Transgenic medaka were maintained in the fish facility at the Department of Biological Sciences, National University of Singapore (DBS, NUS). All breeding and experimental procedures were conducted under protocols approved by the Institute of Animal Care and Use Committee of the National University of Singapore (IACUC; protocol numbers: BR19-0120, R18-0562).

Purification of medaka bone cells by fluorescence activated cell sorting (FACS) and RNA sequencing (RNAseq) was described previously (Buettner et al., 2018; Phan et al., 2020b). Briefly, rankl:HS:cfp transgenic larvae with col10a1:GFP, osx:mCherry, or ctsk:GFP reporter backgrounds were heat-shocked at 9 days post fertilization (dpf) at 39°C for 2 h to collect col10a1-positive osteoblast progenitors and osx-positive osteoblasts at 10 dpf, and ctsk-positive osteoclasts at 12 dpf. Transgenic reporter larvae without Rankl transgene were used as control. mmp13b ^+/+ and mmp13b ^−/− larvae with rankl:HS:cfp/ctsk:mCherry transgenic background were heat-shocked at 9 dpf at 39°C for 2 h and used to collect mCherry-positive osteoclasts at 3 days post heat-shock (dphs). To collect GFP-positive osteoblast progenitors from caudal fins, 2–3 month-old col10a1:GFP transgenic fish were used. Tissue dissociation was performed using a Papain Dissociation System (Worthington Biochemical Corporation) following the manufacturer’s instructions. FACS was conducted using a BD FACS Aria II 5 lasers cytometer. RNA was isolated from FACS purified cells using the Purelink™ RNA micro kit (Invitrogen). RNA quality was measured with an Agilent Bioanalyzer 2100 and Agilent RNA 6000 Pico Kit (Agilent Technologies, United States), and samples with a RIN value > 8.0 were sent to Novogene for RNA sequencing and bioinformatics analysis. The sequencing results were deposited to NCBI’s BioProject, under accession number PRJNA767186.

RNA was isolated from FACS purified cells and whole mount fins at 2 days post amputation (dpa) using a Purelink™ RNA micro kit (Invitrogen) and cDNA was synthesized and amplified using a PreAmp and Reverse Transcription Master Mix Kit (Fluidigm). qPCR was performed with PowerUp SYBR Green Master Mix (Applied Biosystems). All primers used for qPCR are listed in Supplementary Table S3. β-actin was used for normalization. All experiments were performed with three biological samples and three technical repeats each. Samples for RNAseq analysis and qPCR validation were obtained from independent experiments. Data were analyzed using the Bio-Rad CFX Maestro 1.0 software and Student’s t test was performed for statistical analysis.

Riboprobes spanning 659 nucleotides of the mmp13b cDNA (nt458-nt1116; ENSORLG00000014453; http://www.ensembl.org/) were synthesized using a DIG RNA labeling kit (Roche, CAT# 11175025910). The synthesis of a col10a1 riboprobe was described previously (Renn et al., 2013). RNA in situ hybridization was conducted on cryosections as described previously (McMillan et al., 2018) with small modifications. The NBT/BCIP (Sigma) staining solution was added onto the sections and stained at room temperature. Then, the sections were washed with PBS-Tween 20 (PBST) and mounted in Mowiol 4-88 (Calbiochem) for imaging.

Guide RNAs (gRNAs) targeting mmp9 (ENSORLG00000006340) and mmp13b (ENSORLG00000014453) were designed using the CRISPR/Cas9 target online prediction tool (CCTop) (Stemmer et al., 2015) and synthesized by Integrated DNA Technologies (IDT Singapore; sequences in Supplementary Table S3). Three gRNAs (36 ng/μL each) were co-injected into one-cell stage medaka embryos together with tracrRNA (67 ng/μL, IDT) and Cas9 nuclease (0.25 μg/μL, IDT). Primers used for genotyping (mmp9_FP/RP, mmp13b_FP/RP) are listed in Supplementary Table S3.

The selective MMP13 inhibitor CL-82198 hydrochloride (TOCRIS, CAT# 2632) was dissolved in H₂O to make a 50 mM stock solution. The stock solution was stored at room temperature and working solutions were prepared freshly for every use. Pre-treatment was performed on rankl:HS:cfp transgenic larvae from 7–9 dpf with 300 μM CL-82198 (12 μl stock in 2 ml fish medium). Larvae were heat-shocked at 9 dpf to induce Rankl expression, followed by consecutive treatment from 9–12 dpf with 300 μM CL-82198. Control fish were treated with 12 μl H₂O in 2 ml fish medium from 7–12 dpf. All larvae were kept at 30°C in an incubator throughout the treatment period.

12 dpf larvae were transferred to 0.004% calcein staining solution for 1 hour at 30°C in the dark. Then, the larvae were washed with fish medium before mounting in 1.2% low melting agarose for imaging. Alizarin Red bone staining of 12 dpf larvae and 17 dpa fins was performed as previously described (Renn and Winkler, 2009).

