Abstract

   Polycomb repressive complexes (PRCs) are pivotal epigenetic regulators
   that preserve cell identity by restricting transcription responses to
   sub-threshold extracellular signals. Their roles in osteoblast function
   and bone formation remain unclear. Here in aging osteoblasts, we found
   marked activation of PRC1.1 complex, with KDM2B acting as a
   chromatin-binding factor and BCOR and PCGF1 enabling histone H2A
   monoubiquitylation (H2AK119ub1). Osteoblast-specific Kdm2b inactivation
   significantly enhances bone remodeling under steady-state conditions
   and in scenarios of bone loss. This enhancement is attributed to
   H2AK119ub1 downregulation and subsequent Wnt signaling derepression.
   Furthermore, we developed a small molecule termed iBP, that
   specifically inhibits the interaction between BCOR and PCGF1, thereby
   suppressing PRC1.1 activity. Notably, iBP administration promotes bone
   formation in mouse models of bone loss. Therefore, our findings
   identify PRC1.1 as a critical epigenetic brake on bone formation and
   demonstrate that therapeutic targeting of this complex enhances Wnt
   pathway activation, offering a promising strategy against skeletal
   deterioration.

   Subject terms: Diseases, Drug discovery, Histone post-translational
   modifications, Musculoskeletal development
     __________________________________________________________________

   Here they identify a selective PRC1.1 inhibitor, iBP, that treats bone
   loss by improving osteoblast function. iBP releases repression of
   Wnt/β-catenin signaling, enhancing bone formation and repair in aging,
   trauma, and disease models.

Introduction

   The adult skeleton undergoes continuous bone remodeling, involving the
   resorption of mineralized bone by osteoclasts (OCs) and the subsequent
   formation of bone matrix by osteoblasts (OBs), which then become
   mineralized. This dynamic process is vital for maintaining skeletal
   integrity and ensuring mineral homeostasis. An imbalance between OBs
   and OCs leads to defective ossification, which in turn can result in
   conditions such as skeletal dysplasia or an increased risk of
   fractures^[76]1,[77]2. The process of ossification is influenced by
   various factors, including nutrition, pharmaceuticals, hormones,
   inflammation, and mechanical stress on the bones, etc. Consequently,
   disorders such as trauma, infection, rheumatoid arthritis,
   postmenopausal osteoporosis and disuse osteoporosis often present with
   bone loss. The deterioration of skeletal integrity significantly raises
   the risk of pathological fractures and negatively impacts overall
   well-being.

   OBs, as the primary bone-forming cells, play a crucial role in bone
   formation, maintenance, and remodeling through functions such as
   synthesizing the bone matrix, mineralization, producing signaling
   molecules and interacting with OCs^[78]3,[79]4. To develop effective
   and safe therapeutics to modulate bone formation, it is crucial to
   decipher the molecular mechanisms that govern the differentiation and
   functionality of OBs.

   Over the past two decades, genetic studies in humans and mice have
   consistently highlighted Wnt signaling as a critical driver for bone
   formation and regeneration^[80]5,[81]6. Upon binding of Wnt ligands to
   their receptors, β-catenin is stabilized, translocated to the nucleus,
   and forms a complex with transcription factors such as TCF/LEF, thereby
   initiating the transcription of target genes in mesenchymal stem cells
   (MSCs) or OBs for osteogenesis^[82]7. TCF/LEF, in conjunction with
   lineage-specific factors like RUNX2 and OSTERIX/SP7, modulates OB
   differentiation and the expression of bone matrix genes^[83]8–[84]10.
   Deficiency or mutations in Wnt ligands or their corresponding FZD and
   LRP receptors lead to significant impairments in bone
   formation^[85]5,[86]6,[87]11,[88]12. Although numerous therapeutic
   strategies have been developed to target Wnt signaling
   components^[89]5,[90]6, their efficacy is contingent on the chromatin
   environment’s responsiveness^[91]13,[92]14. The epigenetic mechanisms
   that govern the transcriptional response to Wnt/β-catenin signaling and
   OB functions are not yet fully understood.

