SCF^JFK regulates osteoblast differentiation through targeting RUNX2

Abstract

RUNX2 (Runt-related transcription factor 2) is a master regulator of osteogenesis, and its genetic mutations are associated with ~65% cases of cleidocranial dysplasia (CCD), an autosomal dominant abnormal bone development disorder in humans with defective intramembranous bone formation. However, how these mutations affect bone formation remains to be investigated. Here, we report that RUNX2 interacts with the F-box protein JFK and is destabilized by the SKP1-CUL1-JFK E3 ubiquitin ligase complex (SCF^JFK). We find that several CCD-associated mutations of RUNX2 acquire an increased affinity toward SCF^JFK thus an accelerated proteasomal degradation. We demonstrate that SCF^JFK-mediated RUNX2 degradation suppresses osteoblast differentiation. Consistently, Jfk-null mice exhibit increased trabecular bone volume and bone mineral density and are resistant to Runx2 haploinsufficiency-induced CCD-like syndrome. Interestingly, RUNX2 binds to the JFK promoter and represses its transcription, establishing a feedback loop to promote osteoblast differentiation, and remarkably, the CCD-associated RUNX2 mutants also display an impaired transcription repression of JFK. Our study demonstrates SCF^JFK as an E3 ligase for RUNX2 and uncovers a feedback regulatory loop between SCF^JFK and RUNX2 that is implemented in bone development and implicated in CCD, supporting the pursuit of JFK as a potential target to ameliorate CCD-like syndrome.

Introduction

Bone formation is orchestrated mainly by osteoblasts, which are derived from mesenchymal stem cells (MSCs) and are required for intramembranous ossification in the calvarium, clavicles, facial bones, other flat bones, and periosteum, whereas endochondral bone formation involves cartilaginous templates that become progressively replaced by osteoblast‑derived bone.¹ Growth factors, cytokines, and lineage-specific transcription factors—such as RUNX2 and SP7—tightly govern the spatiotemporal patterns of osteoblast differentiation. Dysregulation of these factors underlies various pathological states of bone, including cleidocranial dysplasia (CCD), an autosomal-dominant inherited skeletal disorder. CCD patients typically exhibit persistently open or delayed closure of cranial sutures, underdeveloped or absent clavicles, and multiple dental abnormalities—such as delayed permanent tooth eruption with retained primary teeth, and the presence of supernumerary teeth.² Beyond these hallmark features, compared with unaffected family members, CCD patients may also present with other digital anomalies—for example, shortened fingers, tapered digits, and broadened thumbs—as well as reduced stature. Additional orthopedic manifestations, including pes planus, genu valgum, and scoliosis, are also frequently observed.² Genetic alterations in the RUNX2 gene—including chromosomal translocation, deletion, insertion, nonsense, splice-site, and missense mutations—have been reported in approximately 65% of CCD patients.^1,3,4 However, CCD patients display highly variable phenotypes, from primary dental anomalies to a complete spectrum of CCD features accompanied by osteoporosis, and no clear phenotype‒genotype correlation has emerged that adequately explains the mechanistic involvement of RUNX2 in the pathogenesis of CCD.

As a master regulator of osteogenesis, RUNX2 functions as a transcription factor that binds to DNA elements and recruits cofactor proteins to regulate osteoblast differentiation during embryogenesis and to influence skeletal morphogenesis, chondrogenesis, tooth development, and bone metabolism/maintenance.⁵ Most CCD‑associated RUNX2 mutations are heterozygous and cluster within its Runt domain, which mediates DNA binding and interaction with the co-factor core-binding factor beta (CBFβ). Although it has been proposed that RUNX2 mutations contribute to the pathogenesis of CCD by impairing interactions with DNA or CBFβ and abolishing transcriptional activity,^2,5 some CCD-associated RUNX2 mutants generate protein products that retain at least partial DNA-binding activity, implying that this mechanism does not apply to all RUNX2 mutations. Notably, RUNX2 also undergoes extensive posttranslational regulation, including phosphorylation, sumoylation, and ubiquitination. Dysregulation of these posttranslational modifications alters the transcriptional control of RUNX2, as well as its subcellular localization and stability.⁶ Moreover, the level and stability of cellular RUNX2 are certainly involved in its function during osteoblast differentiation, and several ubiquitin ligases, including WWP1,⁷ WWP2,⁸ SMURF1/2,⁹ and CHIP,¹⁰ have been shown to promote RUNX2 polyubiquitination and subsequent degradation. Clearly, the mechanistic involvement of RUNX2 mutations in the pathogenesis of CCD—particularly regarding how distinct mutations differentially affect its post‑translational modifications and proteasomal degradation—needs further investigation.

Numerous cellular processes, including cell cycle progression, signal transduction, and differentiation, are regulated by targeted proteolysis via ubiquitin‒proteasome pathway, which represents a major regulatory paradigm.¹¹ Dysregulation of this system is often associated with a variety of pathological states, ranging from cancer to metabolic and skeletal disorders.¹² Ubiquitination proceeds through a cascade of enzymatic reactions carried out by E1 (activating), E2 (conjugating) and E3 (ligase) enzymes.¹³ E3 ubiquitin ligases are the most diverse and are structurally and compositionally classified into families, including the HECT, RING-finger, and U-box types.¹⁴ As a multisubunit RING-finger complex, the SCF E3 ubiquitin ligase complex contains the constant subunits—SKP1, CUL1, and RBX1—and a variable F-box protein (FBP). FBPs recognize and specify substrates through the F-box domain, thereby determining the target specificity of the entire complex.¹⁵ Beyond this domain, FBPs harbor C‑terminal motifs for protein‒protein interactions. Kelch domain is known to mediate protein‒protein interactions, and the only Kelch domain‑containing FBP we previously identified in humans is JFK.^16,17 Subsequent studies in our laboratory have shown that JFK is involved in various pathophysiological processes, including tumorigenesis, obesity, and metabolic syndrome, through the targeting of distinct substrates, such as p53, ING4, and ING5.^16–18

Our study reports that SCF^JFK is an E3 ligase for RUNX2. We find that several CCD-associated RUNX2 point mutants exhibit enhanced interactions with SCF^JFK, leading to accelerated proteasomal degradation. We show that SCF^JFK-mediated RUNX2 degradation suppresses osteoblast differentiation and that Jfk knockout promotes embryonic bone development and mitigates Runx2 haploinsufficiency-induced CCD-like syndrome in mice. Furthermore, we demonstrate that RUNX2 represses JFK transcription by binding to its promoter, thereby establishing a feedback loop for osteoblast differentiation and bone development, and that CCD-associated RUNX2 mutants also display impaired transcriptional repression of JFK. Our findings identify JFK as a potential therapeutic target for alleviating CCD‑like syndrome.