Fluorescent images of whole embryos were used for the quantification of osteoclast numbers following the method described previously (Phan et al., 2020a). For quantification of bone lesion sizes, four consecutive vertebral bodies were used for each analysis. The lesion percentage was determined as 100%*[0.5* (lesion area/total area) + 0.5* (number of absorbed arches/4)]. For statistical analysis, a Student’s t test (two-tailed, unpaired) was performed using GraphPad Prism 6.0 (GraphPad Software, La Jolla, CA, United States).

Collagen staining was performed on 20 μm cryosections using a Picro Sirius Red Stain Kit (ab150681; Abcam, UnitedStates) following the manufacturer’s instructions. Cryosection slides were defrosted at room temperature for 5 min before rehydration in water. Sections were incubated with Picro sirius red staining solution for 1 h and thereafter rinsed with acetic acid before dehydration with absolute ethanol and mounting in synthetic resin for imaging.

Immunostaining on 20 μm cryosections was performed as described previously (Phan et al., 2020a). The following primary antibodies were used: Rabbit anti-collagen 1 (ab23730, Abcam), mouse anti-collagen 2 (II-II6B3; Developmental Studies Hybrydoma Bank, United States), rabbit anti-fibronectin (F3648, Sigma, United States), rabbit anti-Tenascin (T2550-23, United States Biological, United States). As secondary antibodies, goat anti-rabbit Alexa Fluor 633 (Thermofisher Scientific) and goat anti-mouse Alexa Fluor 633 (Thermofisher Scientific) were used. The numbers of analyzed sections is listed in Supplementary Table S4.

TRAP staining on 20 μm cryosections was performed as described previously (Ethiraj et al., 2021). This included defrosting slides at room temperature, washing with PBS and incubating the slides with TRAP staining solution for 2 h. After this, slides were washed with PBS and mounted in Mowiol 4-88 for imaging.

Caudal fins were collected from 3 month-old adult medaka with mpeg1:mCherry transgenic background at 2 dpa and cryosectioned to obtain 20 μm longitudinal sections. Sections were stained with TRAP staining reagent for 6 h at room temperature. The same fin sections were imaged before and after TRAP staining.

Collagen in situ zymography was conducted as described before but with slight modifications (Bai et al., 2005; Li et al., 2020). Rankl-transgenic larvae were heat-shocked at 9 dpf and fixed at 12 dpf with 4% PFA at 4°C overnight. 20 μm transverse sections from the trunk were obtained by cryosectioning and dried at room temperature for 2 h, followed by PBST washing and counterstaining with DAPI. DQ collagen substrate conjugated with FITC (ThermoFisher Scientific; D12060) was utilized as substrate for in situ zymography. DQ collagen was dissolved in milliQ water to obtain a 1 mg/ml stock solution. 1% low-melting agarose (LMA) was dissolved in substrate buffer (50 mM Tris-HCl, 5.0 mM CaCl₂, 5.0 μM ZnSO₄; Ph 7.5). 200 μl gel containing 40 μl collagen stock solution, 60 μl substrate buffer, and 100 μl 1% LMA was coated on sections and sections were incubated at 37°C for 18 h. After incubation, sections were washed with PBST and mounted in Mowiol 4-88 for imaging. For negative control, slides were incubated at −20°C for 18 h.

For live fluorescence imaging, larvae were anesthetized with 0.016% Tricaine (MS222, Sigma), mounted in 1.2% low melting agarose on a glass-bottom Petri dish, and imaged using an Olympus FluoView FV3000 confocal microscope and a Nikon SMZ18 stereomicroscope equipped with the NIS-Elements BR 3.0 software. Olympus FluoView FV3000 confocal microscope was also used for time-lapse imaging and imaging of RNA in situ hybridization sections. The Nikon SMZ18 stereomicroscope was used to image Alizarin Red stained larvae and fins. To image Picro sirius red stained and TRAP-stained sections as well as RNA in situ hybridization sections, a Nikon Eclipse 90i upright microscope equipped with NIS-Elements BR 3.0 software was used. Immunostained sections were imaged with a LSM 900 Airyscan II super-resolution confocal laser scanning microscope. Images and time-lapse movies were processed with Bitplane Imaris and Fiji.

To identify genes implicated in bone remodeling, we previously performed transcriptome profiling of bone cells in a medaka osteoporosis model where transgenic Rankl expression is induced by heat shock (Phan et al., 2020b; To et al., 2012). Differentially expressed genes (DEGs) were identified in col10a1 positive osteoblast progenitors and osterix (osx) positive osteoblasts at 1 day post heat-shock (1 dphs; i.e., 10 days post fertilization, dpf) and in cathepsin K (ctsk) positive osteoclasts at 3 dphs (Figure 1A) (Phan et al., 2020b). In the present study, we performed RNA sequencing of col10a1 cells isolated from regenerating caudal fins. In the medaka fin, col10a1 positive osteoblast progenitors are located along segments of the mineralized hemirays and in joints linking the segments. After amputation, col10a1 cells migrate to the amputation site and contribute to the regenerating fin rays (Dasyani et al., 2019; illustrated in Figure 1B). Here, we analyzed the transcriptomes of col10a1 cells purified from fins at 0- and 2-days post amputation (dpa; Supplementary Figure S1). 14,377 transcripts were detected in col10a1 cells (Supplementary Figure S1A), and among them, 3591 were differentially regulated, with 1855 significantly up-regulated at 2 dpa (Supplementary Figure S1B). A KEGG analysis revealed enrichment of, e.g., cell cycle genes, indicative for enhanced cell division of osteoblast progenitors during regeneration (Dasyani et al., 2019) as well as ECM-receptor interactions, among others (Supplementary Figure S1C).

