+, and2+ actinotrichia-forming cells, Figure S8B) 21 contribute to the pool of pre-osteoblasts in the regenerate.We also detected some cells with low GFP protein levels outside of the domain of gfp transcription in the tip region (0-150 mm, Figure S8C), suggesting that runx2:GFP+ pre-osteoblasts contribute to or mix with the distal blastema cell population.We next tested whether there was any overlap between siam:mCherry transcripts and gfp transcripts in osterix:GFP x siam:m-Cherry zebrafish using the same approach, with osterix:GFP+ cells being more proximally located than runx2:GFP+ cells.gfp and mCherry transcript domains were clearly separated (Figure 4H).Nevertheless, GFP and mCherry protein double+ cells were detected proximal to siam:mCherry transcript expression (Figures 4G-4I and S8A).These observations support the hypothesis that distal Wnt-active cells contribute to the osteoblast cell pool in the distal regenerate; however, genetic fate mapping of DMB cells will be required to confirm this.

Fgf signaling controls osteoblast and distal blastema compartment sizes
Next, we tested whether signaling pathway alterations would affect the size and function of these different regenerate domains (Wnt-active tip region versus osterix reporter+ domain).9][40][41] Suppression of Fgf signaling by SU5402 in siam:mCherry x osterix:GFP zebrafish from 3 to 5 dpa led to an overall reduced regenerate length (Figure 5A, arrowhead in Figure 5B).This reduction was attributable to a shortened osterix+ domain, while neither the size and Wnt reporter activity of the osterixÀ, siam+ tip region nor the activity of the osterix:GFP+ cells was affected (Figures 5A-5C and S9B).We let fins recover from the inhibitor treatment to test whether Fgf-suppressed regenerates would catch up in growth.Indeed, fins recovered regenerate length during a two-day recovery period in which both the size and the reporter activity of the distal Wnt-active domain increased (Figures 5D-5F).Furthermore, osterix:GFP signal intensity was higher in regenerates recovering from Fgf suppression, and the osterix:GFP domain had enlarged when compared to 5 dpa, suggesting enhanced osteoblast maturation (and potentially proliferation) during catch up (Figure 5F).These results indicate that the Wnt+ domain (encompassing DMB cells), pre-osteoblasts, and potentially committed osteoblasts compensate for impaired regenerate growth after Fgf signaling suppression which cannot erase the presence of positional identity in the regenerate.

