Deletion at an 1q24 locus reveals a critical role of long noncoding RNA DNM3OS in skeletal development

Skeletal development and maintenance are complex processes known to be coordinated by multiple genetic and epigenetic signaling pathways. However, the role of long non-coding RNAs (lncRNAs), a class of crucial epigenetic regulatory molecules, has been under explored in skeletal biology. Here we report a young patient with short stature, hypothalamic dysfunction and mild macrocephaly, who carries a maternally inherited 690 kb deletion at Chr.1q24.2 encompassing a noncoding RNA gene, DNM3OS, embedded on the opposite strand in an intron of the DYNAMIN 3 (DNM3) gene. We show that lncRNA DNM3OS sustains the proliferation of chondrocytes independent of two co-cistronic microRNAs miR-199a and miR-214. We further show that nerve growth factor (NGF), a known factor of chondrocyte growth, is a key target of DNM3OS-mediated control of chondrocyte proliferation. This work demonstrates that DNM3OS is essential for preventing premature differentiation of chondrocytes required for bone growth through endochondral ossification.


Background
Development of the skeletal system that supports body structures and maintenance of its homeostatic state are highly complex processes orchestrated by an elaborate array of gene activities. A frequently occurring developmental skeletal abnormally is short stature, which can arise as part of systemic diseases [1][2][3]. Height gain as the result of bone elongation is driven by chondrogenesis occurring at the growth plate, which is a cartilaginous structure located near the ends of weight-bearing bones in children [4]. Thus, reduced chondrogenesis at growth plates is the main cause underlying short stature.
The rate of chondrogenesis at growth plates depends on several factors, including nutritional intake, systemic hormonal levels, as well as paracrine growth factors and the extracellular matrix signaling [4][5][6][7]. Consequently, genetic lesions that disrupt the regulation of in these systems will result in short stature. In fact, genetic testing of children with abnormal growth has identified multiple intracellular pathways that involved in chondrocyte differentiation in the growth plate. One of them is the RAS oncogene and the mitogen-activated protein kinase (MAPK) signaling pathway, which integrate signals from several growth factors such as growth hormone (GH), fibroblast growth factors (FGFs) and epithelial growth factor (EGF) [8][9][10][11]. Mutations in this pathway underpin a number of genetic syndromes that are collectively termed 'rasopathies'; these include Noonan syndrome, LEOPARD syndrome, Costello syndrome, as well as cardio-facio-cutaneous and neurofibromatosis-Noonan syndrome, a common element of these syndromes is the postnatal growth failure to varying degrees [12].
Since the beginning of the new millennium, mounting evidence has shown that microRNAs and long non-coding RNAs play critical roles in regulating various cellular processes [13]). These non-coding RNAs regulate gene expression and cell signaling through diverse mechanisms. However, while lncRNAs that regulate epigenetic control of gene expression, development as well as govern human traits have been rapidly identified, few lncRNAs have been implicated in skeletal biology. Here, we reported a patient with short stature, hypothalamic dysfunction and mild macrocephaly, who carries a maternally inherited 690 kb deletion at Chr.1q24.2 encompassing a long noncoding RNA gene, DNM3OS. This lncRNA reading frame is of 7.8 kb in length embedded on the opposite strand within the 14th intron of the DYNAMIN 3 gene (DNM3), and carries two microRNAs: miR-199a-5p and miR-214. Our data show that it is capable of promoting primary chondrocyte proliferation and inhibiting chondrocyte differentiation via up-regulating nerve growth factor (NGF). Our data further show that the mouse counterpart of this lncRNA, Dnm3os, is a bona fide regulatory factor of skeleton development independent of its embedded co-cistronic miR199a and miR214.