Next, to identify genes commonly regulated during bone remodeling under osteoporosis-like lesions and in bone regeneration, we compared the regeneration dataset with the previously obtained osteoporosis dataset (Phan et al., 2020b). This revealed nine genes that were commonly upregulated in col10a1 cells (Figure 1C; Table 1). One of them was mmp13b, which showed high expression levels and a significant fold change in both datasets. Quantitative RT-PCR (qPCR) analysis of FACS purified col10a1 cells validated mmp13b up-regulation consistent with the idea that this metalloprotease is implicated in both bone resorption and repair. We also noticed mmp13b up-regulation in Rankl-induced (Rankl+) ctsk osteoclasts at 12 dpf. However, the baseMean expression levels were considerably lower than in osteoblasts (Figure 1D). To confirm expression of mmp13b in osteoblast progenitors, RNA in situ hybridization of wildtype embryos at 12 dpf showed mmp13b expression along neural arches (Figure 1E, magenta arrow) and the outer surface of the notochord (Figure 1E, cyan arrow), where col10a1 cells reside. Together, this suggests that mmp13b is expressed in osteoblast progenitors, and that its expression is significantly upregulated under Rankl+ and regeneration conditions.

We hypothesized that osteoblast-derived Mmp13b contributes to bone resorption in Rankl embryos and generated knock-out mutants using CRISPR/Cas9 to further characterize its function in medaka (Supplementary Figure S2). Homozygous mmp13b mutants were viable and showed no obvious phenotype during embryogenesis. After Rankl induction, ectopic osteoclasts formed normally at 2 dphs (11 dpf) with similar numbers in mmp13b ^+/+ siblings and mmp13b ^−/− mutants (Figures 2A,B). Next, we assessed the extent of bone resorption after Rankl induction by calcein staining. In mmp13b ^+/+ embryos, excessive bone resorption was evident, with lesions in the centra (Figure 2C; white arrows) and completely resorbed neural arches (yellow arrowheads). In contrast, mmp13b ^−/− mutants exhibited significantly less bone resorption with a 7.5% reduction in bone matrix lesions on average (Figure 2D). This suggests that a mmp13b deficiency does not affect formation of osteoclasts, but that Mmp13b is necessary for osteoclast activity.

To confirm this, we used the MMP13 inhibitor CL-82198 hydrochloride in medaka and examined bone resorption after Rankl induction. Like mmp13b mutants, CL-82198 treated embryos exhibited reduced bone resorption, when compared to controls, with less lesions in the vertebral centra (Figure 2E; white arrows) and mostly intact neural arches (yellow arrowheads). Quantifications revealed a significant reduction of the area with bone lesions in CL-82198 treated larvae compared to controls (Figure 2F), which was consistent with the phenotype in mmp13b knock-out mutants. Together, both knock-out of mmp13b and chemical inhibition with a MMP13 inhibitor in medaka led to impaired bone resorption after Rankl induction, suggesting a possible non-cell autonomous role in controlling osteoclast activity.

In medaka, blocking Mmp13b activity reduced Rankl-induced resorption of mineralized bone matrix. To investigate whether the degradation of non-mineralized collagen was also affected, a Picro-Sirius Red staining was conducted in mmp13b ^−/− mutants at 12 dpf to detect distribution of collagen fibres. During normal bone development (without Rankl induction, Rankl-), collagen staining was observed along the neural and hemal arches (Figure 3A; black arrowheads), and the outer surface of the notochord (blue arrowhead) in both wildtype siblings and homozygous mutants. In contrast, after Rankl induction, mmp13b ^+/+ but not mmp13b ^−/− embryos showed strongly reduced collagen staining (Figure 3A; magenta arrowhead and circle), suggesting robust collagen degradation in wildtype siblings but not in mmp13b ^−/− mutants. A quantification revealed that more than half of the analyzed transverse sections (8/14 = 57%) showed reduced collagen staining for wildtype embryos, while a similar reduction was observed in only 23% of sections (6/26) from mmp13b ^−/− mutants (Figure 3B).