Osteoblast ablation impairs regeneration domain-specifically and regenerates can partially recover
We further investigated the contribution of different osteoblast populations to bone regeneration by ablating either osterix+ cells (proximal regenerate), mmp9+ cells (distal regenerate and joints), or both cell populations simultaneously.Transgenic zebrafish expressing nitroreductase (NTR) in the respective cells 7,8,42 were treated with the substrate nifurpirinol (NFP), 43 resulting in toxic product formation therein.osterix:NTR/mmp9:NTR double ablation from 3 to 5 dpa reduced regenerate length significantly when compared to individual osterix+ and mmp9+ cell ablation (Figures 6A and 6B).Individual osterix:NTR+ cell ablation had a stronger anti-regenerative effect than mmp9:NTR+ cell ablation.This confirmed the importance of osterix+ cells for bone regeneration and corroborated the finding that osterix+ and mmp9+ cells represent distinct fin regenerate cell populations.
We wondered how ablation affected proliferation in the regenerate and characterized the consequences of osterix+ and mmp9+ cell ablation by applying the S-phase marker Bromodeoxyuridine (BrdU) during the last 6 h of ablation.5 dpa double-ablated fin regenerates showed a normal rate of proliferation in the epidermis but reduced mesenchyme (blastema and osteoblast) proliferation (Figures 6C and 6D), pointing to reduced cell expansion as one cause for impaired regeneration, in addition to death of ablated cells.
Next, we asked whether Fgf signaling would be pivotal in the recovery function of the distal regenerate after impairment of regeneration.To test this, we ablated mmp9+ and osterix+ cells from 3 to 5 dpa, determined regenerate length as well as osterix+ and osterixÀ domain sizes, and treated the zebrafish with SU5402 during the recovery phase thereafter (Figure 6E).An increase in regenerate length from 5 to 7 dpa was observed in control (DMSO) conditions without ablation and individual mmp9/osterix cell ablation, as well as after double ablation (Figures 6F, 6G, S10A, and S10B); however, addition of regenerative tissue in double-ablated fins lagged somewhat behind compared to single cell-type ablated fins (Figures 6G and S10B).Notably, Fgf inhibition did not affect recovery in any of the ablation conditions (Figures 6H, S10A, and S10B) indicating that Fgf signaling is dispensable for recovery function.
osterix+ osteoblast ablation impairs bone matrix formation and patterning while mmp9+ cell ablation does not We went on to test whether cell ablation would irreversibly affect specific domains in the elongating regenerate.In order to distinguish recovering from non-recovering regions after ablation, we used the runx2:GFP+ reporter as an indicator for (pre-) osteoblast presence in runx2:GFP x osterix:NTR transgenic zebrafish at 5 dpa (i.e., after 2 days of NTR-mediated osterix+ cell ablation), and at 7 dpa after 2 days of recovery (Figures 7A-7C).At 5 dpa, osterix+ osteoblast-ablated fin regenerates were significantly shorter overall, while the length of individual domains (runx2:GFP+ region proximal to fin ray bifurcations, runx2:GFP+ region distal to fin ray bifurcations, tip region devoid of runx2:GFP signal) was only mildly reduced (Figure 7B).Two recovery days later, the distal runx2:GFP+ region showed normal domain length, while the runx2:GFP+ region proximal to bifurcation was significantly shorter and of the same length as that directly after ablation.In comparison, control-treated fins possessed a more than 2-fold longer pre-bifurcation proximal domain (Figures 7A and 7C).Furthermore, the size of the tip domain was slightly (albeit insignificantly) longer in fins that had undergone osterix+ osteoblast ablation.This indicates that distal regenerate regions less affected by osterix+ cell ablation regrow to pre-amputation size, independent of regeneration defects in proximal parts of the fin.It also suggests that distal regenerate cells might partly compensate for impaired regeneration of proximal fin tissue.Next, we tested how mmp9:GFP+ cells, which partly reside at developing segment joints, would react to osterix+ cell ablation.In control treatment conditions, mmp9:GFP+ cells were visible in up to 3 forming segment joints as well as in the distal portion of the regenerate of mmp9:GFP x osterix:NTR reporter zebrafish (5 dpa, Figure 7D).In contrast, NFP-treated fin regenerates undergoing osterix+ cell ablation lacked confined expression of mmp9:GFP at segment joints and showed much weaker mmp9 expression in a single broader domain at about 50% proximodistal level of the shorter fin regenerates, potentially reflecting dying cells (Figure 7D).At the same time, the segment joint indicator pthlha was not expressed in its usual pattern anymore (Figure S11A).Inspection of osterix+ cell ablated fins showed that segment joints did not form properly, although control-treated fins formed a minimum of one segment joint in a relatively proximal position (Figures 7D and S11B).Notably, in the reverse scenario of mmp9+ cell ablation, segment joint formation was observed (black arrowheads in Figures 6A and S11C), in agreement with a previous report on intact joints after mmp9+ cell ablation. 8This indicates pronounced patterning defects in case osterix+ osteoblasts are depleted from fin regenerates.We let mmp9:GFP x osterix:NTR regenerating zebrafish recover from vehicle/NFP treatment for 2 days.During this time, NFP-treated fins re-initiated segment joint formation in the distal part of the regenerate, resulting in an overall reduced segment joint number (Figures 7E and S11D), again indicating independence of the distal regenerate domain of proximal deterioration.
We asked why regenerate outgrowth and patterning were severely affected after osterix+ cell ablation and hypothesized that bone matrix secretion and maturation were diminished, leading to a lack of structural support in fin regenerates.Inspection of the sc RNA sequencing dataset suggested that chondroitin sulfate (CS), an indicator of matrix maturation, 44 is produced by cells of the BLWE and non-osteoblast blastema and strongly in osteoblasts, as inferred from chsy1 and chpf2 expression (Figures S11E and S11F).Consequently, we stained for the presence of CS, in addition to investigation of the presence of Laminin as an indicator of basement membrane integrity of the BLWE overlying osterix+ osteoblasts. 45,46Ablation of osterix+ osteoblasts led to strong CS reduction in the regenerate while CS levels in stump bone matrix were unaffected (Figure 7F).Laminin expression was maintained after ablation; however, it showed irregular and ill-defined distribution indicating impaired BLWE integrity after ablation (Figure 7F).This indicates that structural ECM integrity is required for appropriate patterning of the different functional domains in the fin regenerate.