Deletion of DNM3OS is associated with developmental delay
A young boy was presented to us with clinical manifestation of short stature, hypothalamic dysfunction and mild macrocephaly reminiscent of Noonan syndrome (Fig. 1a, and Additional file 1). The birth weight and height of the proband (II-2) were both below the 5th percentile among Caucasians, and radiographic examinations first conducted at 2 years of age indicated delayed bone growth; however, laboratory tests found no hormonal imbalance, and tests for mutations in 12 known genes compiled in the Noonan Spectrum Chip were also negative (data not shown). Using oligonucleotide-based array comparative genomic hybridization, we identified a maternally inherited 690 kb deletion at 1q24.3 (Fig. 1b), which falls within the common chromosomal deletion interval reported in a cohort of 9 patients with facial features, prenatal-onset short stature with delayed bone age, single palmar crease, and brachydactyly similar to the proband [14]. Deletions around 1q24-1q25 have been noted for growth deficiency among a myriad of symptoms [15]. Among 5 protein-coding genes and undefined open reading frames in the deleted region, we were drawn to a transcription unit called DNM3OS on the opposite strand in the 14th intron of DNM3, which encodes no protein but two microRNAs, miR-199a-5p and miR-214 (Fig. 1b). In the literature, homozygous deletion of murine Dnm3os was reported to cause severe skeletal defects in new born mice including cranial deformity, which was most likely the cause of postnatal lethality, but it was not clear whether the long noncoding RNA Dnm3os as a whole, or miR-199a-5p or miR-214 plays that essential role for the skeletal development [16]. Previously we demonstrated that miR-214 downregulates N-ras to promote myogenic differentiation at the expense of osteogenic differentiation [17]; however, genetic ablation of miR-214 did not lead to obvious signs of Noonan syndrome-like features including growth delay in mice [18,19] (Additional file 2: Fig. S1). Nevertheless, expression of 7 Noonan syndrome genes in the Ras pathway was drastically increased (Additional file 3: Fig. S2) in MEF cells isolated from miR-214 knock-out mice, in keeping with the presence of miR-214 recognition sites in the 3'UTR or coding regions of these RNA transcripts (Additional file 4: Table S1). Using real-time PCR, we found that the levels of DNM3, DNM3OS, and miR-214 RNA transcripts in the peripheral blood of the proband (II-2) and his carrier mother (I-2) were only 50% of those in his non-carrier brother (II-1) (Fig. 1c-f ). Interestingly, the peripheral expression of 4 out of the same 7 Noonan syndrome genes tested in miR-214 KO mice was significantly increased in carriers II-2 and I-2 relative to the non-carrier II-1 ( Fig. 1g-m). These data suggest that DNM3OS carries an essential function that accounts for the roles of chromosomal 1q23-1q25 interval in skeletal development and although the ablation of miR-214 in mice is not sufficient to cause phenotype that resembles any aspect of the partial Noonan syndrome manifestation seen in the proband, it nevertheless partakes in the regulation of skeletal growth via the Ras pathway. Fig. 1 Deletion of lncRNA-DNM3OS is associated with autosomal dominant inheritance of 1q24 deletion syndrome phenotypes. a Family tree of the proband (II-2). b Chromosomal maps of a cohort of 9 individuals with 1q24-25 deletions. c-f RT-qPCR quantifications of DNM3 mRNA, lncRNA-DNM3OS, miR-214, and miR-199 in the blood. The 3' stem-loop primer for human miR-199 recognizes both miR-199a and miR-199b on chromosome 9, the latter of which is not affected by the deletion. g-m RT-qPCR quantifications of N-RAS, CBL, SHOC2, RAF1, BRAF, SOS1, and PTPN11 in the blood. One-way ANOVA test was used for statistical analysis. **P < 0.01, ***P < 0.001, and ns, not significant (See figure on next page.)