   Among epigenetic repressors, Polycomb group (PcG) proteins have been
   demonstrated to restrict transcription response to extracellular
   cues^[93]15 including Wnt signaling^[94]16–[95]19. Through forming two
   biochemically distinct Polycomb Repressive Complexes (PRC1 and PRC2),
   they maintain a close chromatin environment through distinct
   activities. PRC2 catalyzes the methylation of histone H3 at lysine 27
   (H3K27me), while canonical or non-canonical PRC1 complexes compact
   chromatin or catalyze monoubiquitylation of histone H2A at lysine 119
   (H2AK119ub1) respectively^[96]20,[97]21. PRC1 complexes share RING1
   proteins (RING1A and RING1B) and are categorized into PRC1.1-PRC1.6
   according to the composition of PcG ring-finger domain proteins
   (PCGF1-PCGF6). Canonical complexes PRC1.2 and PRC1.4 contribute
   minimally to H2AK119ub1 but play significant roles in chromatin
   compaction^[98]22,[99]23. Non-canonical PRC1 bind chromatin through
   their unique DNA binding factors. In PRC1.1, FBXL10/KDM2B is
   responsible for recruiting the complex to CpG-rich promoters through
   its -CxxC domain^[100]24,[101]25, with BCOR and PCGF1 being essential
   accessory factors for H2AK119ub1 activity^[102]26,[103]27. In PRC1.6,
   MAX/MGA contribute to chromatin binding while PCGF6 plays diverse roles
   in gene silencing or activation^[104]28–[105]30. The specific
   expression patterns and functions of these PRCs in OBs remain largely
   unknown.

   In this study, we observed an excessive activation of PRC1.1 complex in
   aging bones. Through generation of OB-specific KDM2B-deficient mouse
   models, we found that KDM2B inactivation leads to a significant
   increase of bone mass under physiological or ossification-defective
   conditions through enhanced bone remodeling. These osteogenic effects
   are dependent on the activation of Wnt signaling, associated with a
   decrease of H2AK119ub1 deposition at the Wnt/β-catenin target genes.
   Subsequently, we identified a compound specifically targeting PRC1.1
   that limits H2AK119ub1 activity and activates Wnt signaling pathway.
   Furthermore, this compound promotes bone formation in vivo, presenting
   a promising epigenetic therapy for disorders characterized by bone
   deterioration.

Results

Gain of PRC1.1 function is linked to defective ossification during aging

   Aging is a prevalent risk factor for bone loss, characterized by a
   continuous decline in OB activity. By comparing femurs from aged (18
   months old) and young (3 months old) mice, we observed impaired
   ossification in the aged group, as evidenced by reduced trabecular bone
   revealed through micro-computed tomography (μCT) analysis
   (Fig. [106]1a).

Fig. 1. An aberrant gain of PRC1.1 function is linked to impaired
ossification.

   [107]Fig. 1
   [108]Open in a new tab

   a Representative images of scanning sections of mouse femurs at ages of
   3 and 18 months. b Volcano plot of DEGs in OBs (n = 3 YOUNG and n = 3
   OLD). Upregulated genes are labeled in red and downregulated genes in
   blue (Wald test, p < 0.05; |Fold change| >1.5). OBs were collected
   after 14 days of culture by removing the bone fragments. c Heatmap
   showing mRNA levels of osteocyte aging-associated genes between the two
   groups, colorscale: Z-Score. d GO enrichment analysis of the biological
   processes in downDEGs in aged OBs (Fisher’s exact test, p < 0.05).
   Number of genes and statistical significance are shown. e Heatmap
   showing mRNA levels of ossification-related genes between the two
   groups. Colorscale: Z-Score. f, g WB analyses comparing the levels of
   PRC members and associated histone modifications in OBs from young and
   aged mice. ACTIN and H3 were used as loading controls, respectively. h
   Heatmaps showing H2AK119ub1 signals in chromatin of YOUNG and OLD OBs
   around the peak center (± 5 kb) of H2AK119ub1 target genes. Colors
   represent CUT&Tag RPM, and rows were ranked by CUT&Tag signals in young
   group (n = 2 YOUNG and n = 2 OLD). OBs were collected after 14 days of
   culture by removing the bone fragments. i Boxplots showing
   quantification of differential H2AK119ub1 enrichment at downDEGs (OLD
   vs YOUNG, n = 386, Wilcoxon matched-pairs signed rank test, two tailed,
   p < 0.0001). The boxplots indicate the median (centre line), the third
   and first quartiles (box limits) and 1.5 × IQR above and below the box
   (whiskers). j, k The IGV view of increased H2AK119ub1 enrichment at the
   promoter of representative ossification-related genes and their
   decreased mRNA levels in aged OBs compared with young group, with Actin
   serving as a negative control. Signals representing CUT&Tag and RNA-seq
   RPM. 2 M: 2 months; 3 M:3 months; 8 M:8 months; 18 M:18 months. Source
   data of (f, g, i) are provided as a Source Data file.