DISCUSSION
With this work, we provide a dataset on the transcriptomic landscape of rare mesenchymal (DMB, adjacent proximal blastema, osteoblast) and epidermal cell populations of the 3 dpa zebrafish fin regenerate browsable at the Single Cell Portal (https://singlecell.broadinstitute.org/)8][49] The dataset contains cells which may have been underrepresented in whole fin regenerate sequencing datasets due to the high prevalence of other cells 12,13 but which are important to investigate the lineage specification of bone-forming cells and DMB cells serving as a signaling center throughout regeneration.Of note, the dataset does not contain mesenchymal cells located in the proximity of the amputation plane.As we have used cells from fin regenerates after repeated amputation, the dataset can also be used to study potential changes arising after recurrent injury, although our analysis suggests that gene expression in the distal regenerate is generally not altered between different rounds of amputation.This robustness is in line with previous reports investigating gene expression after repeated amputation. 50,51

Spatial arrangement and diversity of mesenchymal (DMB, adjacent proximal blastema, osteoblast) and BLWE cells in the regenerate
9,47 Other studies, such as by Brown et al., 10 Stewart et al., 11 and Wehner et al. 27 have provided insight on the hierarchy and regulation of distal Runx2+ pre-osteoblasts, and a more proximal osterix+ osteoblast cell pool in the regenerate.Here, we uncovered the complex tissue constitution of the distal regenerate by identifying 5 osteoblast cell populations, including previously described committed osteoblasts, for which we present novel marker genes, two populations of proliferating osteoblasts, osteoblasts with a transcriptomic profile akin to (non-osteoblast) blastema cells, and joint cells located at the segment boundaries.We have analyzed a comparatively high number of osteoblasts (2,321 cells) of different maturity level and proliferative capacity and, as we suggest, proximodistal position.This proximodistal position (Figure 2J) can be inferred from combinatorial gene expression analysis.Importantly, we identified two populations of cells (Blas-tema3 and Osteo1), which form a transition zone between non-osteoblast blastema and osteoblast cell clusters of the regenerate.The observation that BLWE cell clusters separated completely from blastema and osteoblast cell clusters suggests that lateral BLWE do not contribute to the mesenchymal cell pool (and vice versa).
Committed osteoblasts contribute to bone formation in the proximal regenerate and are supported by immature cells located at the tip Here, we used genetic lineage tracing and ablation of osterix+ and mmp9+ cells, including Osteo4, Osteo1, and Osteo2 joint cells, to estimate the contribution of both cell populations to bone formation in the caudal fin regenerate.Thus, this work reveals their complementing contribution to regeneration and expands previous knowledge on the lineage restriction of osterix+ cells 4,7 and the progenitor cell function of  mmp9+ cells in the fin stump and regenerate. 8From 3 to 5 dpa, amplification of Osteo4 cells led to progeny covering $80% of the regenerate's length demonstrating a considerable contribution, supported by a strong reduction in regenerate length after osterix+ cell ablation during regeneration.These observations are in line with an earlier report of impaired regeneration upon osterix+ cell ablation in the fin regenerate post-amputation (Figure S3C in 7 ).At the same time, the restricted distal expansion of Osteo4 progeny suggested a somewhat limited renewal capacity, which is reflected by the partial expression of pcna in this population.In salamanders, transplantation experiments of different Crelabeled populations (Col1a2 vs. Prrx1) suggested that distal skeletal elements of the regenerated limb are derived from non-skeletal connective tissue cells. 47In our experiments, lineage tracing showed the absence of Osteo4 progeny from the distal 250 mm of the regenerate, suggesting alternative sources for this region.Candidate populations are stmn1a+ Osteo3 osteoblasts, which have a high proliferative activity (Figure S4G), and Osteo1 cells, which are found in the transition zone between blastemal and osteoblast cell clusters in our sc RNA sequencing dataset.Notably, strong mmp9 expression is found in cells belonging to these and the Osteo0 clusters (Figure 3F), in addition to previously reported expression of mmp9 in segment joint cells and epidermis. 8Occasional expansion of mmp9:CreERT2-converted cell clones at the very tip of the regenerate (Figures 3E and 3I) suggests contribution of mmp9+ cells to an osteoblast progenitor cell pool in this distal region of the regenerate but also hints at expansion of epidermal mmp9+ cell populations. 