Results

Several CCD-associated RUNX2 mutants exhibit increased polyubiquitination and accelerated degradation

As stated above, RUNX2 mutation is detected in approximately 65% of CCD patients, and more than 200 RUNX2 mutations, including missense, nonsense, and frameshift mutations, have been reported.¹⁹ Intriguingly, some CCD-associated RUNX2 mutants still produce protein products that at least partially maintain their DNA-binding functions.²⁰ Dental follicle cells (DFCs) are MSCs derived from the developing tooth follicle and are recognized for their multipotent differentiation potential, including into osteoblasts, adipocytes, and chondrocytes.²¹ To gain further understanding of the mechanistic involvement of RUNX2 mutations in osteogenesis, we collected primary DFCs from both healthy donors (HDFCs, RUNX2^+/+) and a CCD patient (CDFCs, RUNX2^+/mu) with a point mutation of RUNX2 in the Runt homology domain (RHD) (c.557 G > C, p. R186T) and with typical clinical manifestations of CCD.^22,23 On the basis of a previous report that the R186T mutation of RUNX2 (RUNX2^R186T) was associated with decreased protein level in DFCs,²² we first investigated the protein level and protein stability of RUNX2^R186T. Consistent with a previous study,²² western blotting analysis of isolated primary DFCs confirmed the reduced protein level of RUNX2^R186T in CDFCs (Supplementary Fig. 1a). Cycloheximide (CHX) chase assays in isolated primary DFCs revealed that RUNX2 half-life was significantly reduced in CDFCs compared with that in HDFCs (Fig. 1a). Interestingly, however, with MG132 treatment, the accumulation of RUNX2 was comparable in CDFCs and HDFCs (Fig. 1a), suggesting that the decreased RUNX2 accumulation in DFCs is linked to proteasome-mediated protein degradation. In support of this, in vivo ubiquitination assays based on immunoprecipitation with antibody against RUNX2 revealed a significantly increased level of polyubiquitinated RUNX2 in primary CDFCs compared with that in primary HDFCs (Fig. 1b).

Fig. 1.

Several CCD-associated RUNX2 mutants exhibit increased polyubiquitination and accelerated degradation. a Cycloheximide (CHX) chase assays were performed in primary DFCs collected from healthy donor (HDFCs, RUNX2^+/+) or CCD patient (CDFCs, RUNX2^+/mu) treated with DMSO or the proteasome inhibitor MG132. Quantification was performed by densitometry, and the data are expressed as the RUNX2/β-actin ratio. Data are presented as the mean ± SD (n = 3). The asterisk (*) represents significant difference from the HDFC group. b Cellular extracts from HDFCs or CDFCs treated with DMSO or MG132 were prepared for in vivo ubiquitination assays. c Schematic representation of various mutations within RUNX2 gene. QA polyglutamine and polyalanine domain, RHD Runt homology domain, TAD transcription activation domain. d U2OS cells were treated with RUNX2 3’-UTR siRNAs and subsequently transfected with the indicated Myc-tagged expression constructs. CHX chase assays were performed. Quantification was performed by densitometry, and the data are expressed as the RUNX2/β-actin ratio. Data are presented as the mean ± SD (n = 3). The asterisk (*) represents significant difference from the RUNX2^wt group. e Cellular extracts from U2OS cells treated with RUNX2 3’-UTR siRNAs and subsequently co-transfected with the indicated Myc-tagged expression constructs and HA-Ub in the presence of MG132 were prepared for in vivo ubiquitination assays. Data are presented as the mean ± SD (n = 3)

To investigate the functional significance of the altered protein stability of RUNX2 mutants in the pathogenesis of CCD and to gain a broader view of RUNX2 mutations in the development of CCD, we generated Myc-tagged wild-type RUNX2 (Myc-RUNX2^wt) and 9 RUNX2 mutants (Myc-RUNX2^mu) carrying CCD-associated recurrent mutations in RHD or in the transcription activation domain (TAD) by site-directed mutagenesis: R131G, M175R, R193Q, and Q209H with lost DNA-binding capacity;^2,24,25 F182S and T200I partially maintain DNA-binding capacity;^26,27 R225Q with disrupted subcellular localization;²⁸ the aforementioned R186T mutation with decreased RUNX2 accumulation in DFCs;^20,29 and A362V mutation in TAD (Fig. 1c). CHX chase assays in U2OS cells transfected with Myc-RUNX2^wt or Myc-RUNX2^mu revealed that R131G, M175R, R193Q, Q209H, and A362V mutations did not affect RUNX2, F182S, R186T, T200I, and R225Q mutations all resulted in reduced RUNX2 half-life, albeit to different degrees, with the most dramatic effect detected in R186T and R225Q (Fig. 1d). The increased proteasomal degradation of the RUNX2 mutants could be effectively blocked by MG132 (Fig. 1d). In corroboration, in vivo ubiquitination assays and western blotting revealed that F182S, R186T, T200I, and R225Q were indeed associated with increased polyubiquitination but not protein level with MG132 treatment (Fig. 1e and Supplementary Fig. 1b, c). Together, these data indicate that CCD-associated mutations of RUNX2 could render the protein a decreased stability, which is mechanistically linked to the ubiquitin‒proteasome pathway.

RUNX2 physically interacts with JFK, which is promoted by the RUNX2 R186T mutation

To elucidate the molecular mechanism involved in the RUNX2 mutation-associated increase in the polyubiquitination and proteasomal degradation of RUNX2, we utilized SILAC coupled with LC‒MS/MS to examine differences in the cellular interactome between RUNX2^wt and RUNX2^mu cells. To this end, U2OS cells were maintained in growth media containing “light” or “heavy” arginine and lysine for 2 weeks (> 6 cell divisions) prior to transfection with FLAG-RUNX2^wt or FLAG-RUNX2^R186T, followed by treatment with MG132 before sampling. Whole-cell extracts from light- and heavy-labeled cells were prepared and mixed in equal amounts, then and applied to anti-FLAG affinity columns. We digested the bound proteins with trypsin, and the resulting peptides were subjected to LC‒MS/MS, which resolved them into light and heavy pairs with mass differences dictated by the extent of stable isotope labeling (Fig. 2a). Multiple proteins, including CBFβ, WWP1, SMURF2, JFK, SKP1, CUL1, and RBX1, were found to co-purify with RUNX2 (Fig. 2b). Among these proteins, CBFβ is the heterodimeric partner of RUNX2, and WWP1 and SMURF2 are two E3 ligases known to target RUNX2 for polyubiquitination and degradation;^7–9 interestingly, our proteomics analysis revealed that while the enrichment of WWP1, SMURF2, and CBFβ was comparable or even lower in the RUNX2^R186T interactome than in the RUNX2^wt interactome, JFK,^17,18 as well as SKP1, CUL1, and RBX1, was more enriched in the RUNX2^R186T interactome than in the RUNX2^wt interactome (Fig. 2b). Specifically, LC‒MS/MS revealed 4.76-fold and 10.25-fold increases in the ratios (heavy/light) of JFK and SKP1 peptides, respectively, in cells overexpressing RUNX2^R186T (Fig. 2b). We also performed co-immunoprecipitation assays in U2OS cells transfected with FLAG-RUNX2^wt or FLAG-RUNX2^R186T. Compared with RUNX2^wt, RUNX2^R186T showed an increased interaction with JFK and with the SCF components SKP1, CUL1, and RBX1 (Fig. 2c). Consistently, co-immunoprecipitation assays with cellular extracts from primary DFCs also demonstrated that RUNX2^R186T exhibited an increased interaction with JFK and SCF components (Fig. 2d).

