Notch signaling enhances bone regeneration in the zebrafish mandible

Although bone has some degree of intrinsic healing ability, humans lack the capacity for large-scale skeletal regeneration following major trauma. Surgical resection and/or reconstruction of the mandible is indicated for numerous clinical problems, including primary oral cancers with bony invasion (Shah and Gil, 2009); 25% of patients with oral squamous cell carcinoma are treated with segmental mandibulectomy (Rogers et al., 2004). High-energy trauma (Vaca et al., 2018), osteomyelitis (Marschall et al., 2019), antiresorptive-related osteonecrosis (Khan et al., 2015) and elective orthognathic correction (Alyahya and Swennen, 2019) constitute other indications. The current standard of care for reconstruction of segmental mandible defects is the vascularized free fibular flap (Kim et al., 2013), which provides bony continuity at the expense of increased mortality, increased pain and decreased quality of life (Rogers et al., 2004). There is a significant unmet clinical need for regenerative therapies to augment bone healing, particularly for craniofacial applications for which form is linked not only to function, but also to social competence and mental health (Snyder et al., 2005). Moreover, ageing, radiation/chemotherapy (Hoff et al., 2008) and polytrauma (King et al., 2004) are common co-morbid risk factors for non-union in this patient population. To improve our understanding of how to boost endogenous bone healing, we investigated mandibular healing in an organism that naturally regenerates such defects without exogenous intervention.

Zebrafish (Danio rerio) are an ideal companion model organism to laboratory mice (Busse et al., 2020), in part because of an abundance of molecular and genetic tools, high throughput, and low cost. Similar to the mouse, approximately 70% of all human genes and 80% of disease-causing genes have a zebrafish ortholog (Howe et al., 2013), and zebrafish have cellular skeletons with all human bone cell types (Bergen et al., 2019). Despite this high degree of genetic conservation, and in contrast to mammals, zebrafish are capable of exceptional bone healing, including appendicular regeneration following amputation (Pfefferli and Jaźwińska, 2015). The adult caudal fin undergoes epimorphic regeneration in 2-3 weeks following near-complete amputation, an observation reported in the 18th century (Broussonet, 1789) but now understood with increasingly sophisticated cellular and molecular resolution. Fin ray bones regenerate through both dedifferentiation of existing osteoblasts (Knopf et al., 2011) and transdifferentiation of intra-ray fibroblasts (Singh et al., 2012), highlighting the complex nature of skeletal regeneration in zebrafish.

We have reported that the zebrafish mandible is capable of large-scale bone regeneration in adults (Paul et al., 2016). The mandible is developmentally derived from cranial neural crest cells of the first pharyngeal arch (Schilling et al., 1996). Regeneration of the mandibular bone occurs through the formation of cells with a hybrid phenotype between cartilage and bone, with the morphological cartilage callus directly transforming into bone. Similar hybrid cartilage-bone cells have been observed in regeneration of the murine rib bone: one of the few bones in mammals capable of larger-scale regeneration (Kuwahara et al., 2019). Hedgehog signaling in Sox9-expressing periosteal cells promotes mineralization of this hybrid callus (Paul et al., 2016). However, our understanding of the pathways regulating the formation and maturation of the regenerative fracture callus is still incomplete.

Alagille syndrome (Li et al., 1997), primarily caused by heterozygosity of loss-of-function mutations in jagged 1 (JAG1; the most abundant Notch ligand in bone) results in decreased bone mass (Olsen et al., 2005), pathologic extremity fractures (Bales et al., 2010) and characteristic midfacial hypoplasia (Oda et al., 1997). Conditional genetic mouse models targeting Jag1 in neural crest cells (Hill et al., 2014; Humphreys et al., 2012) and in osteoblasts (Youngstrom et al., 2016) recapitulate many of these skeletal abnormalities. Ongoing pre-clinical investigations by our group (Youngstrom et al., 2017) and others (Kamalakar et al., 2021; Novak et al., 2020) suggest that Notch signaling agonists can improve bone healing outcomes in mice. This prompted us to study the function of this pathway during the formation of regenerative osteochondral mesenchyme in zebrafish.