8The hypothesis that, within the blastema, mmp9+ cells contribute to distal osteoblast progenitors is supported by the observed effects on simultaneous mmp9+ and osterix+ cell ablation, which decreased fin regenerate growth more dramatically than individual ablation.However, depletion of some mmp9-expressing myeloid cells in the fin regenerate 8 might possibly add to the seen additive regeneration defects, as indicated by larval and adult macrophage ablation experiments during zebrafish tissue regeneration. 52,53ur results concerning the lacking expansion of mmp9+ cells at segment joints are unexpected in light of a prior study on the comprehensive potential of mmp9+ cells during fin regeneration.Ando et al. 8 demonstrated a significant contribution of mmp9+ joint cells (of the fin stump) to newly forming fin regenerates.Fate mapping of the same cells in homeostatic fins resulted in broad labeling along the uninjured fin ray.A potential reason for this seeming discrepancy is the different experimental regimen that we have used by initiating a single recombination event post-amputation (not pre-amputation) with potentially much lower mmp9 and CreERT2 expression in joint-forming cells of the early regenerate.Furthermore, a different CreERT2 responder line (with however strong and ubiquitous reporter gene expression upon heat shock) was used.It is possible that mmp9+ joint-forming cells of the early regenerate are less potent than mmp9+ cells of the mature fin ray which might be indicated by the low fraction of proliferating Osteo2 cells at 3 dpa.Another seeming difference of our study and prior work concerns the consequences of NTR-mediated mmp9+ cell ablation. 8We noted a decreased regeneration potential (i.e., shorter fin regenerates) after mmp9+ plus osterix+ cell ablation within the early fin regenerate, while single mmp9+ cell ablation initiated before amputation did not lead to shorter fin regenerates in previous work. 8Again, experimental differences (ablation in the mature fin ray vs. the early regenerate) might cause the opposing observations.Notably, even though double ablation in the early regenerate has strong anti-regenerative effects, fins can recover once ablation treatment is discontinued.This supports the presence of other populations contributing to regeneration, and we suggest that distal blastema cells are of importance in this context.
The origin of osteoblast progenitors in the distal portion of the regenerate has been vague.Distal Runx2+ cells are maintained as a pre-osteoblast cell pool under the influence of Wnt signaling. 11In line with this, we identified a region of runx2:GFP+ pre-osteoblasts showing very weak GFP fluorescence in close proximity to the DMB.The respective cells often co-labeled with mCherry protein produced in siam:mCherry Wnt-responsive cells (Figure S8C).We suggest that these cells have just started to differentiate into pre-osteoblasts and that they are derived from (non-osteoblast) blastema cells.Alternatively, these cells may represent descendants of runx2:GFP+ osteoblasts that have lost gfp expression and have relocated distally.A second population of more proximal mCherry, GFP double+ cells may partly be explained by overlap of their respective transcript domains.However, no such zone of concomitant gfp/mCherry expression was detected in double transgenic osterix:GFP x siam:mCherry zebrafish, in agreement with low Wnt signaling in osterix+ osteoblasts. 11,27Nevertheless, a considerable amount of GFP/mCherry protein double+ cells were detected in the proximal region of the regenerate.We explain this discrepancy between mRNA and protein levels with a relocation of distal blastema cells to a more proximal position, or by a distal movement of osteoblasts toward the tip, resulting in mixing of cells.This is in line with our TEM data showing a gradual phenotype change of cells underlying the BLWE along the proximodistal direction.We conclude that a lineage dependence exists between (non-osteoblast) blastema cells and osteoblasts in the distal fin regenerate and that this transition is laid out by Blastema3 and Osteo1 cells, followed by proliferative Osteo0/3 cells.mfap5, stmn1a, twist2, and runx2:GFP are markers for this transition.Since blastema clusters were continuous in our analysis, actinotrichia-forming cells and even DMB cells may ultimately be involved in this process.Interestingly, photoconversion of DMB kaede-labeled cells, carried out by Wehner et al. (2014), led to a small population of converted cells localizing to a slightly proximal position (Figure S2I in 27 ) and knockdown of the DMB ''marker'' mmp13a results in impairment of osteoblast commitment and differentiation in the regenerate. 29Altogether, we propose a model in which committed osteoblasts maintain themselves in the proximal domain of the regenerate, while the distal domain of the regenerate is populated by immature osteoblast progenitors derived from mmp9+ pre-osteoblasts, and runx2:GFP+ cells that are provided by distal blastema cells.Notably, this distal domain is able to compensate for diminished regenerate growth after a challenge such as Fgf signaling inhibition, demonstrating the adaptability of this region.