Collagen 1 is the major component of non-mineralized collagen fibres (Blair et al., 2016). Therefore, we performed Col1 immunostaining on transverse trunk sections of wildtype control siblings and mmp13b ^−/− embryos at 12 dpf (Figure 3C). Under non-osteoporotic Rankl- conditions, Col1 staining was observed along the outer surface of notochord and notochordal sheath in both wildtype siblings and mmp13b ^−/− mutants. In Rankl + embryos, on the other hand, Col1 was degraded and re-distributed to the distal side of osteoclasts in wildtype siblings (Figure 3C; green arrowheads), which is consistent with the process of transcytosis as described earlier (Salo et al., 1997; Ng et al., 2019). In mmp13b ^−/− embryos, however, redistributed Col1 fragments were significantly fewer in numbers when compared to wildtype siblings, indicating impaired degradation and transcytosis of Col1 in mmp13b mutants. Together, this suggests that Mmp13b is required for osteoclast resorption, including the degradation and transcytosis of non-mineralized collagens and resorption of mineralized bone matrix.

Our findings suggested that osteoclasts form normally but show impaired resorption activities in mmp13b mutants. To understand the causes of reduced activity, we analyzed the maturation and migratory dynamics of osteoclasts in mmp13b mutants. First, ctsk-positive osteoclasts were isolated by FACS from mmp13b ^+/+ and mmp13b ^−/− embryos at 12 dpf, and qPCR was conducted to examine transcription levels of several osteoclast differentiation genes, which we had earlier identified in medaka as upregulated under Rankl+ conditions (Phan et al., 2020a). These included mmp9, rank, ctsk, traf6, TRAP, and il10, as well as ECM-related genes, such as col6a1, col6a3, itga2.2, and itgb8 (Figure 4A). Compared to wildtype siblings, mmp9, rank, ctsk, and traf6 were significantly down-regulated in osteoclasts of mmp13b ^−/− mutants. Furthermore, il10, which is down-regulated in differentiated osteoclasts, had significantly increased transcription levels in mmp13b ^−/- osteoclasts. Together, this suggests that osteoclasts had retained an immature state and failed to fully differentiate in mmp13b mutants. Also, col6a1, col6a3, itga2.2, and itgb8 were significantly down-regulated in osteoclasts of mmp13b ^−/− mutants, opening the possibility that bone cell-ECM interactions are impaired in mutant embryos. Surprisingly, the only osteoclast marker that we found up-regulated in mmp13b mutants, was TRAP. Consistently, also enzymatic TRAP activity could be detected in ctsk-positive cells by histochemical staining in transverse sections of mmp13b ^−/− embryos, while their distribution appeared affected (Supplementary Figure S3).

Next, to assess morphology and migratory dynamics of osteoclasts during bone resorption, live time-lapse imaging was conducted in wildtype siblings (Supplementary Movie S1) and mmp13b ^−/− mutants (Supplementary Movie S2) with a rankl:HS:cfp/ctsk:mCherry background starting from 48 h post heat-shock (hphs) at 11 dpf. Based on these movies, we determined sphericity, area, and migratory speed of osteoclasts. Compared to wildtype, osteoclasts in mmp13b ^−/− mutants generally exhibited higher sphericity (Figure 4B). At around 55 hphs (7 h after time-lapse imaging started), we observed that ctsk-positive cells generally exhibit a range of varying cell sizes and surface areas, and that this range was altered in mutants. For wildtype, most ctsk-positive osteoclasts could be categorized into three major clusters defined by distinct surface areas. Cells could be categorized into cluster 1 (with a surface area of 400–500 μm²), cluster 2 (1,000–2,000 μm²), and cluster 3 (>10,000 μm²) (Figure 4C). Osteoclasts in cluster 1 appeared as small and round cells, cells in cluster 2 had an elongated morphology, and cells in cluster 3 were characterized by large extended surface areas (Figure 4D). In stark contrast, most ctsk-positive cells in mmp13b mutants displayed a shift towards smaller area ranges, which we defined as cluster 4 (300–400 μm²) (Figure 4C), with rounder and smaller appearance (Figure 4D). Comparing osteoclasts in mmp13b ^−/− mutants with that in wildtype, we noticed a significant increase in the percentage of mutant cells in the 300–400 μm² range and a significant decrease in the >10,000 μm² range (Figure 4C). Finally, also the speed of dynamically moving osteoclasts was affected in mutants. In the first 1.5 h of time-lapse analysis (around 48–49.5 hphs), osteoclasts from wildtype siblings (N = 6) moved significantly faster than that from homozygous mutants (N = 5) (Figure 4E; and Supplementary Movies S1, S2). After 1.5 h, when the first highly active osteoclasts start to slow down to attach to bone matrix, there was no significant difference in the speed between wildtype and mutant osteoclasts. In summary, osteoclasts in mmp13b mutants exhibited signs of incomplete maturation upon Rankl induction when compared to wildtype siblings, suggesting that osteoblast-derived Mmp13b promotes osteoclast maturation and initiates osteoclast dynamics during bone remodeling.