Fig. 2.

RUNX2 physically interacts with JFK, which is promoted by the RUNX2 R186T mutation. a Experimental workflow for the quantification of FLAG-RUNX2^wt- or FLAG-RUNX2^R186T-interacting proteins. b Quantification of the total number of peptides associated with RUNX2^wt (light) or RUNX2^R186T (heavy). c Cellular extracts from U2OS cells transfected with empty vector, FLAG-RUNX2^wt, or FLAG-RUNX2^R186T were immunoprecipitated with anti-FLAG immunoaffinity resin and immunoblotted with antibodies against the indicated proteins. Data are presented as the mean ± SD (n = 3). d Cellular extracts from HDFCs or CDFCs were immunoprecipitated with anti-RUNX2 and immunoblotted with antibodies against the indicated proteins. Data are presented as the mean ± SD (n = 3). e The results of the GST pull-down assays with GST-RUNX2^wt or GST-RUNX2^R186 and in vitro transcribed/translated JFK, CUL1, SKP1, or RBX1. f Schematic representation of wild-type JFK and its mutants. The results of the GST pull-down assays with GST-JFK, GST-JFK^ΔF-box, or GST-JFK^ΔKelch and in vitro transcribed/translated RUNX2 are shown. g Equilibrium dissociation constant (Kd) for the interaction of GST-JFK with U2OS cells expressing His-RUNX2^wt or His-RUNX2^R186 measured by SPR. h The overall architecture of the CUL1-RBX1-SKP1-JFK-RUNX2 complex was generated using AlphaFold-predicted models and HDOCK software. i Detailed structural illustrations depicting the interface between JFK and SKP1 and between JFK and RUNX2^wt

To investigate whether RUNX2 is a potential substrate of the SCF^JFK E3 ligase, we examined the direct SCF^JFK complex‒RUNX2 interaction. Consistent with the established function of FBPs as substrate receptors in SCF ubiquitin ligases, GST pull-down assays revealed that both GST-RUNX2^wt and GST-RUNX2^R186T interact with JFK, whereas no binding was detected with CUL1, SKP1, or RBX1 (Fig. 2e). Interestingly, the interaction between JFK and RUNX2^R186T was stronger than that between JFK and RUNX2^wt (Fig. 2e). Additionally, GST pull-down assays using GST-JFK, GST-JFK^ΔF-box (JFK with the F-box deleted), or GST-JFK^ΔKelch (JFK with the Kelch domain deleted) and in vitro transcribed/translated RUNX2 demonstrated that the Kelch domain is responsible for JFK–RUNX2 interaction (Fig. 2f). Furthermore, to quantify the binding affinity between JFK and RUNX2^wt and between JFK and RUNX2^R186T, we performed surface plasmon resonance (SPR) assays using bacterially expressed GST-JFK and eukaryotic U2OS cells expressing/His-tag-purified RUNX2^wt or His-RUNX2^R186T. The results confirmed the specific interaction of JFK with both RUNX2^wt and RUNX2^R186T, with dissociation constants (Kds) of 8.39e-7 μM and 5.33e-7 μM, respectively (Fig. 2g). Importantly, the results of the SPR assays revealed that JFK had a stronger affinity for RUNX2^R186T than for RUNX2^wt, as evidenced by a 1.57-fold greater binding strength for RUNX2^R186T than for RUNX2^wt. These data support this receptor–substrate relationship between JFK and RUNX2 and suggest that this interaction is enhanced by the RUNX2 R186T mutation.

To further confirm the SCF^JFK complex–RUNX2 interaction and to gain structural insight into this interaction, we employed a stepwise docking approach using HDOCK software.³⁰ The SCF complex (PDB ID: 1LDJ) was retrieved from RCSB PDB. This structure was then docked with the AlphaFold-predicted model of JFK to form the intermediate SCF^JFK complex. This intermediate complex was subsequently docked with the AlphaFold-predicted model of RUNX2, yielding the overall architecture of the complex (Fig. 2h). Notably, the interface between JFK and SKP1 revealed that the JFK’s F-box domain (residues 39-93) was positioned in close proximity to SKP1, forming putative hydrogen bonds between Y57 of JFK and D117 of SKP1 (Fig. 2i), which is consistent with the interaction in a typical SCF complex.³¹ Interestingly, the interface between JFK and RUNX2 indicated that the Kelch domains of JFK (residues 107-354 and 667-717) were positioned primarily in close proximity to RUNX2 (Fig. 2i), forming putative hydrogen bonds between N258 of JFK and S165 of RUNX2, as well as between V671/G672 of JFK and Q62 of RUNX2. These structural analyses support the role of the F-box and the Kelch domain in mediating protein‒protein interactions within the SCF^JFK complex, substantiating the notion that RUNX2 physically interacts with JFK and is a substrate of SCF^JFK complex.

SCF^JFK is an E3 ubiquitin ligase for RUNX2 and CCD-associated RUNX2^R186T accelerates its SCF^JFK-mediated degradation

Based on the JFK–RUNX2 interaction in the SCF complex, we further examined whether JFK destabilizes RUNX2 via an SCF-dependent mechanism. Western blotting analysis revealed that JFK dose-dependently decreased the protein level of RUNX2 but not that of RUNX1 or RUNX3, and that this decrease was effectively blocked by MG132 (Fig. 3a). Moreover, western blotting analysis of U2OS cells treated with specific siRNAs targeting the catalytically essential SCF subunits (e.g., CUL1 and SKP1), revealed that JFK overexpression failed to alter the protein level of RUNX2 in CUL1- or SKP1-depleted cells (Fig. 3b, left). These results suggest that RUNX2 degradation is promoted by JFK in an SCF-dependent manner, which is distinct from the E3 ligases previously shown to target RUNX2.^7–10 Moreover, real-time reverse transcription–PCR (qPCR) showed that JFK overexpression did not affect the mRNA level of RUNX2 (Fig. 3b, right), arguing against a dysfunctional transcriptional regulation as the cause of the observed decrease in the level of RUNX2 upon JFK overexpression. Furthermore, CHX chase assays in HDFCs and CDFCs transduced with lentivirally delivered control shRNA or JFK shRNA revealed that JFK knockdown led to a significant increase in RUNX2 half-life (Fig. 3c and Supplementary Fig. 2a). These data suggest that RUNX2 degradation is promoted by JFK via the SCF-dependent ubiquitin–proteasome pathway, and that disruption of JFK function prevents the expedited proteasomal degradation of the CCD-associated RUNX2 mutant.

Fig. 3.