The activity and size of the different regenerate domains depend on the presence of growth factors and reflect the need for continued regenerate growth
Here, we show that manipulation of growth factor signaling affects domain extension in the regenerate, similar to altered domain sizes after osteoblast ablation.Inhibition of Fgf signaling specifically reduced the domain size of osterix+ cells while leaving their reporter activity as well as the size of the osterixÀ, siam+ distal tip region unchanged during inhibition.This indicates a profound effect of Fgf signaling on newly differentiating osterix+ cells.This effect is reversible, and both the size and activity of the siam+ osterixÀ tip region and the activity of osteoblasts in the osterix:nGFP+ domain increase after discontinuation of Fgf suppression.Recovery of fin regenerate length is then completed within a few days, which is accompanied by the presence of an enlarged tip domain, suggesting intrinsic mechanisms detecting the need for enhanced regeneration which is likely carried out by enhanced proliferation at the junction of distal pre-osteoblasts and osterix+ osteoblasts.Notably, positional information cannot be overridden by Fgf suppression.The same is true for fins that have undergone a more dramatic, simultaneous ablation of both mmp9+ and osterix+ cells in fin regenerates.These fins catch up in distal growth when ablation is stopped, albeit at a somewhat slower pace than fins with single mmp9+ or osterix+ cell ablation (Figure 6G).

Regenerate domains compensate for each other as long as structural support is provided
Here, we show that after ablation of osterix+ cells in the proximal regenerate, the runx2:GFP+ cell population distal to ray bifurcation expanded normally, and that the runx2À tip region remained slightly longer in osterix+ cell ablated fins than in control-treated fins.In contrast, the domain proximal to ray bifurcations, which did lose structural support, i.e., bone matrix, did not recover, therefore leading to a significantly reduced domain length.In the future, it will be important to test whether osterix+ cell ablated fins can grow to their pre-amputation size thanks to the tip region or whether regenerates remain significantly shorter because of the lack of proximal tissue.Given that the runx2:GFPÀ tip region is already slightly longer after a 2-day recovery, this region might take over ''responsibility'' for regenerate growth.This would be in agreement with other experimental settings we have tested, in which the tip region of the regenerate increased in size to make up for suppressed regenerate growth.
In our work, we detected patterning defects affecting segment joint formation in case osterix+ osteoblasts lining segments (but not segment boundaries) were ablated.We did not see segment joint formation defects in case of mmp9+ cell (i.e., joint forming cell) ablation, as reported previously. 8This suggests that osterix+ cells and the resulting deposited bone matrix are required for appropriate segmentation of the fin.Altogether, this demonstrates a pivotal function of committed osteoblasts, but also pre-osteoblasts and distal blastema cells, for growth, bone formation, and patterning of the regenerating vertebrate appendage.

Limitations of the study
Here, we have analyzed selected fin regenerate cell populations via FACS enrichment and sc RNA sequencing.A detailed investigation of other fin regenerate mesenchyme populations (i.e., fibroblasts and osteoblasts found close to the amputation plane) along with distal WE cell populations is lacking.Furthermore, genetic fate mapping of DMB cells via CreERT2 should be performed, along with testing for potential dissimilarities of available Cre responder lines.In the future, tissue-specific inhibition of Fgf signaling during regeneration is desirable.

METHOD DETAILS Fin clips
Fish were anesthetized in Tricaine 0.02% (Sigma-Aldrich, #A5040) and the caudal fin was resected at 50 % of its length using a scalpel.Animals were transferred to fish water and allowed to regenerate at 28 C. 41 3, 5 and 7 dpa regenerates were obtained from fin-clipped male and female zebrafish.

Tissue dissociation and flow cytometry
Fin regenerates of quadruple transgenic fish (siam:mCherry, shh:GFP, osterix:CreERT2-p2a-mCherry, runx2:GFP) were harvested at 3 dpa, cut into small pieces with a scalpel and transferred into 1 ml collagenase-dispase solution (1 mg/ml in PBS, Roche #10269638001) for 10 min at 28 C. The sample was pipetted slowly up and down with an elongated, flame polished Pasteur pipette.The procedure was repeated 4 times with decreasing inner tip diameters of the Pasteur pipettes until a homogenous solution was obtained.The dissociates were poured onto an equilibrated 70 mm cell strainer and collected in 10 ml HBSS solution (without CaCl 2 and MgCl 2 , Gibco #12082739).After centrifugation (15 minutes, 1800 rpm, 4 C), the supernatant was discarded and the remaining cell pellet resuspended in 500 ml 2 % BSA in PBS.Calcein violet (1ml 10mM, Invitrogen #C34858) was added to the cell solution and incubated for 30 minutes.Calcein violet, GFP and mCherry+ cells were collected in 50 ml 2% BSA in PBS via fluorescence-activated cell sorting (BD LSR Fortessa) and processed for single-cell RNA sequencing analysis based on 10X Genomics (10X Chromium system, 10X library preparation according to the manufacturer's instructions).