Notch signaling has been previously implicated in zebrafish caudal fin regeneration, a condition in which jag1a and her6 (orthologous to human HES1) are highly expressed (Schebesta et al., 2006). In this model, Notch1a signaling functions to maintain bone-derived blastema cells in a proliferative and undifferentiated state (Grotek et al., 2013), and regulates formation and remodeling of the supporting vascular plexus (Kametani et al., 2015). Whereas dedifferentiated osteoblasts are responsible for intramembranous bone regeneration at numerous skeletal sites in zebrafish (Geurtzen et al., 2014), chondroid regeneration of the mandible is mechanistically distinct. To investigate the function of Notch signaling within adult mandibular defect healing, we used pharmacological and genetic approaches to modify Notch signaling following experimental mandibulectomy. We find that Notch signaling is required for cartilage callus formation, and that augmented Notch signaling early in repair can improve mandibular bone regeneration, likely by accelerating the conversion of the hybrid cartilage callus to bone.

We produced a novel transgenic reporter fish harnessing the promoter of her6, a direct transcriptional target of canonical Notch receptor signaling. We verified overlap between the her6:mCherry reporter and the endogenous Notch1a intracellular domain (NICD) at 10 days post-resection (dpr) at the injury margin by immunohistochemistry (IHC) (Fig. 1A). To understand the temporal regulation of Notch signaling in the context of mandible regeneration, her6:mCherry animals subjected to unilateral mandibulectomy were harvested for histological analyses (Safranin O staining  and mCherry IHC) at 0, 4, 7, 10 and 14 dpr (Fig. 1B). At 0 dpr, her6:mCherry⁺ cells were visible in the skin, weakly within Meckel's chondrocytes, and among the stromal cells on the medial aspect of the Meckel's cartilage. Between 4 and 7 dpr, a small number of her6:mCherry⁺ cells were visible lining the bone injury margins. At 10 dpr, the strongest her6:mCherry activity was observed in condensations of repair chondroblasts, as well as among the majority of the fibroblastic mesenchyme. her6:mCherry reporter activity began to wane at 14 dpr as the callus matured.

Because the greatest her6:mCherry activity was observed by IHC at 10 dpr, this time point was selected for subsequent bulk gene expression analysis by quantitative polymerase chain reaction (qPCR) (Fig. 1C). The 10 dpr callus is a highly dynamic and heterogeneous environment and, relative to paired intact contralateral mandibles, upregulation of numerous endochondral, matrix/remodeling and Notch-associated genes was observed. For example, sox9a was upregulated 3.16±1.68-fold (P=0.028), col1a2 was upregulated 4.80±3.45-fold (P=0.007), sp7 was upregulated 3.48±1.40-fold (P=0.003) and tnc was upregulated 6.50±3.25-fold (P=0.002). Upregulation of bglap did not reach statistical significance (P=0.059), but spp1 did (P=0.043). Analogous to our understanding of mammalian bone (Youngstrom et al., 2017), jag1b was the most abundantly expressed ligand in both the resected and contralateral mandible samples. However, jag2b and dlc were upregulated in the context of injury, whereas jag1b was not. notch1a was the only receptor to be significantly upregulated within the mandible callus. Notch ligands/receptors listed in Table S1 but not reported in Fig. 1C were below the limit of detection by qPCR. These data associate Notch signaling with the early healing response following segmental mandibulectomy.

To test the hypothesis that postoperative Notch signaling directly regulates mandible regeneration, complementary strategies were used to block and over-activate this pathway (Fig. 2A). In one experiment, three 16-h treatments of 20 µM dibenzazepine (DBZ) in dimethyl sulfoxide (DMSO) vehicle were used to pharmacologically inhibit activation of all Notch receptors every 48 h beginning 4 h post-resection. To conditionally overexpress the active NICD in all tissues using the Gal4/UAS system, hsp:NICD fish were treated with three 30-min heat shocks every 48 h beginning 2 h prior to surgery. DMSO-treated fish or hsp:NICD(−) fish subjected to the same heat-shock regimens were used as controls for each experiment, respectively.