SCF^JFK is an E3 ubiquitin ligase for RUNX2 and CCD-associated RUNX2^R186T accelerates its SCF^JFK-mediated degradation. a U2OS cells were transfected with increasing amounts of FLAG-JFK and exposed to DMSO or MG132. b U2OS cells treated with the indicated siRNAs, followed by transfection with vector or FLAG-JFK. Cellular extracts were prepared for western blotting (left), and total RNA was extracted for qPCR (right). Data are presented as the mean ± SD (n = 3). c HDFCs or CDFCs transduced with lentivirally delivered control shRNA or JFK shRNA were exposed to CHX for the indicated times. Cellular extracts were prepared for western blotting. Quantification was performed by densitometry, and the data are expressed as the RUNX2/β-actin ratio. Data are presented as the mean ± SD (n = 3). The asterisk (*) and octothorp (#) represent significant differences from the HDFC and CDFC control group, respectively. d U2OS cells were co-transfected with the indicated plasmids and exposed to MG132 before performing in vivo ubiquitination assays. e Purified recombinant GST-JFK, GST-JFK^ΔF-box, or GST-JFK^ΔKelch was incubated with in vitro transcribed/translated RUNX2 for in vitro ubiquitination assays. f HDFCs or CDFCs transduced with lentivirally delivered control shRNA or JFK shRNA were exposed to MG132. Cell lysates were then immunoprecipitated with anti-RUNX2 and immunoblotted with anti-Ub

To further elucidate the receptor–substrate relationship and the ubiquitin–proteasome-dependent mechanism of RUNX2 degradation mediated by JFK, we investigated whether the observed JFK‑induced decrease in RUNX2 stability results from RUNX2 ubiquitination. We co-transfected U2OS cells with FLAG-JFK, FLAG-JFK^ΔF-box, or FLAG-JFK^ΔKelch, along with HA-Ub or its mutational variant HA-UbK48R, which is defective in polyubiquitin chain assembly.³² Immunoprecipitation with anti-RUNX2 showed that wild-type JFK, but not the F-box or Kelch deletion mutant, promoted RUNX2 polyubiquitination (Fig. 3d). In addition, JFK-induced polyubiquitination of RUNX2 was not observed when wild-type ubiquitin was substituted with UbK48R (Fig. 3d). Moreover, in vitro ubiquitination assays revealed that wild-type JFK, but not the F-box or Kelch deletion mutant, enhanced RUNX2 polyubiquitination (Fig. 3e). Importantly, in vivo ubiquitination assays in HDFCs and CDFCs revealed that JFK knockdown led to reduced RUNX2 polyubiquitination in CDFCs to the extent comparable to wild-type RUNX2 polyubiquitination in HDFCs (Fig. 3f). Collectively, these results establish SCF^JFK as an E3 ubiquitin ligase that destabilizes RUNX2 through polyubiquitination and support the argument that the CCD-associated mutation of RUNX2 accelerates its proteasomal degradation by SCF^JFK.

Jfk ablation expedites embryonic osteogenesis and mitigates Runx2 haploinsufficiency-induced CCD-like syndrome in mice

Runx2 germline deletion is lethal owing to impaired formation of a mature mineralized skeleton,³³ while its haploinsufficiency in mice recapitulates the phenotypic features of CCD.⁹ To gain further support for the physiological relevance of SCF^JFK-mediated RUNX2 degradation, we determined whether disruption of the JFK-RUNX2 axis affects embryonic development and the development of CCD. To this end, we generated Jfk knockout (Jfk^−/−) mice via embryonic stem cells targeting in combination with the Cre/LoxP recombination system.³⁴ As reported in our previous study,³⁵ Jfk^−/− mice were born at Mendelian frequencies and showed no overt anatomical abnormalities but consistently exhibited a lean body phenotype irrespective of sex.³⁵

Jfk^+/+/Runx2^+/− and Jfk^−/−/Runx2^+/− mice were subsequently generated by crossing heterozygous Jfk knockout (Jfk^+/−) mice with heterozygous Runx2 knockout (Runx2^+/−) mice. Alizarin red/Alcian blue staining showed that compared with wild-type littermates, Jfk ablation led to accelerated closure of the fontanelle, suggesting that Jfk ablation expedites membranous ossification of the skull (Fig. 4a, b). Additionally, compared to wild-type littermates, Jfk^+/+/Runx2^+/− mouse pups had larger defect areas in the skull and shorter clavicle and scapula lengths, suggesting that these animals experienced delayed calvarium formation, aplastic clavicle, and hypoplastic scapula, phenotypic manifestations of the CCD syndrome that were partially ameliorated by Jfk knockout (Fig. 4a, b). Moreover, microquantitative computed tomography (μ-QCT) analysis of femurs isolated from 6-week-old Jfk^+/+/Runx2^+/+, Jfk^−/−/Runx2^+/+, Jfk^+/+/Runx2^+/−, and Jfk^−/−/Runx2^+/− mice revealed that compared to those of wild-type mice, the bone mass of Jfk^−/− mice was significantly increased (Fig. 4c, d). Further analysis of Jfk^−/− mice revealed that the trabecular number (Tb.N), cortical bone thickness (Ct.Th), bone mineral density (BMD), trabecular thickness (Tb.Th), trabecular bone volume (BV/TV), cortical bone volume (Ct.BV) and tissue volume (Ct.TV) were significantly increased, whereas the trabecular separation (Tb.Sp) was significantly decreased (Fig. 4c, d). Compared with wild-type mice, Jfk^+/+/Runx2^+/− mice exhibited significantly decreased bone mass, whereas such decreases were not evident in Jfk^−/−/Runx2^+/− mice (Fig. 4c, d). Notably, calcein labeling of femurs revealed that Jfk deficiency increased the mineral apposition rate (MAR), whereas Runx2 haploinsufficiency was associated with a reduced MAR (Fig. 4e). The impaired MAR associated with Runx2 haploinsufficiency was partially rescued by concurrent Jfk ablation (Fig. 4e). In concordance, western blotting analysis of protein extracts from femoral bones revealed that Jfk knockout resulted in increased Runx2 protein level (Fig. 4f). Importantly, the Runx2 protein level was greater in the Jfk^−/−/Runx2^+/− group than in the Jfk^+/+/Runx2^+/− group, supporting our notion that genetic ablation of Jfk alleviates the JFK-mediated proteasomal degradation of RUNX2 (Fig. 4f). These data support JFK-mediated RUNX2 degradation, suggesting that disruption of Jfk function could counteract Runx2 haploinsufficiency-induced CCD-like syndrome.

Fig. 4.