Transcript levels for mmp13b were elevated upon Rankl induction as shown by RNAseq analysis and qPCR. To test whether upregulated RNA levels also translated into increased enzymatic metalloprotease (MMP) activity, we performed in situ zymography using a DQ-collagen substrate as described previously (Li et al., 2020). Control cryosections incubated at −20°C showed no or very little background enzymatic activity (Figure 5A). In contrast, strong enzymatic activity was observed in mmp13b ^+/+ embryos after Rankl induction around muscle fibres, neural progenitors, chordoblasts inside the notochord and cells lining the notochord on the outside (Figure 5B). In mmp13b ^−/− mutants, on the other hand, enzymatic activity in chordoblasts and notochord-lining cells was markedly decreased, while levels in muscle fibres and neural progenitors remained high (Figure 5C). This suggests that enzymatic metalloprotease activity observed in cells of the vertebral column is mostly due to Mmp13b, which is consistent with high mmp13b transcript levels observed, e.g., by in situ hybridization (compare Figure 1E). Rankl-induction increased MMP activity in these cells in mmp13b ^+/+ siblings but not mmp13b ^−/− mutants (Figures 5D–F). Next, we determined the identity of cells around the notochord that showed elevated enzymatic activity. For this, we performed analysis in transgenic lines expressing mCherry in osx-positive osteoblasts or ctsk-positive osteoclasts. This showed that notochord-lining cells with the highest enzymatic activity were osteoblasts (Figures 5G,G’), while osteoclasts had low levels of enzymatic MMP activity and were located one or 2 cell diameters away or directly adjacent to cells with high MMP activity (case 2 and 3, respectively; yellow arrowheads, Figures 5H,I).

Together, this strongly suggests that Rankl induction leads to elevated Mmp13b activity primarily in osteoblasts and that it is osteoblast-derived Mmp13b that promotes the maturation of adjacent osteoclasts and triggers their dynamic behavior required for bone remodeling.

RNAseq analysis suggested that among other mmp genes, mmp13b was strongly upregulated in col10a1 expressing osteoblast progenitors in the regenerating caudal fin at 2 dpa (Figure 6A; Supplementary Table S2). RNA in situ hybridization was used to confirm that mmp13b was expressed in osteoblast progenitors along the hemiray segments (Figures 6B,C). We noticed that mmp13b is also expressed in joint cells of the regenerating proximal fins (asterisks in Figure 6B). To assess whether mmp13b is required for fin regeneration, the fins of 2–3-month-old mutant fish were amputated and the length of regenerates were measured at 17 dpa. Compared with wildtype siblings, the regenerate lengths were significantly shorter in homozygous mutants, however the bony fin rays were mineralized normally (Figures 6D,E). Taken together, mmp13b is expressed in osteoblast progenitors of fin ray segments and joints, and its knock-out resulted in delayed fin regeneration in medaka.

We next assessed whether a deficiency in mmp13b alters the ECM structure in fins and thereby affects the efficiency of regeneration. We analyzed matrix proteins that are known substrates of Mmp13b, such as collagen 1 (Hayami et al., 2008), collagen 2 (Howes et al., 2014), fibronectin (Zhang et al., 2012), and tenascin-C (Yamamoto et al., 2016), in fin rays of mmp13b ^+/+ and mmp13b ^−/− fish at 0 and 2 dpa. In control fins at 0 dpa, no presence of collagen 1 and collagen 2 was detected in joint regions (Figure 7A; empty arrowheads). In contrast, collagen 1 and collagen 2 were strongly expressed in joints of mmp13b ^−/− fins (Figure 7A; white arrowheads). This suggests that the collagen composition is altered in the joint regions of mmp13b mutants. Also, at 2 dpa, excess collagen 1 and collagen 2 was detectable in homozygous but not wildtype joints. In addition, also fibronectin showed enhanced expression in the bone matrix region in mutant fins (Figure 7A; yellow arrows) at 2 dpa when compared to wildtype (yellow empty arrows). For tenascin-C, in the fibroblast-containing region in the center of the fin ray (yellow box), its distribution appeared to be more aggregated in mmp13b ^−/− fins compared to mmp13b ^+/+ fins. Taken together, these findings suggest that a mmp13b deficiency results in an altered ECM composition in fin rays. This includes an excess of matrix proteins especially in the joints, which harbor progenitor cells needed for regeneration (Ando et al., 2017; Dasyani et al., 2019). To examine whether an altered ECM structure in the joints of mmp13b ^−/− deficient fins affected the number of osteoblast progenitors, we analyzed expression of the progenitor marker col10a1 by in situ hybridization in fins at 6 dpa. This showed that, as reported earlier (Dasyani et al., 2019), col10a1 is expressed in joints of wildtype fins (Figure 7B, blue arrow). In contrast, in mmp13b ^−/− fins, col10a1 expression in joints was strongly reduced (Figure 7B; blue box, blue arrows). This suggests that the number of col10a1-positive osteoblast progenitors is reduced in the joints of mmp13b deficient mutants.