   To assess changes in gene expression, we performed RNA sequencing
   (RNA-Seq) analysis on primary OBs cultured in vitro for two weeks,
   focusing on the differentially expressed genes (DEGs). In aged OBs, 201
   upregulated and 386 downregulated genes were identified relative to the
   young group (Fig. [109]1b and Supplementary Data [110]1). In line with
   these findings, the expression of numerous genes associated with aged
   OCs, as previously defined by Zhang et al.^[111]31, was altered in aged
   OBs. These alterations included downregulation of genes such as Alpl,
   Bglap, Bmp2 and upregulation of genes including Ccr1, Ccl6
   (Fig. [112]1c). Gene Oncology (GO) analysis of downregulated genes in
   aged OBs (P < 0.05) indicated a significant enrichment in ossification
   processes (Fig. [113]1d), underscoring the regulatory roles of
   identified genes in bone formation. The heatmap in Fig. [114]1e
   illustrates the downregulation of leading genes in ossification.
   Considering individual variability among mice, future studies should
   control for variables including genetic background, specific age, and
   environmental factors, to ensure reproducibility. These findings
   suggest that the reduction in bone mass is intimately linked to the
   diminished ossification capabilities of OBs. Investigating the specific
   contributions of these mouse-specific features will also be crucial for
   translating our findings to broader biological contexts.

   Recognizing the established suppressive roles of PRCs on osteogenic
   signaling during OB differentiation^[115]32,[116]33, we compared the
   protein levels of PRC components in OBs isolated from mice of varying
   ages. As expected, Western Blot (WB) analysis revealed a significant
   rise in P16 expression levels with aging, a hallmark of cell
   senescence^[117]34. For reasons that remain to be elucidated, there was
   a noticeable decrease in the protein levels of EZH2 and SUZ12, which
   are core components of PRC2. Among the principal PRC1 members, RING1B
   levels showed a slight increase with age, while the levels of
   NSPc1/PCGF1 and KDM2B^[118]25,[119]35, were significantly elevated. In
   alignment with these findings, there was a specific and substantial
   increase in H2AK119ub1 levels and a modest decrease in H3K27me3 levels
   in aging OBs. The gain of H2AK119ub1 is primarily attributed to
   elevated levels of PCGF1, as other key members of non-canonical PRC1,
   such as PCGF3 and PCGF6, show either unchanged or only minor
   downregulation in expression (Fig. [120]1f, g). These data link the
   marked activation of PRC1.1 to bone aging.

   Genome-wide profiling through Cleavage Under Targets and Tagmentation
   (CUT&Tag) analysis with proper spike-in normalization demonstrated a
   widespread increase of H2AK119ub1 signals in aged OBs compared to
   the young group (Fig. [121]1h and Supplementary Data [122]2). Among the
   386 downregulated genes in aged OBs, we observed a significant
   elevation of H2AK119ub1 enrichment (Fig. [123]1i). Ossification genes
   serve as illustrative examples while the unaffected gene Actin serves
   as a negative control (Fig. [124]1j, k). Collectively, these findings
   suggest that PRC1.1 activation raises the threshold for transcriptional
   activation by osteogenic signals, thereby suppressing the ossification
   process.