X-ray microtomography (µCT) analysis of the callus at 32 dpr revealed striking differences in regenerate bone quantity, which were proportional to post-operative Notch signaling levels (Fig. 2B). DBZ treatment resulted in 58.6% less bone volume (BV) (P=0.017), 57.3% less bone surface area (BS) (P=0.007) and 59.5% less tissue mineral content (TMC) (P=0.018) relative to DMSO-treated siblings. This phenotype was even more severe in magnitude when DBZ treatment was consolidated into single 48-h treatments targeted within a range of times between 0 and 5 dpr (Fig. S1). Although the 48-h treatments were not powered to detect significant differences from each other, the 1-3 dpr treatment window had the lowest mean BV. In contrast to the reductions in regenerate bone observed following DBZ treatment, heat-shocked hsp:NICD⁺ animals acquired 89.6% more BV (P=0.013), 69.7% more BS (P=0.020), and 87.9% more TMC (P=0.019) relative to heat shocked hsp:NICD(−) siblings. Tissue mineral density (TMD) was not changed in either experiment, indicating that post-operative Notch signaling regulates quantity but not quality of regenerate bone tissue at 32 dpr.

Phenotypically wild-type animals generally achieve filling of the defect space with bone by 32 dpr. Perturbations of the Notch pathway in the early post-operative period interfered with this month-long process. DBZ treatment resulted in mandible non-union, whereas NICD overexpression increased bone formation. Fig. 2C,D depicts representative low- and high-healers in each group/experiment at 32 dpr. Incidentally, we also identified increased bone mass in T1 vertebral bodies of hsp:NICD⁺ animals (Fig. S2), suggesting global changes to skeletal metabolism.

To characterize further the DBZ-induced reduction in regenerative bone formation in her6:mCherry animals, Safranin O histology and mCherry IHC was performed on tissue sections from DMSO- and DBZ-treated animals (Fig. 3A,B). At 10 and 18 dpr, significantly reduced callus cellularity was observed in the DBZ group, and the robust her6:mcherry⁺ chondroid callus noted in the DMSO group (as well as in wild-type animals in Fig. 1B) was absent. In the DBZ group, her6:mCherry⁺ cells were only present in the skin and Meckel's chondrocytes at 18 dpr. Between 18 and 32 dpr, the cartilaginous callus in the DMSO group had typically resolved into bone, with no remaining cartilage. There was limited bone formation in the DBZ group, and delayed her6:mCherry⁺ activity was visible on bone surfaces and adjacent mesenchyme at 32 dpr in the DBZ group.

To determine whether arrest of callus formation was caused by defects in proliferation and/or apoptosis, 5-bromo-2′-deoxyuridine (BrdU) and caspase-3 IHC was performed at 6 dpr (Fig. 3C). There were fewer BrdU⁺ proliferating cells in the DBZ group (P=0.021), whereas apoptosis was not affected. Gene expression analysis of DBZ versus DMSO animals at 10 dpr confirmed transcriptional downregulation of her6 (69.0±13.9%), as well as a broad profile of osteochondral genes (Fig. 3D, Fig. S3). Representative genes and their percentage reductions include: sox9a down 19.7±8.2% (P=0.004), col1a2 down 77.8±4.3% (P<0.001), sp7 down 66.0±12.4% (P=0.001), among others. The dominant Notch ligand, jag1b, was downregulated 68.5±3.4% (P=0.001; Fig. S3). Thus, a broad spectrum of callus osteochondral markers and extracellular products were downregulated in DBZ-treated animals.

Fluorescence in situ hybridization (ISH) was used to assess the differentiation state of the hybrid cartilage-bone callus, using col2a1a and col1a1a as early markers of cartilage and bone, respectively. These genes are co-expressed in repair chondroblasts (Paul et al., 2016), and labeled the mid-callus mesenchyme of the DMSO group. DBZ treatment decreased expression of both markers relative to DMSO, and notably increased the col2a1a/col1a1a ratio (P=0.028) (Fig. 3E). We conclude that Notch signaling within a narrow post-operative window is required for the proliferative expansion and subsequent osteogenic conversion of the regenerative mandible callus. Failure of these processes as a result of Notch inhibition leads to atrophic non-union.