Fin regeneration depends on the recruitment of mpeg1-positive macrophages to the amputation site (Petrie et al., 2014). These macrophages also express tartrate resistant acidic phosphatase (TRAP) and cathepsin K (ctsk) (Blum and Begemann, 2015). qPCR analysis showed that transcript levels of both trap and ctsk were significantly downregulated in mmp13b ^−/− fins at 2 dpa (Figure 7C). Consistent with this, also histochemical TRAP staining revealed a reduction or absence of TRAP-positive cells in the regeneration blastema of mmp13b ^−/− fins, when compared to wildtype (Figure 7D; blue arrowheads). We also analyzed distribution of mpeg1:mCherry/TRAP double-positive cells in regenerating fins at 2 dpa (Supplementary Figure S5). This revealed the presence of both mpeg1/TRAP double-positive cells as well as TRAP-positive/mpeg1-negative cells. We conclude that the latter are osteoclasts that have differentiated from macrophages and that therefore at least a subset of TRAP-positive cells are osteoclasts. Together, this suggests that a deficiency in Mmp13b results in a modified ECM composition in regenerating fins with a concomitant reduction in osteoblast progenitor and macrophage/osteoclast numbers, consequently leading to a delay in fin regeneration.

Medaka mmp13b ^−/− mutants develop normally into adulthood with no obvious deficiencies in bone remodeling. This suggests that other Mmps compensate for the loss of Mmp13b, similar to the situation in mice. There, Mmp9 ^−/−;Mmp13 ^−/− double mutants exhibited more severe defects in growth plate formation and bone development compared to the respective single mutants (Stickens et al., 2004). Interestingly, our RNAseq data showed strong expression and up-regulation of mmp9 in osteoblast progenitors and osteoblasts after Rankl induction (Supplementary Table S1). As this opened the possibility of a synergism with mmp13b, we generated mmp9 ^−/−;mmp13b ^−/− double mutants in medaka to identify potential compensatory effects on endogenous bone remodeling. At embryonic stages, mmp9 ^−/−;mmp13b ^−/− double mutants developed normally. However, severe bone defects were observed in adult mutants at 5 months. Alizarin red staining revealed normally mineralized bone in wildtype controls and mmp9 ^−/− as well as mmp13b ^−/− single mutants. In contrast, mmp9 ^−/−;mmp13b ^−/− double mutants showed an increase of mineralized bone mass in the arches of vertebral bodies (Figure 8A). Transverse views of single vertebral bodies revealed excess bone matrix in the neural arches of mmp9 ^−/−;mmp13b ^−/− mutants, reducing the size of the canal that contains the spinal cord (Figure 8A; black arrow). Also, hemal arches had increased bone mass resulting in an obstruction of the hemal canal and dislocation of blood vessels (Figure 8B; black arrows). This phenotype was likely caused by reduced bone resorbing activity of the normally formed osteoclasts, as evident upon Rankl induction in mmp9 ^−/−;mmp13b ^−/− double mutants (Figures 8C,D; Supplementary Figure S4). Together, this suggests that mmp9 and mmp13b work redundantly and are required for activity of endogenous osteoclasts as well as under osteoporotic conditions. Interestingly, mmp9 ^−/−;mmp13b ^−/− double mutants also showed delays in fin regeneration (Supplementary Figure S5).

Transcriptome profiling of bone cells under osteoporotic conditions revealed that two mmp13 orthologs, mmp13 (ENSORLG00000020553) and mmp13b (ENSORLG00000014453) are expressed in medaka. The zebrafish genome also has two mmp13 genes, mmp13a (ENSDARG00000012395) and mmp13b (ENSDARG00000100794). Earlier studies have focused on mmp13a, as mmp13b is expressed at low levels and transcripts cannot be detected by qPCR (Zhang et al., 2017) or in situ hybridization (Li et al., 2020). Zhang et al. (2017) reported that mmp13a was up-regulated in atp6v1h-deficient zebrafish that have an osteoporosis-like phenotype suggesting a potential role in bone resorption.

In medaka, on the other hand, mmp13a (annotated as mmp13 in ENSEMBL, see above) showed low expression levels in osteoblasts and was not differentially regulated upon Rankl induction (Supplementary Table S1). Therefore, we focused our study on mmp13b and investigated its function in bone formation and remodeling. RNA in situ hybridization confirmed that mmp13b is expressed in osteoblast progenitors and osteoblasts. In addition, in situ zymography confirmed strongly elevated MMP enzymatic activity in osteoblasts upon Rankl induction. The MMP activity was strongly reduced in mmp13b mutants. Together, this suggests that osteoblast progenitors and osteoblasts produce enzymatically active Mmp13b.