Jfk ablation expedites embryonic osteogenesis and mitigates Runx2 haploinsufficiency-induced CCD-like syndrome in mice. a Alizarin red/Alcian blue staining of the whole-mount skeletons, skull, clavicle, scapula, and forelimb of Jfk^+/+/Runx2^+/+, Jfk^−/−/Runx2^+/+, Jfk^+/+/Runx2^+/−, and Jfk^−/−/Runx2^+/− newborn pups. Scale bar = 5 mm. b Quantification of body length, skull defect area, clavicle length, and scapula length in Jfk^+/+/Runx2^+/+, Jfk^−/−/Runx2^+/+, Jfk^+/+/Runx2^+/−, and Jfk^−/−/Runx2^+/− newborn pups. Data were analyzed via two-way ANOVA with Sidak’s post hoc test and are presented as the mean ± SD (n = 6). The asterisk (*), octothorp (#), and dollar ($) represent significant differences from the Jfk^+/+/Runx2^+/+ group, Jfk^−/−/Runx2^+/+ group, and Jfk^+/+/Runx2^+/− group, respectively. No statistically significant interaction was detected. c Representative 3D μ-QCT images of the trabecular bone and cortical bone of femurs from 6-week-old Jfk^+/+/Runx2^+/+, Jfk^−/−/Runx2^+/+, Jfk^+/+/Runx2^+/−, and Jfk^−/−/Runx2^+/− mice. d 3D μ-QCT analysis of the distal femurs from the indicated groups for Tb.N, Ct.Th, BMD, Tb.Th, BV/TV, Tb.Sp, Ct.BV, and Ct.TV. Data are presented as the mean ± SD (n = 6). No statistically significant interaction was detected. e Representative calcein-labeled images of femurs from the indicated groups. Quantification of the MAR is shown. Data are presented as the mean ± S.D. (n = 6). Scale bar = 20 μm. No statistically significant interaction was detected. f Total proteins from bone tissues of the indicated groups were prepared for western blotting analysis

JFK-mediated RUNX2 degradation suppresses osteoblast differentiation

As RUNX2 is critically involved in osteoblast differentiation and early tooth development,³⁶ to elucidate the mechanistic involvement of SCF^JFK–RUNX2 in bone development and CCD pathogenesis, we next investigated the functional effects of the JFK-RUNX2 axis on osteoblast differentiation. To this end, bone marrow mesenchymal stem cells (BMSCs) isolated from the femurs and tibiae of 4-week-old wild-type mice and mouse pre-osteoblastic MC3T3-E1 (subclone 14) cells were induced for osteoblast differentiation.^37,38 Time-course western blotting analysis revealed that Jfk exhibited an opposite pattern of change compared to Runx2, decreasing initially as Runx2 protein level increased (day 7, day 14, and day 21) and then rising as Runx2 protein level declined (day 28) in BMSCs undergoing osteoblast differentiation, as determined by Alizarin red S (ARS) staining (Fig. 5a).³⁹ These results are consistent with those of JFK-mediated RUNX2 degradation and highlight the critical role of the JFK-RUNX2 axis during osteoblast differentiation.

Fig. 5.

JFK-mediated RUNX2 degradation suppresses osteoblast differentiation. a BMSCs from wild-type mice were induced for osteoblast differentiation for the indicated times prior to ARS staining (left) or western blotting analysis (right). b MC3T3-E1 cells transduced with lentivirally delivered plasmids were induced for osteoblast differentiation for 7 days before ALP staining (upper) or for 28 days prior to ARS staining (lower). c ALP or ARS staining of ex vivo cultured BMSCs isolated from the femurs and tibias of 4-week-old wild-type or Jfk^−/− mice after induction for osteoblast differentiation for 7 days or 21 days. d BMSCs from wild-type or Jfk^−/− mice were induced for osteoblast differentiation for 7 or 21 days. Total RNA or protein was prepared for qPCR measurement (upper) or western blotting analysis (lower). Data are presented as the mean ± SD (n = 3)

To further confirm the involvement of the JFK-RUNX2 axis in osteoblast differentiation, MC3T3-E1 cells were transduced with lentivirally delivered vector/control shRNA, FLAG-Jfk, FLAG-Jfk^ΔF-box, FLAG-Jfk^ΔKelch, Jfk shRNA, and/or Runx2 shRNA, and osteoblast differentiation was induced. ALP staining at 7 days of osteogenic induction served as an early marker of osteogenesis,⁴⁰ while ARS staining at 28 days of induction was used to detect calcium deposition as a late marker for osteogenesis. These data showed that wild-type Jfk, but not the F-box or Kelch deletion mutant, resulted in significantly reduced ALP activity and calcium deposition (Fig. 5b), whereas Jfk knockdown accelerated osteoblast differentiation and matrix mineralization, which were at least partially abrogated upon simultaneous knockdown of Runx2 (Fig. 5b and Supplementary Fig. 2b).

To further investigate the role of the JFK-RUNX2 axis in osteoblast differentiation, BMSCs were isolated from wild-type or Jfk^−/− mice. ALP and ARS staining revealed that Jfk knockout promoted osteoblast differentiation and bone matrix formation, as demonstrated by increased ALP activity and calcium deposition (Fig. 5c). Consistently, the expression levels of Runx2 as well as other osteogenesis markers including Sp7, Alp, Col1a2, and Bglap, increased in BMSCs from Jfk^−/− mice at day 0, day 7, and day 21 of osteogenic induction, as measured by qPCR and western blotting (Fig. 5d, Supplementary Fig. 2c, d and Supplementary Table 1). These results support that JFK-mediated degradation of RUNX2 suppresses osteoblast differentiation and matrix mineralization.

JFK is transcriptionally repressed by RUNX2

RUNX2 contributes to bone formation by transcriptionally regulating a diverse set of osteoblast-specific target genes, including the transcriptional activation of SPP1 and COL1A2 and the transcriptional repression of p21 and NR2F2.^41,42 To further explore the mechanistic involvement of the JFK-RUNX2 axis in osteoblast differentiation, we performed CUT&Tag assays using antibodies against Runx2 in MC3T3-E1 cells. After alignment using Bowtie2, MACS2 analysis yielded a total of 54,140 Runx2-specific binding peaks. Peak distribution analysis revealed that 39.23% mapped to promoters (≤ 3 kb), 28.29% to introns, and 28.09% to intergenic regions (Fig. 6a). The genomic landscape of Runx2 occupancy at promoter sequences was analyzed by deepTools (Fig. 6b). These analyses revealed that Runx2 was enriched at the promoters of Col1a2, Spp1, and p21 (Fig. 6c), which are known as osteoblast-associated genes.^41,43 Notably, CUT&Tag revealed that Runx2 was also enriched at the Jfk promoter (Fig. 6c), indicating that Runx2 transcriptionally regulates Jfk.

Fig. 6.