The cellular basis of enhanced healing in Notch1a transgenic animals was then assessed. Samples from heat-shocked hsp:NICD⁺ and hsp:NICD(−) animals were first analyzed histologically by Safranin O staining and Sox9a, NICD and PCNA IHC at 6, 8, 10 and 32 dpr (Fig. 4A,B, Fig. S4A,B, Fig. S5). Although the initial response from the injury margin involved Sox9a⁺ cells, from 8 to 10 dpr these cells exhibited divergent differentiation trajectories. By 10 dpr, there was less Safranin O⁺ cartilage in the hsp:NICD⁺ group, and the percentage of mesenchymal cells retaining Sox9a positivity was 70.8±11.1% in hsp:NICD(−) animals versus 27.0±14.1% in their hsp:NICD⁺ siblings (P<0.001) (Fig. 4C). PCNA IHC also showed a temporal shift in overall proliferation, peaking at 6 dpr rather than 10 dpr (Fig. S5). Furthermore, osteoclast histomorphometry was assessed by tartrate-resistant acid phosphatase (TRAP) staining, and the absence of a phenotype supports an osteoblast lineage-specific phenotype in the hsp:NICD⁺ group (Fig. S4D). Thus, overactivation of NICD drives callus mesenchymal cells to a particularly osteogenic fate, with evidence of increased proliferation early in the injury response followed by accelerated callus ossification.

Proteomic analysis was performed on total callus lysate from 4 dpr heat-shocked hsp:NICD⁺ and hsp:NICD(−) samples to capture changes in protein levels within hours of the final heat shock. In total, 196 proteins were dysregulated by NICD overexpression (Fig. 4D); the three most highly enriched gene ontology (GO) biological process terms were ‘aspartate metabolic process’, ‘dicarboxylic acid metabolic process’ and ‘glucose metabolic process’. Closer review of significantly upregulated proteins revealed a metabolic landscape characteristic of aerobic glycolysis. Described as analogous to the Warburg effect, this involves a shift from oxidative phosphorylation to rapid but inefficient cytoplasmic glycolysis regardless of oxygen tension (Esen and Long, 2014). Dysregulated proteins in the NICD group (Fig. 4D) included a 2.9-fold increase in L-lactate dehydrogenase A chain (Ldha; P<0.001), a 1.8-fold increase in creatine kinase (Ckmb; P<0.001), a 1.5-fold increase in aldehyde dehydrogenase multifunctional fusion protein (Aldh4a1; P<0.001), a 1.7-fold increase in ATP synthase subunit (Atp5l; P=0.007), a 1.5-fold increase in NADH dehydrogenase (ubiquinone) flavoprotein 1, mitochondrial (Ndufv1; P=0.011), a 1.5-fold increase in aspartate aminotransferase (Got1; P=0.012) and a 1.4-fold increase in phosphoglucomutase 1 (Pgm1; P=0.027), among many related changes. We conclude that Notch overexpression results in upregulation of enzymes for aerobic glycolysis, increasing energy availability for accelerated bone regeneration (Lee et al., 2017; Slaninova et al., 2016; Lu et al., 2013; Shyh-Chang et al., 2013). Together with other hits having known functions in bone (Fig. S4C), these proteomic data encourage further investigation of the molecular basis of rapid regenerative ossification.

We posit a nuanced role for Notch signaling in multiple phases of regeneration (Youngstrom and Hankenson, 2019). During mandible regeneration in zebrafish, we report that regenerate bone quantity scales with Notch signaling owing to its positive regulatory roles during both callus proliferation and osteoblast differentiation/anabolism.

Recent work in the field of fracture healing has highlighted remarkable phenotypic complexity of cells involved in bone repair (Matthews et al., 2021; Wong et al., 2020). Our colleagues recently reported that conditional overexpression of NICD using ACTA2-Cre^ERT2 results in accelerated ossification of fracture cartilage in mice (Novak et al., 2020). In addition to validating the common Notch requirement between mice and zebrafish, we expand on this study by showing that initial callus size is regulated by Notch, and also revealing significant, lasting improvements in mandibular bone regeneration from activation of Notch signaling within a narrow postoperative window. Together with our previous work (Paul et al., 2016), we hypothesize an iterative role of Notch-Hedgehog-Notch signaling in osteochondral repair.

Our approach to conditional overexpression of NICD was temporally restricted relative to experiments by Grotek et al. that resulted in lineage arrest of dedifferentiated caudal fin osteoblasts with sustained post-operative activation of hsp:NICD (Grotek et al., 2013). The developmental and regenerative programs of fin rays and the mandible are significantly divergent, and the initial mandible callus is formed by Sox9a⁺ cells. Because zebrafish cartilage cells show not only significant plasticity during regeneration, but also contribute to developmental osteoblasts and adipocytes (Giovannone et al., 2019), our work specifically implicates Notch signaling in both recruitment and ossification of Sox9a⁺ hybrid callus cells. Although we favor a model in which Notch signaling causes accelerated ossification of cartilage, an alternative explanation is that Notch signaling selects for a subpopulation of particularly osteogenic progenitor cells. The stage-specific role of Notch signaling, and its intersection with other pathways that regulate osteochondral repair, will likely be fruitful areas of future investigation.