Its expression in osteoblasts could reflect a possible role for Mmp13b in bone formation. Such a role was previously shown in mice where Mmp13 is expressed in hypertrophic chondrocytes and osteoblasts and contributes to endochondral bone ossification and fracture healing (Yamagiwa et al., 1999; Stickens et al., 2004). In Mmp13 ^−/− mice, hypertrophic chondrocytes exhibited a delayed exit from the growth plate and accumulated in a terminally differentiated state, leading to an expanded hypertrophic chondrocyte zone (Inada et al., 2004; Stickens et al., 2004). In addition, osteoblast-derived Mmp13 is responsible for proper cartilage degradation and its inactivation results in increased trabecular bone in mice (Stickens et al., 2004). Beyond normal bone development, Mmp13 also promoted fracture healing through degrading cartilage ECM components, such as aggrecan (Kosaki et al., 2007). In our study, however, mmp13b mutant embryos developed normally suggesting that this matrix metalloprotease is dispensable for normal bone formation. Other Mmps could compensate for the loss of mmp13b in bone formation. Interestingly, however, mmp9, which among other mmp genes, was the most strongly expressed and upregulated in osteoblasts under osteoporotic conditions (Supplementary Table S1), does not appear to compensate for the loss of mmp13b in bone formation. Medaka mmp9−/−;mmp13b−/− double mutant embryos developed normally suggesting that these proteases are not required for regular bone formation. Instead, mmp9−/−;mmp13b−/− double mutants exhibited severely increased bone mass in the vertebral bodies at 5 months, with an obstruction of neural and hemal canals in the vertebral arches leading to a dislocation of blood vessels. This high bone mass phenotype is intriguingly similar to the osteopetrosis-like symptoms that we had earlier reported for an osteoclast-deficient medaka model (To et al., 2015). This thus strongly suggests that mmp9 and mmp13b act synergistically to control the activity of endogenous osteoclasts and control bone remodeling at adult stages.

Medaka mmp13b single mutants instead showed defects under osteoporosis-like conditions. After Rankl induction, ectopic osteoclast formed normally but bone resorption was reduced in mmp13b ^−/− mutants, and this reduction could be recapitulated by chemical Mmp13 inhibition. This strongly suggests that Mmp13b plays a critical role in osteoclast activation under osteoporotic conditions. The role of Mmp13 in regulating osteoclast differentiation and resorption has been studied in various tumor models. For example, MMP13 was found to be induced in breast tumors (Freije et al., 1994), where its up-regulation at the interface between tumor tissue and bone matrix contributed to bone metastasis through promoting osteoclast differentiation (Nannuru et al., 2010; Pivetta et al., 2011). Another study showed that myeloma-derived MMP13 regulates the fusion of osteoclasts and bone resorption (Fu et al., 2016). This shows that MMP13 produced by tumor cells activates osteoclasts and promotes bone metastasis.

In medaka, we demonstrated that the maturation of osteoclasts depends on Mmp13b. We analyzed ten known osteoclast-relevant genes that were earlier shown to be upregulated when macrophages differentiate into mature osteoclasts (Phan et al., 2020a). In mmp13b mutants, these genes were down-regulated after Rankl induction while il10, a known osteoclast inhibitor (Evans and Fox, 2007; Zhang et al., 2014) was up-regulated. Together, this strongly suggests that Mmp13b is required for the maturation of osteoclasts under osteoporosis-like conditions. However, the observation that trap, a well-established osteoclast marker (Andersson and Marks, 1989), was up-regulated indicates that osteoclast differentiation was not entirely blocked in mmp13b mutants, although bone resorption was strongly reduced.

Also, the morphology of osteoclasts was affected in mmp13b mutants. Osteoclasts showed a larger sphericity in mmp13b ^−/− mutants when compared to wildtype, which could reflect the more immature state of these cells. In addition, also the size or surface area of osteoclasts was affected in mmp13b mutants. In wildtype embryos, ctsk-positive osteoclasts were found to exhibit three distinct size ranges, which we defined as immature (cluster 1), mono-nucleated (cluster 2) or multi-nucleated osteoclasts (cluster 3). Importantly, in mmp13b mutants, a fourth size range (cluster 4) was detected, which was similar to cluster 1 and consistent with the idea that these cells were immature. Finally, also the motility of ctsk cells was affected in mmp13b mutants. In wildtype siblings, osteoclasts exhibit a dynamic motility pattern, with a relatively fast speed (approx. 8 nm/s) in the first 1.5 h of the time-lapse analysis (around 48–49.5 hphs), which is consistent to earlier reports (Phan et al., 2020a), followed by a slowing down to a consistent speed of approx. 2 nm/s. This change in speed might reflect the transition of an immature osteoclast approaching the vertebral column to a differentiated osteoclast making contact to mineralized matrix (Phan et al., 2020a). In mmp13b mutants, on the other hand, no elevated speed was observed suggesting possible deficiencies in motility of the immature osteoclasts. In conclusion, our findings on expression profiles, sphericity, size and motility strongly suggest that osteoclasts fail to fully mature in mmp13b mutants.

Similar osteoclast defects were observed in medaka embryos treated with the MMP13 inhibitor CL-82198 hydrochloride, which had earlier been used for the treatment of osteoarthritis (Wang et al., 2013). These inhibitor experiments thus confirmed our mutant analysis and demonstrate that the medaka model can be used for future screening of compounds that interfere with Mmp13 activity to potentially prevent osteoporosis or tumor-caused osteolysis.