Genome-wide analysis of RUNX2 target genes. a CUT&Tag in MC3T3-E1 cells with Runx2 antibodies. The genomic distribution of Runx2 occupancy was analyzed using MACS2. b DeepTools was used to visualize the binding density of Runx2. The heatmap shows the CUT&Tag tag counts on Runx2 binding peaks in MC3T3-E1 cells (n = 2). c CUT&Tag tracks at the indicated target gene promoters in MC3T3-E1 cells. The peak regions corresponding to Runx2 binding sites are highlighted by red rectangles. d Genomic distribution of Runx2 determined by ChIP-seq (GSE41920) in MC3T3-E1 cells was analyzed using MACS2. e DeepTools was used to visualize the binding density of Runx2. The heatmap shows the ChIP-seq tag counts on Runx2 binding peaks in MC3T3-E1 cells. f ChIP-seq tracks at the indicated target gene promoters in MC3T3-E1 cells. The peak regions corresponding to Runx2 binding sites are highlighted by red rectangles

To support this notion, we analyzed the genome-wide transcription targets of Runx2 in MC3T3-E1 cells on day 15 of osteogenic induction using published chromatin immunoprecipitation-based deep sequencing (ChIP-seq) datasets (GSE41920).⁴⁴ 56,110 Runx2-specific binding peaks were identified by MACS2 analysis. Peak distribution analysis revealed that 26.47% mapped to promoters (≤ 3 kb), 33.78% to introns, and 35.36% to intergenic regions (Fig. 6d). The genomic occupancy of Runx2 at promoter sequences was analyzed by deepTools (Fig. 6e). Consistent with our CUT&Tag results, these analyses also revealed that Runx2 was enriched at the Jfk promoter, in addition to its occupancy at the promoters of Col1a2, Spp1, and p21 (Fig. 6f). To confirm the occupancy of Runx2 at the promoter of Jfk, we conducted quantitative ChIP (qChIP) analysis in MC3T3-E1 cells. Indeed, we detected strong occupancy of Runx2 at the promoters of Jfk, Col1a2, Spp1, and p21 (Fig. 7a). Additionally, qPCR and western blotting in MC3T3-E1 cells transduced with lentivirally delivered control shRNA or Runx2 shRNA showed that Runx2 knockdown resulted in elevated expression of Jfk and p21 but reduced expression of Col1a2 and Spp1 (Fig. 7b), suggesting that Jfk is transcriptionally repressed by Runx2.

Fig. 7.

JFK is transcriptionally repressed by RUNX2. a Soluble chromatin from MC3T3-E1 cells was prepared for qChIP analysis of Runx2 occupancy at the indicated gene promoters. Data were analyzed via two-way ANOVA with Sidak’s post hoc test and are presented as the mean ± SD (n = 3). b Total RNA or protein from MC3T3-E1 cells transduced with lentivirally delivered control or Runx2 shRNA was prepared for qPCR or western blotting analysis, respectively. Data are presented as the mean ± SD (n = 3). c Soluble chromatin from HDFCs or CDFCs was prepared for qChIP analysis of RUNX2 occupancy at the indicated gene promoters. Data are presented as the mean ± SD (n = 3). d Total RNAs and proteins from HDFCs and CDFCs were prepared for qPCR or western blotting, respectively. Data are presented as the mean ± SD (n = 3). e Schematic diagram illustrating the RUNX2 binding motif within the JFK promoter region is shown. U2OS cells transfected with the indicated plasmids were assayed for firefly and renilla luciferase activity. Data are presented as the mean ± SD (n = 3). f U2OS cells treated with RUNX2 3’-UTR siRNAs and subsequently transfected with the indicated plasmids were assayed for firefly and renilla luciferase activity. Data were analyzed via one-way ANOVA and are presented as the mean ± SD (n = 3). The asterisk (*) and octothorp (#) indicate significant differences from the vector group and the RUNX2^wt group, respectively. g Total RNA and protein from U2OS cells treated with RUNX2 3’-UTR siRNAs and subsequently transfected with the indicated Myc-tagged expression constructs was prepared for qPCR and western blotting, respectively. Data were analyzed via one-way ANOVA and are presented as the mean ± SD (n = 3)

To examine the impact of the CCD-associated RUNX2^R186T on the transcription regulation of JFK, we performed qChIP analysis in the aforementioned HDFCs and CDFCs. The results showed that compared to that in HDFCs, the enrichment of RUNX2 at the promoters of JFK, COL1A2, and SPP1 in CDFCs was significantly reduced (Fig. 7c). In addition, compared with those in HDFCs, the expression of JFK in CDFCs was upregulated, and the expression of COL1A2 and SPP1 was downregulated (Fig. 7d), indicating that CCD-associated RUNX2^R186T impairs RUNX2‑mediated transcriptional repression of JFK.

To further substantiate the transcriptional repression of JFK by RUNX2, we cloned an ~800 bp fragment (chr1: 16,353,481‒16,354,281) of the JFK promoter containing a RUNX2-binding sequence, TGTGGT,⁴⁵ and generated a JFK promoter-driven luciferase reporter (JFK^wt-Luc), as well as a luciferase reporter driven by the JFK promoter lacking the canonical RUNX2-binding sequence (JFK^mu-Luc), to test the transcriptional activity of RUNX2^wt and RUNX2^mu (R131G, M175R, R193Q, Q209H, and A362V mutants) on the JFK promoter-driven luciferase reporter. Luciferase reporter assays showed that RUNX2^wt, but not RUNX2^mu, was able to repress the JFK promoter-driven luciferase activity, an effect dependent on the RUNX2 binding sequence (Fig. 7e, f).

Consistently, qPCR and western blotting analyses also showed that JFK expression was decreased in wild-type RUNX2-overexpressing cells, an effect that was not evident when R131G, M175R, R193Q, Q209H, or A362V mutant of RUNX2 was overexpressed (Fig. 7g), which is consistent with the DNA-binding dysfunction of these CCD-associated RUNX2 mutants.²⁰ Together, these data support that RUNX2 binds to JFK promoter and functions as a transcriptional repressor of JFK.

Discussion

Various clinical manifestations of heritable skeletal diseases ultimately reflect the basic mechanisms underlying bone and cartilage formation as well as skeletal development and bone homeostasis. CCD is an autosomal-dominant inheritance disorder,^2,46,47 with genetic mutations of RUNX2, a master regulator of osteogenesis, occurring in approximately 65% of CCD patients.^1,3,4 Although it is thought that CCD-associated RUNX2 mutations result in impaired transcriptional activity for its target genes, some mutations still retain partial transcriptional capability.^{2,27,29,48,49} Thus, the mechanism(s) through which these mutations contribute to CCD pathogenesis require further investigation. Our current study demonstrated that RUNX2 is subjected to proteasome-mediated degradation by SCF^JFK. Importantly, our proteomic analyses revealed that the CCD-associated RUNX2^R186T mutant is more susceptible to being targeted by the SCF^JFK E3 ligase, pointing to a role for JFK-directed RUNX2 degradation in the pathogenesis of CCD. Future high-resolution structural studies, such as cryo-electron microscopy, would elucidate and compare the three-dimensional architectures of the SCF^JFK complex in the context of wild-type versus mutated RUNX2.

Several ubiquitin ligases, including WWP1,⁷ WWP2,⁸ SMURF1/2,⁹ and CHIP,¹⁰ have been shown to target RUNX2 for polyubiquitination. However, whether and how these E3 ligases are involved in the pathogenesis of CCD and whether and how mutations in RUNX2 affect its stability have yet to be determined. Our results in the present study indicate that SCF^JFK functions as an E3 ligase for RUNX2. Importantly, we showed that Jfk knockout is associated with increased Runx2 levels and partially rescues aberrant embryonic bone development due to Runx2 haploinsufficiency, whereas Runx2 haploinsufficiency increases Jfk expression, which is consistent with the transcriptional repression of JFK by RUNX2. Intriguingly, we showed that while compared to wild-type RUNX2, RUNX2 R186T mutant acquires a higher affinity toward JFK, it does not affect the interaction of RUNX2 with other E3 ligases, including WWP1 and SMURF2. Nevertheless, systematically characterizing how different ubiquitin ligases function in promoting RUNX2 degradation in a time- and/or cell context-dependent manner should be informative. Moreover, targeted substrate profiling of JFK and other ubiquitin ligases (WWP2, WWP1, CHIP, and SMURF1/2) could elucidate how diverse biological processes (e.g., cartilage homeostasis, chronic inflammation, and/or bone metabolism) collectively shape RUNX2’s functional specification in cell fate determination and elucidate proteasome dynamics during bone development. Such investigations would likely also reveal the upstream events that impact RUNX2 stability.