Murine Dnm3os is required for maintaining the proliferative potential of articular chondrocytes
Both human and mouse Dnm3os encodes a 7.8 kb RNA transcript, which was identified due to its depleted expression in mouse limb buds lacking basic helixloop-helix transcription factor Twist [20]. First appearing in limb buds and other future skeletal elements [17], robust Dnm3os expression was reported in perichondrial cells and periarticular chondrocytes at the cartilage growth plate [16]. Using RNA-FISH, we found that full length Dnm3os RNA is confined to the nucleus (Fig. 2a). During embryonic development, chondrocytes in the long bone growth plate undergo an orderly proliferation and differentiation process that eventually gives rise to trabecular bones [21]. This chondrogenic process can be faithfully recapitulated in vitro using isolated immature primary mouse articular chondrocytes (iMAC) [22]. We observed that expression of Dnm3os and its co-cistronic miR-199a-5p and miR-214 progressively increased (Fig. 2b, c) following several passages in the maintenance medium as the cells gradually reverted to a "dedifferentiated" state marked by the switch of cell surface collagen types from Col2a1 to Col1a1 (Fig. 2d, e). Correspondingly, the expression of chondrogenic transcription factor SOX9 gradually decreases with passaging (Additional file 5: Fig. S3A). When introducing shDnm3os, the switch of cell surface collagen types from Col2a1 to Col1a1 was slightly slowed, suggesting that Dnm3os was not a key fact for dedifferentiated (Additional file 5: Fig. S3B). Conversely, expression levels of these non-coding RNAs all decreased dramatically after the cells were induced to differentiate into hypertrophic chondrocytes marked by Col10a1 and MMP13 (Fig. 2f, g). These spatial and temporal expression patterns of Dnm3os are consistent with an essential role in promoting the proliferation but suppressing the differentiation of chondrocytes. To ascertain such a function, we silenced Dnm3os expression using shRNAs in the iMAC before inducing their differentiation. At the end of two weeks, immunofluorescence microscopy indicated that over 60% of scrambled shRNA transfected cells (marked by GFP) still retained Col2a1 and Sox9, but these percentages dropped to below 20 and 40%, respectively, in the cells transfected with shDnm3os ( Fig. 2h, i, l, m). Remarkably, many GFP positive cells that received shDnm3os completely lost Col2a1 and Sox9 (Fig. 2h, i, l, m), indicating that they had differentiated further into hypertrophic chondrocytes (Fig. 2b). Elevated expression of Col10a and Mmp13 confirmed the role of silencing Dnm3os in promoting chondrogenic differentiation (Fig. 2j, k, n, o). The expression level of Dnm3os maintained at a relative low level in the cells transfected with shDnm3os (Fig. 2p). Moreover, EdU assay showed Dnm3os silencing decreased the proliferation of iMAC by 10% (Additional file 5: Fig. S3 C-E).
In addition to controlling Ras pathway genes (Additional file 3:  Dnm3os is required for maintaining the proliferative potential of ATDC5 cells To corroborate the above observation in primary mouse articular chondrocytes, we took the advantage of ATDC5 cells that are widely used as an in vitro model for chondrocyte differentiation to test the pro-proliferative role of Dnm3os. Alcian blue staining images collected by light microscopy over an 18-day time-course indicated that insulin-supplemented differentiation medium cultivation successfully induced the differentiation of ATDC5 cells to mature hypertrophic chondrocytes (Additional file 8: Fig. S6A). Decreased Col2a1 and increased Col1a1 expression confirmed the differentiation on the biomarker (Additional file 8: Fig. S6B, C). Consistent with primary mouse articular chondrocytes, the expression of Dnm3os in ATDC5 cells gradually decreased with the chondrogenic differentiation (Additional file 8: Fig. S6D).
Overexpression of FL and the DKO mutant Dnm3os activated ATDC5 cells proliferation and impeded the chondrogenesis as evident by 2-fold increased EDU positive cells (Fig. 4a, b and Additional file 5: Fig. S3I) and absent Alcian blue staining (Fig. 4c), while silencing Dnm3os with shRNA inhibited the proliferation of ATDC5 cells (Fig. 4d, e). We further employed RT-QPCR to examined the expression of Sox9 in FL and DKO transfected ATDC5 cells and iMAC cells. Both overexpression of FL and DKO increased the expression level of Sox9 in ATDC5 cells, while the effect in iMAC cells were barely observed (Additional file 8: Fig. S6E, F). However, the expression of Sox9 decreased with iMAC cells passaging , and one day later the cells were labeled with EdU for 8 hours before visualization. The frequency of nuclear EdU labeling is determined by examination of at least three random fields, magnification ×400 and at least 300 cells and nuclei in each group. One-way ANOVA test was used for statistical analysis. **P < 0.01, ***P < 0.001