Single-cell RNA sequencing analysis
For each experiment about 8000 cells from zebrafish fin regenerates were flow-sorted into BSA-coated PCR tubes containing 1 ml of PBS with 0.04 % BSA.All cells were carefully mixed with reverse transcription mix before loading them in a Chromium Single Cell A Chip on the 10X Genomics Chromium controller 57 and processed further following the guidelines of the 10X Genomics user manual for single cell 3' RNA-seq v2.In short, the droplets were directly subjected to reverse transcription, the emulsion was broken and cDNA was purified using silane beads.After amplification of cDNA with 12 cycles, it underwent a purification with 0.6 volume of SPRI select beads.After quality check and quantification using the Fragment Analyzer (Agilent), 30 ng cDNA were used to prepare sc RNA-seq libraries -involving fragmentation, dA-Tailing, adapter ligation and a 12 cycles indexing PCR based on manufacturer's guidelines.After quantification, both libraries were sequenced on an Illumina Nextseq500 system in paired-end mode with 26 bp/57 bp (for read 1 and 2 respectively), thus generating $60-90 mio.fragments for the transcriptome library on average.A custom reference based on GRCz10, Ensembl annotation e98 was created, by first adding the sequences of gfp and mCherry as separate chromosomes to the fa file and the gtf file and then building the cellranger reference using cellranger mkref.Next, fastq files were processed with cellranger count from 10X genomics (https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/using/count) version 3.0.0using the custom reference.This resulted in a dataset with 2532, 3601, 2481 and 3413 mean counts per cell and median number of 300, 532, 975 and 508 detected genes per cell, and 7028, 4707, 2846 and 4535 cells for regeneration experiments Reg1, Reg2, Reg3 and Reg4, respectively.
For the downstream analysis of the 10X data, current best practices were followed. 14The filtered gene counts matrices were read with scanpy 1.6.0. 58Next, cells were filtered based on the number of total counts, the number of detected genes and with sample specific thresholds: Only cells with more than 2000, 1000, 1000 and 1000 total counts as well as more than 600, 300, 300 and 400 detected genes for samples Reg1, Reg2, Reg3 and Reg4, respectively, were kept for the analysis.Furthermore, cells with more than 5% of mitochondrial reads were filtered.After quality control, our dataset consisted of 1342, 1410, 2102 and 1814 cells from the respective regeneration experiment.Only genes which were detected in more than 3 cells (counting all samples together) were kept for downstream analysis.Normalization was performed with the scanpy function sc.pp.normalize_total and the data was log-transformed with sc.pp.log1p.Highly variable genes were detected with sc.pp.highly_variable_genes setting n_top_genes=4000.Principal component analysis was performed on the highly variable genes.A neighbor graph was constructed with sc.pp.neighbors setting n_neighbors=30 and n_pcs=15.Next, a UMAP was constructed with sc.tl.umap setting min_dist=0.9. 59Clustering was done using sc.tl.leiden with resolution=0.2.Marker genes were computed with sc.tl.rank_genes_groups.Two clusters were identified as blood vessels and immune cells (autofluorescent cells) and those were excluded from further analysis, leaving 1219 (Reg1), 1331 (Reg2), 2007 (Reg3) and 1705 cells (Reg4).Next, principal component analysis, neighborhood graph construction, UMAP computation and marker gene detection were repeated with the same functions and parameters as above.In result, the presented UMAPs show 691 BLWE cells (Basal), 3250 (non-osteoblast) blastema cells (Blastema), and 2321 osteoblasts (Osteo).The 3 main clusters were sub-clustered using sc.tl.leiden setting the restrict_to parameter to the respective cluster.The resolution parameter was set to 0.25, 0.1 and 0.35 for the blastema, BLWE and osteoblast clusters, respectively.For each sub-cluster, marker genes were computed as above but using only the cells of the respective main cluster as a reference.Trajectory inference was performed with sc.tl.paga and the paga plot was created sc.pl.paga_compare setting threshold=0.1. 15To compute RNA velocities, we used the velocyto.pycommand line interface on the cellranger bam files to create a loom file that contains spliced and unspliced reads. 60We read this loom file with scvelo and merged it with the AnnData object that was created by the scanpy analysis above. 61The resulting data was processed according to the scvelo tutorial: In detail, filter_and_normalize was run with min_shared_counts=20 and n_top_genes=2000, next moments were computed, and the velocities were computed with mode ''stochastic''.A velocity graph was computed and the result was visualised on the UMAP using pl.velocity_embed-ding_stream.For differential expression analysis, we used a limma-voom workflow: only highly expressed genes were considered, i.e. genes with a cpm value>1 in more than 25% of the cells. 62The AnnData object was converted to an edgeR DGEList object using anndata2ri (https:// github.com/theislab/anndata2ri)and scran. 63,64Normalisation factors of the raw count matrix were computed with edgeR's calcNormFactors function.The number of detected genes per cell was scaled to zero mean and unit variance and added as a co-factor to the design formula.The other factor that was added was the condition (Reg1 vs Reg2-4 pooled together).Next, the DGE list was transformed using limma's voom function. 65Next, edgeR's functions lmFit, contrasts.fit,treat (with lfc=log2(1.5)) and topTreat were used to generate a list of differentially expressed genes.Next, pseudospace coordinates were computed.For the lateral coordinate, the gene set lamb1a, wnt5b, shha and phlda2 were used.A laterality score was computed using the scanpy function sc.tl.score_genes.Next, the transcriptome data object was subset to the lateral gene set.Diffusion pseudotime was computed by running sc.pp.neighbors with n_neighbors=50, sc.tl.diffmap and sc.tl.dpt on the subsetted transcriptome data.For sc.tl.dpt, the cell with the highest laterality score was used as the root.The lateral coordinate of pseudospace was then computed by a rank transformation of the pseudotime.For the distal coordinate, the same computation was performed using the gene set aldh1a2, wnt5a, fgf3, fgf10a, igf2b, msx3, msx1b, cdh4, dkk1b, wnt3a, dkk1a, dlx5a, junba, junbb, msx2b, spry4 and wnt10a.