Our finding that JFK negatively affects RUNX2 stability, thereby regulating osteoblast differentiation in BMSCs as well as in MC3T3-E1 cells highlights the important role of the JFK-RUNX2 axis in osteoblast differentiation. Given that MSCs have immunosuppressive characteristics, can home in to injury sites, and can differentiate into various cell types, they are now understood as valuable tools in regenerative medicine, for example, in the treatment of osteoporosis and other skeletal disorders.⁵⁰ Our current study, in conjunction with our recently reported finding that JFK promotes excessive hepatic lipid accumulation and links to obesity and metabolic syndrome,³⁵ highlights the possibility that disruption of JFK or the SCF complex components in MSCs could be a potential therapeutic strategy to ameliorate CCD-like syndrome and protect against obesity. Consistent with these findings, the imbalanced lineage commitment of MSCs, including a shift in commitment to adipogenesis, has been associated with aging, immobility, and an increased need for corticosteroid therapy in osteoporosis patients in the clinic.⁵¹ Thus, perhaps therapeutic modulation to maintain a relatively low JFK level in BMSCs could be pursued to protect against the pathogenesis of osteoporosis by extending the normal osteoblast differentiation capacity of at-risk individuals (e.g., professional athletes and geriatrics). Given that the function of JFK may not be strictly or linearly correlated with its own expression level but rather with the integrity and availability of the entire SCF^JFK complex, dissection of JFK function in MSC proliferation and, subsequently, during distinct MSC developmental lineages (i.e., osteoblasts or adipocytes) might support the definition of JFK as a “molecular rheostat” that modulates the osteoblast–adipocyte differentiation balance. Additionally, future studies using skeletal tissue-specific conditional knockout models would be desirable to more precisely elucidate the autonomous role of JFK in osteogenesis.

Importantly, beyond the role of SCF^JFK in promoting the degradation of RUNX2 in osteoblast differentiation and CCD pathogenesis, we discovered that wild-type RUNX2 represses JFK transcription. Consistent with the impaired RUNX2-mediated osteoblast differentiation drives CCD, we found that some CCD-associated mutations in RUNX2 resulted in impaired transcriptional repression of JFK. Together, these results support the notion that RUNX2 transcriptionally represses JFK through binding to its promoter, suggesting the presence of a regulatory feedback loop in which JFK mediates the degradation of RUNX2, an activity that is blocked by RUNX2 itself through transcriptional repression of JFK during normal osteoblast differentiation (Fig. 8). In CCD syndrome, RUNX2 mutations render RUNX2 more susceptible to SCF^JFK-directed decay and/or impaired transcriptional repression of JFK, thereby tipping the balance toward RUNX2 destabilization. Thus, disruption of this feedback loop—which normally prevents RUNX2 degradation through SCF^JFK during normal osteoblast differentiation and skeletal development—also drives the pathogenesis of CCD. It is plausible that disruption of this feedback loop per se may help to explain the pathogenic influence of other mutations known for CCD,² or perhaps even explain pathogenic processes in other skeletal development disorders. Our study’s characterization of this loop therefore opens the door for rapid screening assays to better characterize the biochemical basis of pathogenic RUNX2 mutations. Our illustration of the multilayered interaction between SCF^JFK and RUNX2 degradation through the ubiquitin‒proteasome pathway and the direct repression of JFK transcription support that JFK functions as a critical node in the regulatory hierarchy of osteoblast differentiation.

Fig. 8.

Proposed model of the SCF^JFK-RUNX2 feedback loop in osteoblast differentiation. SCF^JFK is an E3 ligase that targets RUNX2 for degradation, and RUNX2, in turn, transcriptionally represses JFK, establishing a feedback loop to control osteoblast differentiation and bone development. The SCF^JFK-RUNX2 feedback loop regulates normal osteoblast differentiation and skeletal development by balancing the activity of RUNX2 and JFK, whereas the CCD-associated mutation leads to increased susceptibility of RUNX2 to SCF^JFK-mediated degradation and/or impaired transcriptional repression of JFK, thereby driving the pathogenesis of CCD. A schematic diagram of the proposed model was manually illustrated in Adobe Photoshop

In summary, our study identifies SCF^JFK as an E3 ligase for RUNX2. We found that RUNX2 transcriptionally represses JFK. We showed that CCD-associated RUNX2 mutations are more susceptible to SCF^JFK-targeted degradation and/or suffer from impaired transcriptional repression of JFK. Our results support the existence of a feedback loop regulating osteoblast differentiation and bone development, offering mechanistic understanding of how SCF^JFK-RUNX2 loop imbalance drives the pathogenesis of CCD. Beyond understanding the pathophysiological significance of RUNX2 stability, our work supports targeting JFK as a potential therapeutic strategy for skeletal disorders.

Materials and methods

Antibodies, reagents, and plasmids

The antibodies used were anti-FLAG (Sigma); anti-HA (MBL); anti-RUNX2, anti-ubiquitin, anti-CUL1, anti-SKP1, anti-RBX1, anti-RUNX1, anti-RUNX3, anti-COL1A2, anti-SP7, anti-ALP, anti-p21, and anti-β-actin (Santa Cruz); anti-Myc-Tag (Cell Signaling Technology); and anti-BGLAP, anti-RUNX2, and anti-JFK (ABclonal). Protein A/G Sepharose CL-4B beads (Amersham Biosciences), the protease inhibitor cocktail (Roche).

Wild‑type and deletion mutant JFK cDNAs were PCR‑amplified and ligated into the pcDNA3.1 vector, which contains FLAG. Wild‑type RUNX2 cDNAs were PCR‑amplified and ligated into the pcDNA3.1 vector that contained Myc or His. A QuikChange Multi-Site Directed Mutagenesis Kit (Agilent) was used to generate the RUNX2 mutants. GST-tagged wild-type or deletion mutants of JFK and wild-type or mutant RUNX2 expression plasmids were produced by inserting the respective coding sequences into the pGEX-4T-3 vector.