KDM2B inactivation in OBs leads to enhanced bone formation

   To investigate whether PRC1.1 limits OB functions, we managed to
   generate a mouse model of conditional ablation of PRC1.1. Accordingly,
   we created the conditional knockout allele of Kdm2b by flanking exon13
   that encodes the CxxC-domain with two loxP sequences (Supplementary
   Fig. [125]1a)^[126]18. These floxed mice Kdm2b^flox/flox were then
   crossed with the widely used OB-targeting Ocn-Cre mice to generate
   Kdm2b^fl/flOcn^Cre conditional knockout mice (referred to as CKO mice)
   (Fig. [127]2a and Supplementary Fig. [128]1b). This crossbreeding
   enabled the specific deletion of the KDM2B-CxxC domain, thereby
   abrogating its chromatin regulatory activity^[129]25,[130]36 in OBs.
   Notably, CKO mice showed no discernible differences in body size or
   weight compared to Ocn-Cre mice, referred to as control (CON) at
   various ages, including 9 and 10 weeks and between 6 to 12 months
   (Supplementary Fig. [131]1c).

Fig. 2. KDM2B inactivation in OBs leads to increased bone mass and enhanced
bone formation.

   [132]Fig. 2
   [133]Open in a new tab

   a Schematic diagram of construction of CKO mice. b Statistical analysis
   of femur length of 9-week-old mice (n = 3 per genotype). c
   Representative images of 3D reconstruction and scanning sections of
   femurs from control and CKO mice at the age of 9 weeks. d μCT analysis
   of trabecular parameters (n = 3 per genotype). e Representative images
   of H&E staining of the whole femur (left) and metaphysis (right) from
   CKO and control mice. f Representative images of femurs from control
   and CKO mice, the yellow areas represent trabecular bone and the gray
   areas represent cortical bone. g μCT analysis of cortical bone
   parameters (n = 3 per genotype). h, i Schematic diagram of the timeline
   of the calcein double labeling experiment. Representative images (h)
   and quantification of calcein double labeling parameters (i) (n = 3
   mice per genotype). j Three-point bending experiment, with blue arrows
   indicating the CKO group and yellow arrows indicating the CON group. k
   Statistical analysis of the maximum load bearing capacity of fresh
   unfixed femurs from three-point bending tests (n = 3 mice per
   genotype). Statistical significance was assessed using Student’s t
   tests, two-tailed, in (b, d, g, i, k), error bars are presented as mean
   values ± SD. Scale bar, 1 mm in (e) and 100 μm in (h). CON: Ocn-Cre
   CKO: Kdm2b^fl/fl Ocn-Cre, (a) and (h) created with Biorender
   ([134]https://biorender.com/w07p252,
   [135]https://BioRender.com/ibb92gm). Source data of (b, d, e, g, i, j,
   k) are provided as a Source Data file.

   Strikingly, at around 10 months of age, the femur length of CKO mice
   was significantly increased (Fig. [136]2b), suggesting potential
   alterations in the bone’s internal structure. μCT analysis revealed a
   substantial enhancement in trabecular bone volume in CKO mice at
   various ages, with three-dimension (3D) reconstruction at 9 weeks of
   age (Supplementary Figs. [137]1d and 2c). Bone metrological analysis
   demonstrated significant increases in trabecular number (Tb.N),
   thickness (Tb.Th), and bone volume/total volume (BV/TV) ratio, along
   with a significant reduction in trabecular separation (Tb.Sp) in CKO
   mice (Fig. [138]2d). Hematoxylin-Eosin (H&E) staining of distal femur
   sections confirmed the increased bone mass in CKO mice (Fig. [139]2e).
   Heterozygous (HE) mice, however, showed no apparent differences in
   trabecular bone architecture compared to the control group
   (Supplementary Fig. [140]1e). In contrast to trabecular bone, no
   significant differences in cortical bone were observed between CKO and
   control mice, as evidenced by the lack of statistical variation in
   absolute or relative Cortical bone Area (Ct.Ar) (Fig. [141]2f, g).
   Thus, these data reveal a role for KDM2B in the negative regulation of
   trabecular bone mass.

   To directly assess new bone formation, we administered calcein
   injections 13 days and 3 days prior to the mice sacrifice, followed by
   the visualization of fluorescent bands (Fig. [142]2h). According to the
   distance between double labels of mineralization, the bone formation
   rate/bone surface (BFR/BS) and bone formation rate/total volume
   (BFR/TV) were significantly increased in CKO mice (Fig. [143]2i).
   Additionally, we compared the structural integrity of femurs from CKO
   and control mice using a three-point bending test. By comparing the
   compression loading, we noticed that the CKO femurs could withhold
   significantly higher forces (Fig. [144]2j, k), indicating enhanced bone
   stiffness compared to controls.