Fig. 4
LncRNA-Dnm3os is required for maintaining the proliferative potential of the ATDC5 cells. a EdU labeling of proliferating ATDC5 cells and b calculation thereof. The ATDC5 cells were transfected by RK5, FL and DKO individually, and one day later the cells were labeled with EdU for 8 hours before visualization. c Alcian blue staining of cartilage matrix produced by differentiated ATDC5 cells. d EdU labeling of proliferating ATDC5 cells and e calculation thereof. The ATDC5 cells were transfected with scrambled shRNA, shDnm3os-1 and shDnm3os-2 individually, and one day later the cells were labeled with EdU for 8 h before visualization. The frequency of nuclear EdU labeling in (a) and (d) were determined by examination of at least three random fields, magnification ×400 and at least 300 cells and nuclei in each group. One-way ANOVA test was used for statistical analysis. ***P < 0.001, ****P < 0.0001, and ns, not significant (Additional file 5: Fig. S3A), which suggested the overexpression of FL and DKO to some extent counteracted the reduction of Sox9 caused by passage. These data suggest that Dnm3os contributes to the maintenance of ATDC5 and iMAC cells as chondrogenic progenitors and plays an independent role in regulating chondrogenesis instead of simply acting as a precursor of miRNAs.

NGF is a potential Dnm3os-regulated gene in chondrocytes
To substantiate the direct link between Dnm3os and chondrogenesis, we performed RNA-seq in FL and DKO mutant Dnm3os transfected primary mouse articular chondrocytes with 2 biological replicates. The result showed 250 and 211 differentially expressed genes in FL and DKO overexpressed cells relative to the RK5 transfected control cells, respectively (Fig. 5a), and of these, 155 were commonly expressed in those two cells transfected with Dnm3os vectors, suggesting that this lncRNA functions as an independent regulator of chondrocytes (Fig. 5a) and DKO acts as a regulator at the transcriptional level. To predict the biological function of Dnm3os, we categorized these 155 genes into defined pathways and we found cell cycle appeared in the top altered pathways, which confirmed the proliferation effect of Dnm3os (Additional file 8: Fig. S6G). To further provide biological insight to the differential expression genes, the 155 genes were assigned into defined pathways by using Reactome databases (Fig. 5b). Among the top 20 significantly altered pathways ranked in the dot plot enrichment map, signal transduction, metabolism and extracellular matrix organization, whose dynamics are key to tissue morphogenesis appeared with majority gene enrichment and statistical significance. We further dissected 8 of the top enriched pathways and found matrix metalloproteinase 3, collagen genes including Col8a2, Col23a and growth factors including NGF, FGF, as visualized by the heat map (Fig. 5c). The expression changes of these genes suggested the ongoing extracellular matrix remodeling and accelerated cell proliferation. Subsequently, we picked up 14 genes which have been reported to play a role in chondrocyte proliferation or differentiation and verified the change by QPCR analysis (Fig. 5d, e). To exclude the miR-214 target genes, we detected the expression of these 14 genes in miR-214 transfected ATDC5 cells simultaneously, and found NGF was a potential target gene regulated by lncRNA-Dnm3os specifically (Fig. 5e).