Gene knockout mice

All mice were kept under specific‑pathogen‑free conditions. Jfk was targeted for homologous recombination in C57BL/6 embryonic stem cells by introducing LoxP sites flanking exon 5.⁵² Jfk-deficient mice were generated as previously described.³⁵ Genotyping analyses of Jfk-deficient mice were carried out via PCR on tail‑tip genomic DNA, and the primers used were 5′-AGTGCTCTTAACTGCCTAGTA-3′ (forward) and 5′-GGGCGTGTCTTAATGTCCTAAA-3′ (reverse). Runx2^+/− mice were kindly provided by Prof. Weiguo Zou (University of Chinese Academy of Sciences).⁵³ Genotyping analyses of Runx2^+/− mice were carried out via PCR on tail‑tip genomic DNA, and the primers used were 5′-CACGGAGCACAGGAAGTTGGG-3′ (forward) and 5′-AAGATGGATTGCACGCAGGTTCTC-3′ (reverse) and 5′-CACGGAGCACAGGAAGTTGGG-3′ (forward) and 5′-TGAGCGACGTGAGCCCGGTGG-3′. Animal handling and procedures were approved by the Institutional Animal Care Committee of Peking University Health Center (LA2022074).

Cell culture

Cells were cultured as recommended by the ATCC. U2OS and DFC cells were maintained in DMEM supplemented with 10% fetal bovine serum (FBS). MC3T3-E1 cells and BMSCs were maintained in α-MEM supplemented with 10% FBS. The cells were maintained in a humidified 5% CO₂ incubator at 37 °C. The human sample-related studies were approved by the Ethics Committee of Peking University School and Hospital of Stomatology (PKUSSIRB-2012004).

Primary DFCs collected from healthy donor (HDFCs, RUNX2^+/+) and CCD patient (CDFCs, RUNX2^+/mu) were kindly provided by Prof. Shuguo Zheng (Peking University). The isolation of DFCs followed established protocols.²² In brief, dental follicle tissues were collected during the surgical extraction of impacted maxillary incisors. The tissues were digested with 0.3% collagenase type I (Sigma‒Aldrich) (40 min, 37 °C) to obtain a single-cell suspension. Isolated cells were cultured in DMEM plus 10% FBS and penicillin-streptomycin (Gibco), at 37 °C in a humidified 5% CO₂ incubator.

Co-immunoprecipitation

Co-immunoprecipitation was carried out according to the established procedure.¹⁶ In brief, we incubated cells for 30 min in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 0.3% NP-40, 2 mM EDTA, pH 7.5) containing protease inhibitor cocktail, then centrifuged the mixture (14,000 × g, 5 min, 4 °C). Proteins were immunoprecipitated with control or specific anti-RUNX2 antibody (Santa Cruz; sc-101145) at 4 °C for 16 h with rotation. After adding 60 µL of 50% protein G agarose beads for 2 h, the beads were washed five times with lysis buffer, then pelleted by centrifugation (500 × g, 5 min, 4 °C). Precipitates were eluted in 2 × SDS-PAGE loading buffer by boiling for 10 min, separated by SDS-PAGE, and immunoblotted with appropriate antibodies (anti-CUL1, Santa Cruz, sc-17775; anti-SKP1, Santa Cruz, sc-5281; anti-RBX1, Santa Cruz, sc-393640).

GST pull-down assay

GST fusion constructs were expressed in BL21 E. coli bacteria, and crude bacterial lysates were prepared by sonication in lysis buffer (50 mM Tris-HCl, 1.5 mM EDTA, 1 mM dithiothreitol, 10% (v/v) glycerol, and 0.4 M NaCl, pH 7.4) containing a protease inhibitor cocktail. In vitro transcription/translation experiments were carried out with rabbit reticulocyte lysate (TNT systems, Promega) following the manufacturer’s protocol. Briefly, the GST fusion protein was immobilized on 50 μl of 50% glutathione-Sepharose 4B slurry beads in GST pull-down binding buffer (10 mM HEPES, 3 mM MgCl₂, 100 mM KCl, 5 mM EDTA, 5% glycerol, and 0.5% CA630, pH 7.6). We incubated the beads with rotation (1 h, 4 °C), then washed them three times with GST pull-down binding buffer and resuspended them in the same buffer. Next, we added the in vitro transcribed/translated protein and continued incubation with rotation (2 h, 4 °C). After three additional washes with binding buffer, the bound proteins were eluted by boiling in 30 µL of 2 × sample loading buffer and separated by SDS-PAGE.

Isolation of mouse BMSCs

The femurs and tibias from four-week-old mice sacrificed by cervical dislocation were quickly separated. BMSCs were flushed from the bone marrow cavity using a syringe, seeded into 35-mm culture dishes, and incubated at 37 °C in 5% CO₂. After 48 h, we removed non-adherent cells by washing with PBS and cultured the adherent cells in α-MEM containing 10% FBS.

Osteoblast differentiation

MC3T3-E1 cells or BMSCs were cultured in α-MEM medium containing ascorbic acid (50 μg/ml), dexamethasone (100 nM), and β-glycerophosphate (10 mM) for 7, 14, 21, or 28 days, and the culture medium was changed every two days.

ALP assay and ARS staining

ALP activity assays were performed 7 days after osteogenic induction. Cells were fixed with 10% paraformaldehyde (10 min, 25 °C), followed by three washes with PBS. ALP activity was tested by colorimetric detection using nitro-blue tetrazolium chloride (NBT) and 5-bromo-4-chloro-3′-indolyphosphate p-toluidine salt (BCIP) as chemical substrates following the manufacturer’s protocol (Beyotime Institute of Biotechnology). Calcium deposit nodule formation was detected by ARS staining at 7, 14, 21, or 28 days after the induction of osteoblast differentiation. Briefly, cells were fixed in 10% paraformaldehyde (10 min, 25 °C), followed by three washes with PBS and staining with 1% alizarin red (Solarbio) for 15–30 min.

Alizarin red/Alcian blue staining

Jfk^+/+/Runx2^+/+, Jfk^−/−/Runx2^+/+, Jfk^+/+/Runx2^+/−, and Jfk^−/−/Runx2^+/− newborns were eviscerated, and the skin and organs were dissected away. The skeletons were fixed in 95% ethanol and then immersed in acetone for another day. Following a 2-day stain in Alizarin Red/Alcian blue solution, the skeletons were fixed and cleared three times with 95% ethanol, treated with 1% KOH for 3 days, then sequentially cleared in 1% KOH/glycerol mixtures (20%, 50%, and 90% glycerol, each for 1 day), and finally stored in 100% glycerol.⁵³

Statistical analysis

In this study, the data are reported as the mean ± SD unless otherwise noted. Comparisons between the groups were analyzed with two-tailed Student’s t tests, one-way ANOVA, or two-way ANOVA with Sidak’s post hoc test. SPSS 22.0 or GraphPad Prism (v 8.0.2) were used for statistical tests. Statistical significance for group differences is represented by *p < 0.05. In vitro assays were performed in three independent experimental replicates unless otherwise stated.

Author contributions

L.H. and W.S.L. conceived and directed the project; Z.S.S. and Y.Z. performed the experiments and data analyses; Y.L., X.Y.Y., J.C.W., D.W.F., D.Z., D.Z.T., F.P., H.S.Z, S.G.Z., and X.J.T. offered technical assistance.

Data availability

CUT&Tag data are accessible via GEO (NCBI) under accession number GSE318880. All data supporting the study’s conclusions are provided in the article and/or Supplementary Materials. Additional supporting data are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41392-026-02837-8.