   Notably, tartrate-resistant acid phosphatase (TRAP) staining of femurs
   unveiled a significant increase in the number of OCs in CKO mice as
   measured by OC per tissue area (N.Oc/T.Ar) (Supplementary Fig. [145]1f,
   g). Moreover, the serum levels of RANKL, a well-recognized bone
   resorption factor were significantly elevated in CKO mice
   (Supplementary Fig. [146]1h). These data support that the enhanced bone
   density observed in CKO mice is not due to decreased bone resorption
   but attributed to elevated OB function and augmented bone formation.
   Consistent with this, OB-secreted RANKL has been shown to induce OC
   differentiation^[147]13. Therefore, these findings indicate that KDM2B
   limits bone remodeling under steady states.

KDM2B inactivation boosts OB functions to prevent bone loss and enhance bone
repair

   Next, we wondered how KDM2B-inactive OBs respond to diverse
   microenvironments of defective ossification. Initially, we compared the
   bone architecture of the femur between CKO mice and WT mice at 18
   months of age. μCT scanning unveiled that the trabecular bone loss
   observed in control mice was significantly mitigated in CKO mice
   (Fig. [148]3a). A substantial increase in BV and Tb.N, along with a
   decrease in Tb.Sp, was noted in CKO mice relative to the control group,
   though Tb.Th remained unchanged (Fig. [149]3b). Furthermore, to model
   postmenopausal osteoporosis, we performed ovariectomy (OVX) on both CON
   and CKO mice. μCT analysis indicated that CKO mice demonstrated
   significant resistance to OVX-induced bone loss (Fig. [150]3c, d).
   Thus, KDM2B-inactivated OBs effectively counteract bone loss associated
   with aging or hormonal deficiency.

Fig. 3. KDM2B inactivation boosts OB functions to counteract bone loss and
enhance bone repair.

   [151]Fig. 3
   [152]Open in a new tab

   a Representative images of 3D reconstruction and scanning sections of
   femurs from control and CKO mice at the age of 18 months. b μCT
   analysis of trabecular parameters (n = 3 mice per genotype). c
   Representative images of three-dimensional reconstruction and scanning
   sections of trabecular bone from OVX-CON, OVX-CKO, sham-CON, sham-CKO.
   d Statistical analysis to compare the BV/TV ratio in designated groups
   (n = 3 mice per genotype). e Representative images of scanning sections
   in coronal and sagittal plane of femur of bone defects (three columns
   on the left) and images of 3D reconstruction of repaired circular bone
   defect (column on the right) from control and CKO mice. f μCT analysis
   of BV/TV ratio of newly formed bone between control and CKO group
   (n = 3 mice per genotype). g Representative images of H&E staining of
   femur of repaired bone defect from control and CKO mice, box areas
   shown at a higher magnification. (C cortex, BM bone marrow).
   Statistical significance was assessed using Student’s t tests,
   two-tailed, in (b, f), error bars are presented as mean values ± SD.
   Statistical significance was assessed using Two-Way ANOVA, two-tailed,
   in (d), error bars are presented as mean values ± SD. Scale bar, 1 mm
   in low-power field and 100 μm in high-power field of 3 g. sham, sham
   operation. OVX, Ovariectomy. CON Ocn-Cre, CKO Kdm2b^fl/fl Ocn-Cre.
   Source data of (b, d, f, g) are provided as a Source Data file.

   To determine how KDM2B-deficient OBs react to acute stimuli, we created
   a circular bone defect (1-mm diameter) in the midshaft of the femur
   using low-speed dental handpieces in both CON and CKO mice
   (Supplementary Fig. [153]2a, b). As shown in Fig. [154]3e, CKO mice
   exhibited a greater presence of bone-like substances around the
   circular defects compared to the CON group. The BV/TV value of newly
   formed bone was approximately three times higher in CKO mice than
   in the CON group (Fig. [155]3f). Sagittal plane H&E staining clearly
   confirmed increased osteoid and new bone formation in CKO mice
   following acute bone trauma (Fig. [156]3g). Moreover, TRAP staining
   revealed a significant increase in the number of OCs in new bone,
   suggesting an enhanced bone turnover in CKO mice (Supplementary
   Fig. [157]2c, d). Consequently, the heightened OB functionality due to
   KDM2B inactivation not only prevents bone loss but also expedites the
   bone repair process after acute trauma.