Zebrafish provide a useful model for understanding efficient mechanisms of skeletal repair. By way of molecular or genetic modulation of post-operative Notch signaling in vivo, we conclude that the extraordinary bone regeneration observed in the zebrafish craniofacial skeleton is contingent on functional Notch signaling. Furthermore, conditional genetic overactivation of this pathway enhances bone accrual by control of proliferation, differentiation and metabolism in osteochondral lineages. Continued investigation of this phenotype could lead to discoveries applicable to mammalian regenerative medicine. Finally, owing to the similarities between mandible and long-bone healing (Wong et al., 2020), relatively high-throughput zebrafish surgical models have potential utility for screening fracture therapeutics in vivo at scale.

Two PCR products forming a 7.4 kb fragment immediately upstream of the her6 transcriptional start site were amplified from the bacmid DKEY-111K7 and sequentially sub-cloned into a Tol2 transgenesis vector (Kawakami, 2004) containing the mCherry coding sequence. This fragment contains four RBPjk/Notch-responsive elements based on the consensus sequence YGTGRGAAM. The following primers were used: SacII-her6-7.4kb-For, ATACCGCGGgcatctgtttgaaaactcccc; NotI-her6-3kb-Rev, ATAGCGGCCGCataagttggcggttcattccg; NotI-her6-3kb-for, ATAGCGGCCGCgaatgaaccgccaacttatcc; SpeI-her6-ATG-rev, AATACTAGTtctctgggaagtatcctttcg (lowercase letters represent target genomic sequence). The final construct was co-injected with mRNA encoding the Tol2 transposase into one-cell-stage embryos. Embryos that showed transient transgene expression were raised and mated to establish founders for this novel transgenic line.

In this study, 385 zebrafish (Danio rerio, laboratory strain AB) with a mean age of approximately 6 months were used with IACUC approval from the University of Connecticut Health Center. Experimental endpoints are reported in dpr. The following transgenic lines were used in this study: Tg(her6:mCherry)^sd64, Tg(hsp70l:Gal4)^kca4 and Tg(UAS:myc-Notch1a-intra)^kca3; the latter two constitute a conditional transcriptional activator and effector pair (Scheer and Campos-Ortega, 1999). For simplification in the text, reporter animals are transcribed as her6:mCherry, and hsp70l:Gal4; UAS:myc-Notch1a-intra animals as hsp:NICD⁺. her6:mCherry animals were genotyped by native fluorescence, and hsp:NICD⁺ and hsp:NICD(−) animals were genotyped by tail biopsy using published primers (Barske et al., 2016).

her6:mCherry zebrafish were used to perform initial histological assessment of regeneration, and for experiments involving histology of DBZ-treated animals hsp:NICD⁺ and hsp:NICD(−) zebrafish were used for conditional genetic overactivation of Notch signaling. Wild-type zebrafish on the same AB9 background were used for all other experiments. Animals of both sexes were used indiscriminately, and all transgenic animals had grossly normal phenotypes. Fish were housed in a custom facility with automated environmental controls (Aquaneering, San Diego, CA, USA). Euthanasia was performed by submersion in 400 mg/l tricaine methanesulfonate (MS-222; Syndel) in system water, which was then moved over ice.

DBZ (CAS 209984-56-5; Syncom) is a γ-secretase inhibitor used to pharmacologically inhibit cleavage of the active intracellular domain of Notch receptors. To treat fish with 20 µM DBZ, 400 µl of 50 mM DBZ stock in DMSO (BP231; Fisher BioReagents) was added per liter of system water (final DMSO concentration 0.1%). DMSO without DBZ was used as a vehicle negative control. Initial DBZ treatments were applied in three 16-h intervals, 48 h apart, beginning 4 h after mandibulectomy. Follow-up DBZ treatment windows were single 48-h exposures at a range of times between 0 and 5 dpr.