RNA in situ hybridization and histology
The plasmids to obtain probes for fgf24 (restriction digest with NotI, transcription with T7), and pthlha (restriction digest with BamHI, transcription with SP6) have been published. 32,66mCherry (restriction digest with HindIII and transcription with T7), and gfp (restriction digest with EcoRI, transcription with T3) plasmids were provided by Gilbert Weidinger.The CreERT2 probe plasmid (restriction digest with AgeI, transcription with T7) was provided by Stefan Hans.The tnc in situ probe plasmid was generated by Rene ´Bernitz and provided by Daniel Wehner.The pthlha in situ probe plasmid 32 was provided by Marie-Andre ´e Akimenko.Additional probes have been synthesized from pCR-Blunt II-TOPO vectors after cloning of the respective sequences by using the Zero Blunt TOPO PCR cloning Kit (Invitrogen) according to the manufacturer's instructions.The forward and reverse oligonucleotides to amplify the cDNA sequences are listed below (Table Oligonucleotides), along with the corresponding fragment sizes.
For whole mount RNA in situ hybridization, fin regenerates were fixed in 4 % PFA in PBS overnight at 4 C.After fixation, fins were washed in PBS and dehydrated in methanol for storage at -20 C. Fins were rehydrated gradually in methanol solutions in PBT (PBS, 0.1% Tween 20) of 75 %, 50 % and 25 % at room temperature for 5 minutes each and washed with PBT (4x 5 minutes).Fins were digested for 20 minutes using ) was combined with Fast Red staining for simultaneous detection of two transcripts.After staining with NBT/BCIP, fins were briefly rinsed twice in PBT and transferred to new tubes.This was followed by incubation in 0.1M Glycin/HCL, pH 2,2 + 0.1% Tween (2x 5 minutes) and washing with PBT (4x 5 minutes).200ml Anti-Fluorescein-AP, Fab fragments antibody (Roche, 11426339810) was added (1:2000 dilution in PBT + 2 mg/ml BSA + 2% sheep serum) and incubated over night at 4 C.The fins were kept dark and washed in PBT (2x 5 minutes, 6x 30 minutes).Fins were then washed in 0.1M Tris pH 8.2 + 0.1% Tween (3x 5 minutes).SIGMAFASTä Fast Red TR/Napthol AS-MX Tablets (Sigma, F4648) were dissolved according to the manufacturer's instructions.Specimens were kept dark and at room temperature in 500 ml Fast Red staining solution until the staining had developed.The staining reaction was stopped by two brief washes with PBT and one wash in STOP solution.Fins were stored in 80% glycerol/20% STOP solution thereafter and processed for cryosectioning.Cryosections were imaged with a Zeiss 10x/0.45Plan-Apochromat air objective on an ApoTome1 equipped with a Zeiss AxioCam MRc color CCD camera and a Zeiss ZEN blue (v 2012) software.