KDM2B inactivation enhances OB functions via Wnt/β-catenin signaling

   Subsequently, we confirmed that KDM2B deficiency indeed results in the
   suppression of PRC1.1 activity. To accurately quantify the
   transcriptomic and epigenomic alterations following KDM2B inactivation,
   we utilized highly purified OBs for high-throughput sequencing
   analysis. To achieve this, we crossed Ocn^Cre or Kdm2b^fl/fl Ocn^Cre
   mice with Ai14 reporter mice, which enables OB-specific expression of
   tdTomato following Cre-mediated recombination (Fig. [158]4a).

Fig. 4. KDM2B inactivation enhances OB functions through the activation Wnt
signaling.

   [159]Fig. 4
   [160]Open in a new tab

   a Schematic of tdTomato-positive mouse generation. b Fluorescent images
   of trabecular and cortical bone sections from tdTomato mice. White
   triangles highlight Tomato-positive OBs in the trabecular region (green
   dashed line, ~400 μm²), while yellow triangles indicate OBs in the
   cortical region (yellow box, ~100 μm²). c Quantification of
   tdTomato-positive cells in periosteum of trabecula and cortex bone
   (n = 3 mice per genotype). d Workflow for isolation, sorting and
   analysis of tdTomato-positive OBs. e Volcano plot of DEGs in OBs
   isolated from 3-months-old mice (n = 3 CON and n = 3 CKO). Upregulated
   genes are labeled in red and downregulated genes in blue (Wald test,
   p < 0.05; |Foldchange| >1.5) OBs were collected after 14 days of
   culture by removing the bone fragments. f and h GO enrichment analysis
   for biological processes and KEGG enrichment analysis in upDEGs (CKO vs
   CON (YOUNG), Fisher’s exact test, p < 0.05). Number of genes and
   statistical significance are shown. g and j RT-qPCR analysis of
   representative genes in CKO and control OBs. i The heatmap showing the
   levels of leading genes in Wnt/β-catenin signaling pathway. Gene
   expression levels are shown as relative Z-Scores between two groups. k
   WB analysis of β-catenin levels in total protein lysates of femur
   tissue and tdTomato-positive OBs of control and CKO mice. GAPDH was
   used as loading control. l WB analysis of cytoplasmic and nuclear
   β-catenin in tdTomato-positive OBs from control and CKO mice. TUBULIN
   and H3 were used as fraction-specific controls. m WB analysis of KDM2B
   levels in KDM2B-WT and KDM2B-CxxC deletion mutant overexpressing
   MC3T3-E1 cells. n ChIP-qPCR analysis comparing β-catenin enrichment at
   target genes in differentiated MC3T3-E1 cells (n = 3/group). o
   Representative images of 3D reconstruction of trabecula bone from CKO
   mice treated with PBS or LGK974 for 1 month. p µCT analysis of
   trabecular parameters (n = 4 mice/group). Statistical significance was
   assessed using Student’s t tests, two-tailed, in (c, g, j, p), error
   bars are presented as mean values ± SD. Statistical significance was
   assessed using Two-Way ANOVA, two-tailed, in (n), error bars are
   presented as mean values ± SD. Scale bar, 100 μm. CON: Ocn ^Cre
   td-Tomato, CKO: Kdm2b^fl/fl Ocn ^Cre td-Tomato in (b–i, o, p); CON:
   empty vector in (m). a and d created with Biorender
   ([161]https://biorender.com/w07p252,
   [162]https://biorender.com/p69f867). Source data of (c, g, j, k, l, m,
   n, p) are provided as a Source Data file.