Dnm3os-induced NGF maintains the proliferative potential of ATDC5 cells
As a neuropeptide, NGF was involved in cartilage metabolism and is reported to mediate the chondrogenic differentiation of mesenchymal stem cells [23]. To determine if NGF play a role in Dnm3os mediated chondrocytes proliferation, we treat FL, DKO mutant and miR-214 transfected ATDC5 cells with PD90780, an inhibitor of NGF. 24 hours after treatment, we labeled the cells with EdU and found that inhibition of NGF blocked FL and DKO mutant Dnm3os-induced chondrocytes proliferation (Fig. 6a, b). On the other hand, the chondrogenesis was promoted by PD90780 as evident by increased staining by alcian blue in FL and DKO mutant Dnm3os transfected cells after 18-days differentiation (Fig. 6c,  d). Consistently, FL and DKO mutant Dnm3os-induced collagen types switch from Col1a1 to Col2a1 was interrupted by PD90780 (Fig. 6e, f ). MMP9, a target gene of NGF was downregulated with treatment of PD90780 and validated the inhibitory effect of the compound (Fig. 6g).
RT-QPCR analysis of Col10a1 confirmed the that inhibition of NGF in FL and DKO constructs overexpressing cells promoted differentiation (Fig. 6h). These data collectively indicated these chondrocytes have been released from the bondage of Dnm3os and entered the differentiation state.

Discussion
Skeletal development is a complex process that exquisitely controlled both spatially and temporally by cell signaling networks and gene regulation programs. In the past decades, studies of congenital human disease reveal a great deal of genes that involved in bone growth including Ras-MAPK pathway, Wnt and Hedgehog signaling [24,25]. However, the contribution of non-coding RNA especially lncRNA to skeletal development still remains unclear. Here, we present evidence that shows LncRNA-Dnm3os is required for maintaining the proliferative potential of articular chondrocytes, and we demonstrate that Dnm3os has specific gene targets such as NGF, which impeded chondrocytes to differentiation state. Thus, Dnm3os defines a new class of lncRNAs that serve as transcriptional regulator in addition to produce microRNAs, thereby forming a regulatory network that maintains a proper pool of proliferating chondrocytes to supply bone growth through endochondral ossification, which account for the short stature.
In 2011, Burkardt et al. reported nine patients with a core clinical symptom of mental retardation, microcephaly accompanied by short stature, and identified a crucial deletion region spanning 1.9 Mb at 1q24.3q25.1 [14]. The deleted region contains 13 genes including Dynamin 3 (DNM3) and CENPL,which encodes a protein essential for centromeric function, mitotic progression and synaptic reaction. Later, Ashraf T described 2 patients with 1q24 microdeletions and the skeletal phenotype, but had normal intellect or mild learning impairment [26]. Genetic testing of these 2 patients narrows the skeletal abnormalities to a region containing only DNM3 and a transcript union in the opposite strand of DNM3, which called DNM3 opposite strand (Dnm3os). This unit can be transcribed into a long non-coding RNA Dnm3os (LncRNA-Dnm3os), which was described as a precursor of two microRNAs: miR-199a-5p and miR-214. It was reported that Dnm3os, miR214 and miR199a-5p are abundant in the skeleton system containing limb and skull [27]. Dnm3os deletion mice exhibited several skeletal abnormalities, including craniofacial hypoplasia and defects of dorsal neural arches [16]. Down regulation of miR199a-214 cluster, especially miR-214 was considered to be responsible of the phenotype. While for quite a long time, LncRNA-Dnm3os was described as a precursor of these two miRNAs. However, miR-214 KO mice were born at Mendelian ratios and displayed a minimal reduction in body weight compared with WT littermates. Our experiments revealed scientific explanations for these paradoxical phenomena. First, LncRNA-Dnm3os may have compensatory effect of skeletal development since LncRNA-Dnm3os shares most of the target genes with miR-214, and genetic deletion of the miR199a-214 cluster did not abolish the regulatory effect of LncRNA-Dnm3os (Figs. 3 and 4). The other possible explanation is LncRNA-Dnm3os has its own target genes which involved in skeletal development. Thus, the Dnm3os transcript unit defines a regulatory network between lncRNA and miRNAs.
During the past decades, emerging evidence suggests that lncRNAs can play a crucial role in manipulating various cellular processes. In particular, lncRNAs can serve as mater gene regulators at transcriptional and posttranscriptional levels,participating in embryonic development and occurrence of diseases. In some cases, lncRNAs can act as baits of microRNA and sequestrate microRNA for target mRNAs transcriptional repression. Other lncRNAs including BACE1 AS regulate gene expression by competing with miRNAs. And for some lncRNAs, their degradation can be triggered by micro-RNA. Recently, research works defined the interaction between nucleolin, ILF-2 and lncRNA-Dnm3os by RNA pull-down assays with macrophage nuclear lysates, which indicates that lncRNA-Dnm3os is more than a precursor of microRNAs. Due to the obvious and direct link to vertebrate skeleton development, lncRNA-Dnm3os may also have unique function in this biological process.