To overactivate Notch signaling, hsp:NICD⁺ and hsp:NICD(−) animals were treated with heat, causing transient systemic overexpression of the (activated) NICD by heat shock-mediated Gal4/UAS transgenics. Controls were hsp⁺ animals lacking the NICD transgene and heat-treated in parallel. Animals were transferred to 38°C system water for three 30-min intervals, 48 h apart, beginning 2 h prior to mandibulectomy.

A cohort of DMSO±DBZ fish were immersed in system water containing 4.5 mg/ml BrdU (5002, Sigma-Aldrich) for 1 h and washed with a continuous flow of system water for 30 min prior to euthanasia.

Samples were fixed in 10% neutral buffered formalin and processed for histology. Fish were decapitated, decalcified in 20% EDTA, pH 7.4, dehydrated in stepped solutions of ethanol/xylenes, and embedded in paraffin. Blocks were sectioned at 5 µm using a microtome, and deparaffinized prior to staining. All histological images are representative of at least n=5 biological replicates per group. Approximate surgical margins are marked with yellow arrows.

For cartilage Safranin O histology, nuclei were stained using modified Weigert's iron hematoxylin differentiated in acid-ethanol, 0.02% Fast Green was used as a counterstain, and 1% Safranin O was used to stain the proteoglycan-rich extracellular matrix of cartilage. Slides were mounted in toluene media. Representative sections were imaged using an Axio Scan.Z1 slide scanner (Zeiss Microscopy). Dotted squares in Safranin O images represent approximate source locations of IHC insets from adjacent slides.

For IHC (indirect immunofluorescence), slides were subjected to membrane permeabilization in acetone and antigen retrieval in citrate, pH 6.0, or commercial reagent (550524; BD Biosciences) in a pressure cooker for 3 min. Two percent goat serum with 1% DMSO and phosphate-buffered saline with 0.05% Tween-20 (BP337; Fisher BioReagents) or commercial buffer (37580, Thermo Fisher Scientific) was used for blocking. Novus NBP2-25158 chicken anti-mCherry antibody (Centennial) was used as a primary antibody for her6:mCherry fish at 1:250 and incubated at 4°C overnight. Abcam AB175477 goat anti-chicken Alexa Fluor 568 was used as a secondary antibody at 1:500 at 4°C for 1 h. Additional antibodies used were: GeneTex GTX48504 goat anti-Notch1a ICD (recognizing amino acids 2314-2327, QNMAPMQHQNISHH, near the C-terminus; NP_571516.1), Sigma-Aldrich MoBU-1 mouse anti-BrdU, BD Biosciences 559565 rabbit anti-caspase-3, GeneTex GTX128370 rabbit anti-Sox9a, Invitrogen 133900 mouse anti-PCNA, and Invitrogen secondary antibodies A32849, A11011, A11001, A11008 and A32787. Slides were mounted with DAPI glycerol-based media (P36966, Molecular Probes). Images were acquired on a Leica DMI6000B fluorescence microscope. Antibody lot numbers are reported in Table S2. Quantification was performed by single-channel color thresholding and particle counting in ImageJ, manually confirmed. Numbers are reported per 0.144 mm² field. her6:mCherry(−) animals and IHC slides lacking primary antibodies were negative controls.

For fluorescence ISH, col1a1a and col2a1a transcripts were detected using a previously published protocol (Paul, et al., 2016). Briefly, pCR-Blunt II-TOPO plasmids with kanamycin resistance-containing blunt probe insertions were propagated in DH5α cells (18265017, Invitrogen). Plasmid DNA was extracted, linearized and purified (Thermo Fisher Scientific K0502, K0691). DIG/DNP-labeled riboprobes (PerkinElmer NEL555001EA; Roche 11277073910,) were transcribed, isolated and sequentially hybridized. Fluorescence was visualized following tyramide signal amplification (Perkin Elmer NEL747A001KT, NEF832001EA, NEL704A001KT, NEL705A001KT). Two fields per sample, one anterior and one posterior, were segmented for callus mesenchyme and counted in ImageJ.

A commercial TRAP kit (387A, Sigma-Aldrich) was used to stain osteoclasts. Samples were coverslipped using aqueous media (VectaMount AQ) and imaged within 24 h. Histomorphometry was performed in Osteomeasure7 (OsteoMetrics) using an Olympus BX53 with Wacom pen computer. A single mid-callus field was captured and analyzed for each specimen using a 20× objective. Parameters are reported per 0.26 mm² field.