In mammals, MMP13 has similar activities in osteocytes and osteoblasts. In both cell types, MMP13 activity is required to remove and remodel the surrounding ECM. On the other hand, our study suggests that in the anosteocytic bone of medaka, osteoblast-derived Mmp13b in addition functions in a non-cell autonomous manner to activate osteoclasts. Whether in osteocyte-inhabited bone, osteocyte-derived MMP13 plays similar roles in osteoclast activation remains to be investigated.

Comparing the transcriptome profiles of osteoblasts during bone remodeling under osteoporosis-like conditions and fin regeneration, we identified a significant up-regulation of mmp13b in col10a1 positive osteoblast progenitors. These col10a1 cells contribute substantially to the regeneration of fin ray segments and joints in medaka (Dasyani et al., 2019). RNA in situ hybridization confirmed mmp13b expression in osteoblasts of segments and fin ray joints. Earlier studies had reported the important role of joint cells in fin regeneration. In zebrafish, mmp9-expressing joint cells represent a progenitor pool for fin regeneration (Ando et al., 2017). Like the situation in mmp9 mutant zebrafish, in our study, we also observed delayed fin regeneration in medaka mmp13b ^−/− single and mmp9 ^−/−;mmp13b ^−/− double mutants. While being expressed in osteoblast progenitors, we propose a direct role for Mmp13b protein in generating an ECM environment that promotes regeneration.

Mmp13 has earlier been implicated in regeneration processes, including skeletal (Behonick et al., 2007), muscle (Lei et al., 2013), heart (Xu et al., 2018), and fin regeneration (Li et al., 2020). In zebrafish, a knock-down of mmp13a resulted in delayed fin growth and impaired osteoblast differentiation (Li et al., 2020). Hence, our findings in stable mmp13b knock-out medaka mutants are consistent with morpholino-based findings in zebrafish.

During fin regeneration, the ECM in the blastema is undergoing extensive remodeling and different matrix proteins are dynamically produced and degraded to generate a distinct ECM environment (Mercer et al., 2012; Govindan and Iovine, 2015). For example, hyaluronic acid (HA), fibronectin (FN) and tenascin C (TNC) are up-regulated to form a transitional matrix for cell proliferation and differentiation during early regeneration stages (Calve et al., 2010). On the other hand, collagen expression is rapidly reduced (Govindan and Iovine, 2015). In our study, we analyzed expression levels and distribution of collagen 1, collagen 2, fibronectin, and tenascin-C in the regenerating fin rays. We found that in regenerating wildtype fins at two stages (0 and 2 dpa), almost no collagen 1 and collagen 2 protein was detectable in the joint regions. This contrasted with mmp13b mutants, where excess collagen 1 and collagen 2 was observed in joints, suggesting impaired collagen degradation when Mmp13b activity is reduced. Two matrix proteins important for the transitional ECM, fibronectin and tenascin-C, also exhibited altered distribution in wildtype siblings and mmp13b mutants. At 2 dpa, fibronectin was upregulated in both wildtype and mmp13b mutants when compared to 0 dpa (data not shown). However, compared to wildtype where fibronectin was found along bone lining osteoblasts, enhanced fibronectin levels were observed in the bone matrix region in mmp13b mutants. Finally, tenascin-C appeared more aggregated in the fibroblast-containing region of mmp13b mutants when compared to wildtype. In addition to these ECM changes, we also observed reduced numbers of col10a1-positive osteoblast progenitors in newly formed joint regions of mutants. Insufficient numbers of progenitors in the joint regions are expected to delay fin regeneration, as joints have been identified as the reservoir of cells needed for fin regeneration (Ando et al., 2017). Together, we hypothesize that Mmp13b is needed to remodel the ECM and allow recruitment of osteoblast progenitors to the joints where they are needed for regeneration.

The initial phases of fin regeneration in zebrafish depend on the recruitment of macrophages (Nguyen-Chi et al., 2017), while at later stages TRAP-positive osteoclasts are detected in the regenerating fins where they contribute to remodeling (Blum and Begemann, 2015; Ando et al., 2017). In our study, we also observed TRAP-positive cells in the regenerating fins at an early stage. While we are unable to distinguish whether these TRAP-positive cells represent macrophages or osteoclasts, no such cells were observed in mmp13b mutants. We thus conclude that remodeling the ECM through Mmp13b is not only required for recruitment of osteoblasts but also of other cell types needed for regeneration.

In conclusion, our study using transcriptome profiling in a medaka model has identified Mmp13b as an important regulator needed for ECM remodeling during tissue regeneration and for maturation of osteoclasts during osteoporotic conditions. The fact that mmp13b was found strongly expressed in osteoblast progenitors in both situations implies interesting scenarios involving cell autonomous and non-cell autonomous roles, respectively. While in regeneration, mmp13b-positive osteoblasts seem to remodel the ECM to attract more osteoblasts into the joint regions, under osteoporotic conditions, osteoblast-derived Mmp13b controls maturation of osteoclasts. Whether the latter is also through ECM remodeling or alternatively through proteolytic activation of other factors needed for osteoclast maturation, remains to be analyzed.