Conclusions
This study demonstrated that lncRNA-DNM3OS maintains the proliferation and restrains premature differentiation of chondrocytes independent of the co-cistronic microRNAs miR-199a and miR-214. In addition, mechanistic studies showed NGF as a key target of lncRNA-DNM3OS that supports chondrocyte proliferation. Combined with our findings, lncRNA-DNM3OS likely plays an important role in regulating skeletal development by triggering NGF signaling. Future studies are required to ascertain whether there are more particular genes or signaling pathways regulated by lncRNA-DNM3OS.

Single family case studies
The studies of this case obtained the informed consent from all subjects. For publication of photos, consent was obtained from the patients.

Animals
The studies with animals follow the guidelines and ethical regulations. The research program and study protocols were approved by Animal Care and Ethical Committee of Nanjing Medical University (approval number 14,030,111).

Isolation of mouse articular chondrocytes and culture
Primary chondrocytes were isolated from 5 to 6 days old mice as described. Briefly, cartilages from tibial plateaus and femoral condyles were excised and all extraneous soft tissues were removed. To isolate chondrocytes, the cleaned cartilages were digested two consecutive times in cell culture medium with 3 mg/ml collagenase D (Roche, Indianapolis, 11088858001) at 37 °C for 45 min, and then overnight with 0.5 mg/ml collagenase D. The next day, the dislodged cells were passed through a 2 ml Pasteur pipet successively to disperse aggregates, then through a sterile 48-µm mesh before collected by centrifugation for 10 min at 400×g, 20 °C. The cells were plated out at a density of 25 × 10 3 /cm 2 . Chondrogenic differentiation was induced by replacing culture medium with DMEM/ F12 (1:1) (Gibco,C11330500BT), supplemented with 10% FBS, 1% pen&strep (Gibco, 15140122), 10 µg/ml insulin, 10 µg/ml transferrin, and 3 × 10 − 8 M sodium selenite (Sigma, I3146). The differentiation medium was changed every 3 days.