Immediately following euthanasia at 10 dpr, the injury site (callus), and for some experiments, matched contralateral mandible tissue, was microdissected under a Leica M165FC stereomicroscope; n=4 replicate pools per group of n=11-13 fish each were prepared in 1 ml of TRIzol (Thermo Fisher Scientific) at 4°C, and homogenized in a bead mill (Minilys, Bertin Technologies) prior to storage at −80°C. Total RNA was isolated by standard AGPC extraction and purified by solid phase extraction (7326830, Bio-Rad Laboratories) with on-column DNase digestion. RNA was reverse-transcribed (RT; Bio-Rad, 1708840), and 80 ng of cDNA per reaction was run in at least duplicate in a Bio-Rad CFX96 Real-Time System using SYBR Green chemistry (Bio-Rad, 1725270). (−)RT sample templates served as negative controls.

Primer pairs (synthesized by Integrated DNA Technologies) were either identified from the literature (Casadei et al., 2011; Lin et al., 2013; Wiweger et al., 2014) or custom-designed in NCBI Primer-BLAST to span exons (Table S1). Primers were validated by melt curve analysis and gel electrophoresis of qPCR products. Expression levels are reported as 2^−dCt with respect to the housekeeping gene actb2, which allows reference-free interpretation (Youngstrom et al., 2020). Raw dCt values were used for statistical analysis.

Samples were imaged on a Scanco µCT40 specimen scanner at 12 µm³ resolution at the following settings: 55 k Vp, 145 µA, 300 ms integration time, 500 projections/180°. Bone within regenerating mandibular defects was quantified by slice-by-slice manual segmentation of all new bone within and adjacent to the injury margins, which were detected by radiodensity and with respect to anatomical landmarks. The first precaudal/thoracic (T1) vertebra of each fish was also analyzed: the hourglass-shaped vertebral bodies were manually segmented, slice-by-slice, from all processes (Suniaga et al., 2018). BV, BS, TMC and TMD are reported from these defined volumes of interest, with a lower threshold of 450 mgHA/ccm. All µCT data represent a minimum of n=8 biological replicates per group.

Total protein was isolated from hsp:NICD⁺ and hsp:NICD(−) animals (n=5 per group) at 4 dpr. Microdissected callus tissue was lysed with commercial protein extraction reagent (M-PER 78501, Thermo Fisher Scientific) in the presence of protease inhibitors, ultrasonicated, and quantified by BCA assay (Pierce Rapid Gold, Thermo Fisher Scientific). 200 µg aliquots were analyzed by an UltiMate 3000 RSLCnano ultra-high performance liquid chromatography system coupled to a Q Exactive HF tandem mass spectrometer (UPLC-MS/MS; Thermo Fisher Scientific). Raw data were searched against the Danio rerio UniProt reference proteome (UP000000437) using MaxQuant v1.6.10.43 (Max Planck Institute of Biochemistry). Data visualization and statistical analysis was performed in Scaffold Q+S v5.0.0 (Proteome Software). GO terms were generated using an unranked list of enriched (P<0.05) targets as foreground against the entire identified protein list as background via the Gene Ontology enRIchment anaLysis and visuaLizAtion tool (GOrilla) (Eden et al., 2009). Full experimental details are included in supplementary Materials and Methods, and raw data are available via ProteomeXchange with identifier PXD027280.

Prism 9 (GraphPad Software) was used for statistical analysis and graph production. Wild-type 10 dpr qPCR comparisons between matched resected and contralateral mandibles were analyzed using ratio-paired, two-tailed t-tests of raw dCt values. All µCT, qPCR, histologic and proteomic data were analyzed using unpaired, two-tailed t-tests with Welch's correction. Each experimental group (DBZ or hsp:NICD+) was only compared with its associated control. Statistical significance for all experiments was set at P≤0.05, graphically depicted by asterisks within figures.

Megan Matsutani, Claire E. Arata, Sanja Novak and Emily A. Gardner provided valuable logistical and technical assistance. We also thank Archana Sanjay, Ivo Kalajzic and Parker K. Acevedo for helpful discussions.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199995.