Electron microscopy
Fin regenerates were cut off distal to the amputation site and were fixed in 4% PFA in 100 mM phosphate buffer, pH 7.4.Samples were further dissected for embedding into epoxy resin and processed according to a modified protocol for serial block face SEM 67 using osmium tetroxide (OsO 4 ), thiocarbohydrazide (TCH), and again OsO 4 to generate enhanced membrane contrast. 68,69In brief, samples were postfixed overnight in modified Karnovsky fixative (2% glutaraldehyde/2% formaldehyde in 50 mM HEPES, pH 7.4), followed by post-fixation in a 2% aqueous OsO 4 solution containing 1.5% potassium ferrocyanide and 2mM CaCl 2 (30 minutes on ice), washes in water, 1% TCH in water (20 minutes at room temperature), washes in water and a second osmium contrasting step in 2% OsO 4 /water (30 minutes on ice).Samples were washed and en-bloc contrasted with 1% uranyl acetate/water for 2 h on ice, washed in water, and dehydrated in a graded series of ethanol/water mixtures (30%, 50%, 70%, 90%, 96%), followed by 3 changes in pure ethanol on molecular sieve.Samples were infiltrated into the epon substitute EMBed 812 (resin/ethanol mixtures: 1:3, 1:1, 3:1 for 1h each, followed by pure resin overnight and for 5hrs), embedded into flat embedding molds, and cured at 65 C overnight.Ultrathin sections (70 nm) were prepared with a Leica UC6 ultramicrotome (Leica Microsystems, Wetzlar, Germany) and a diamond knife (Diatome, Nidau, Switzerland), collected on formvar-coated slot grids, and stained with lead citrate 70 and uranyl acetate.In total, eight regenerating fin rays from five regenerates were analyzed.From each ray, three grids with six to nine sections were prepared and imaged.Mounted sections were analyzed with a JEM 1400Plus transmission electron microscope (JEOL, Freising, Germany) at 80 kV and images were taken with a Ruby digital camera (JEOL).

Drug treatments
Fgf inhibition was performed with 17 mM SU5402 (Selleckchem) or the vehicle control dimethylsulfoxide (DMSO, final percentage 0.1 %) in fish water at the times indicated.Solutions were changed daily.50 mg/ml 5-Bromo-2-deoxyuridine (BrdU) (Sigma-Aldrich #B5002) stock solutions were prepared in DMSO and used at 5 mM (Figures 5A-5F and 7A-7F) or 2.5 mM (Figures 6C-6H) in selected experiments, either in fish water, in fish water supplemented with NFP (Sigma-Aldrich) or DMSO (3 % DMSO maximum), or in fish water supplemented with SU5402 or DMSO.BrdU treatment was performed during the last 6 hours of day 5 post amputation by incubation.NTR mediated ablation was performed with 1 mM NFP in fish water.Fin regenerates were either fixed and/or photographed at 5 dpa or zebrafish were allowed to recover from treatment for 2 days before fin regenerates were fixed and/or photographed.

Statistical analysis
Statistical analysis was performed using Prism 6 (GraphPad Software, La Jolla, CA, USA), with the statistical tests and corresponding p values reported in the figures and respective legends.All values represent the mean G SD unless otherwise stated.Additional information is listed in Table S9.

Live imaging, quantification of fluorescent cells and plot profile measurements
For quantification of GFP expression fish were anesthetized with 0.02% Tricaine (MS222) and their fins were imaged with a Zeiss SteREO Discovery.V12 microscope equipped with a AxioCam MRm camera and AxioVison software version 4.7.1.0.Identical settings for magnification, exposure time, gain, and contrast were used.Live imaging of siam:mCherry x runx2:GFP and siam:mCherry x osterix:GFP transgenic zebrafish was performed on a Dragonfly Spinning Disk confocal equipped with an Andor Zyla PLUS monochrome sCMOS camera and Fusion software with a z interval of 2 mm and a 30x silicone objective.GFP and mCherry double+ cells were counted in 50 mm intervals beginning with the most distal mCherry signal (0 mm) up to 400 mm.Quantification of reporter fluorescence along fin rays of transgenic siam:mCherry x osterix:GFP zebrafish was done with the plot profile tool in Image J/Fiji.