Plasmids construction and cell transfection
The DNA fragment containing full length Dnm3os sequence was amplified by PCR from mouse genomic DNA and inserted between the BamHI and SalI sites in the pRK5 vector to generate pRK5-Dnm3os. PCR-based deletion strategy was used to generate the miR-199a-5p and miR-214 deletion mutant (DKO). The PCR primers used are as follows.  Table S2) of shDnm3os were cloned into pRetro-H1G shRNA expression vector. ShRNA with the sense sequence 5′-agatctTTA GTA TAG ATA ATA TTT CtacctgacccataGAA ATA TTAT CTA TAC TAT TTT Tggtacc-3′ which lacks complementary sequences in the human genome, was used as control (scrambled shRNA).
Blood RNA extraction and RT-qPCR 10 ml blood from each subject was drawn with a BD Vacutainer CPT Cell Preparation Tube containing sodium citrate. Lymphocytes and monocytes were separated from the plasma in Ficoll solution (Sigma, F2637). Briefly, the blood samples were diluted with 1:1 sterile PBS, and then carefully poured onto 10 ml Ficoll solution in a 50 ml centrifuge tube (the blood must remain on top, do not mix). The tubes were centrifuged for 20 min at 350xg, and lymphocytes and monocytes between plasma and Ficoll layers were harvested using a sterile pipette. The cells were washed twice with PBS, and the RNA was extracted using the RNAiso reagent (Takara,9109) to template cDNA synthesis using the PrimeScript RT reagent kit (Takara,RR014). SYBR green real-time qPCR reactions (Vazyme,Q111) were carried out on a ABI7500 Real-Time PCR system. The cycling condition was 95°C for 5 minutes, followed by 40 amplification cycles of 95°C, 15 second and 60°C, 1 minute. For each data point, triplicate reactions were carried out and the experiment was repeated three times to assess the statistical significance. RT-qPCR primer sequences are listed as follows.

Edu incorporation assay
To measure cell growth, 24 hours after transfection in 24-well plates, 20 mM EdU (Ribobio,C10310-1) was added for 8 hour. The cells were then fixed in 3.7% formaldehyde, then washed with PBS and permeablized. 500 µl Click-iT reaction cocktail (430 ml 1xClick-iT reaction buffer, 20 ml CuSO 4 , 1.2 ml Alexa Fluor ® azide, and 50 ml reaction buffer additive) was added to each well and incubated for 45 min at the room temperature in the dark. The cells were counter-stained with DAPI for nuclei and visualized under an inverted fluorescence microscope. Images were processed with Image J and the percentage of EdU incorporation was calculated based on the number of EdU positive (red) and total (DAPI) cells.

Alcian blue staining and quantification
Chondrocytes were cultured for 14 days in chondrogenic differentiation medium, then fixed in 4% formalin for 10 min. After washing twice with PBS, the cells were incubated with 3% acetic acid for 10 min and stained with 1% alcian blue in 3% acetic acid (pH 2.5) for 30 min and photographed. For quantification, the stained cells were washed twice, and the alcian blue dye was extracted with 500 ml dimethyl sulfoxide (Sigma,D8418). Absorbance was measured at 650 nm.

Alkaline phosphatase assay
Histochemical detection of alkaline phosphatase activity was performed on cells that were fixed for 2 min in 4% paraformaldehyde at room temperature. After washing with TBST, the cells were incubated for 30 min in in 0.1 M Tris-HCl, pH 8.5, containing 0.1 mg/ ml Naphthol AS-MX phosphate (Sigma,N4875), 0.5% N,N-dimethylformamide (Sigma,D4551), 2 mM MgCl 2 , and 0.6 mg/ml fast blue BB salt (Sigma,D9805), and then photographed.

RNA sequencing and data analysis
RNA from Rk5-vector, Full length and DKO mutant Dnm3os transfected primary mouse articular chondrocytes were extracted using TRIzol (Life Technol-ogies), followed by purification using a RNeasy Mini Kit (Qiagen,74104). RNA-seq was performed using primary mouse articular chondrocytes from two individual animals. RNA-seq libraries were prepared using the Illumina TruSeq RNA Library Prep Kit v2 and sequenced by a HiSeq 2500 sequencer. RNA-seq reads were aligned to mm10 using TopHat with default settings (.cbcb.umd. edu/).

Statistical analysis
Prism 8.0 (GraphPad software) was employed for analyses. Data from a minimum of three independent experiments were presented as mean ± s.d. Animals in the experiment were randomly selected and grouped. An unpaired two-tailed Student's t-test was used for analyzing two data sets, and one-way analysis of variance was used for more sets. The significance threshold was set at 0.05 (p < 0.05).