   After genotyping (Supplementary Fig. [163]3a), the phenotype of the
   femur in CKO-tdTomato mice was confirmed by μCT (Supplementary
   Fig. [164]3b). The numbers of tdTomato-positive OBs were increased in
   both the trabecular and cortical bone of CKO mice (Fig. [165]4b, c),
   suggesting that the enhanced OB function also stimulates OB
   differentiation as positive feedback. Femurs from CON and CKO-tdTomato
   mice were minced into small fragments for OB culture. Within a few
   days, tdTomato-positive cells were observed migrating out of the
   cultured bone fragments (Supplementary Fig. [166]3c).
   Immunofluorescence (IF) staining confirmed the expression of RUNX2 and
   ALP, two osteogenic markers, in tdTomato-positive cells (Supplementary
   Fig. [167]3d). The adherent cells were subsequently harvested and
   collected after 14 days of culture, sorted by flow cytometry, and
   subjected to RNA-Seq analysis (Fig. [168]4d).

   In either young or aged CKO OBs, hundreds of DEGs were identified
   compared to their respective controls (Fig. [169]4e, Supplementary
   Fig. [170]3e and Supplementary Data [171]3). Notably, 43% of the aged
   OB-downregulated genes (166 out of 386) are upregulated in aged CKO
   group (Supplementary Fig. [172]3f). GO analysis indicated the
   involvement of upregulated genes in young and old CKO OBs with skeletal
   system morphogenesis, embryonic skeletal system development, and
   skeletal system development (Fig. [173]4f and Supplementary
   Fig. [174]3g). RT-qPCR analysis confirmed the upregulated expression of
   osteogenic genes (Fig. [175]4g) in CKO OBs. Collectively, these
   analyses demonstrate that KDM2B ablation significantly reverses the
   deterioration of OB functions. Given the partial reversing effect,
   further studies are warranted to elucidate the interplay between PRC1.1
   and other regulatory mechanisms in the context of aging.

   Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis showed that
   upregulated genes were enriched in Wnt signaling pathways (Fig. [176]4h
   and Supplementary Fig. [177]3g). As illustrated by heatmap and
   confirmed RT-qPCR analysis, multiple components of the Wnt pathway,
   including ligands, receptors, and signal transducers, were
   significantly activated upon KDM2B loss (Fig. [178]4i, j). Moreover,
   protein levels of β-catenin were increased in both femur tissues and
   tdTomato-positive OBs from CKO mice (Fig. [179]4k). To quantify the
   subcellular distribution of β-catenin, we separated nucleoplasmic and
   cytoplasmic fractions of OBs. WB analysis showed that β-catenin levels
   were significantly increased in the nucleus and slightly decreased in
   the cytoplasm (Fig. [180]4l).

   Runx2 and Sp7 recognized as osteogenic target genes downstream of
   β-catenin^[181]8,[182]13, were further investigated for their
   regulation by KDM2B. We overexpressed KDM2B-WT and KDM2B-ΔCxxC mutant
   in 3T3-E1 cells, a pre-OB cell line (Fig. [183]4m), and induced OB
   differentiation. Chromatin Immunoprecipitation (ChIP)-qPCR analysis of
   the differentiated cells revealed a significant increase of β-catenin
   binding at the Runx2 and Sp7 promoter regions in the KDM2B-ΔCxxC mutant
   group when compared to the KDM2B-WT group (Fig. [184]4n). These data
   suggest that KDM2B ablation strengthens the Wnt signaling pathway.

   To substantiate the roles of activated Wnt signaling in OB functions,
   we treated the CKO mice with LGK974, a potent and specific inhibitor
   blocking WNT secretion^[185]37,[186]38. After two weeks of LGK974
   intraperitoneal injection (3 mg/kg), IF staining confirmed decreased
   β-catenin intensity in OBs (Supplementary Fig. [187]3h). μCT scanning
   showed that trabecular bone was significantly reduced by LGK974
   treatment. Statistical analyses for trabecular bone microarchitecture
   unveiled a significant decrease in BV/TV, Tb.Th, and Tb.N and unchanged
   Tb.Sp (Fig. [188]4o, p). These data confirm the essential role of Wnt
   signaling in the enhanced OB functions resulting from KDM2B
   inactivation.

H2AK119ub1 loss in KDM2B-inactive OBs is associated with the Wnt/β-catenin
target gene derepression

   To find out whether the enhanced Wnt signaling in KDM2B-deficient OBs
   is attributed to the dampened function of PRC1.1, we simultaneously
   performed H2AK119ub1 CUT&Tag analysis (Fig. [189]4d). As expected,
   H2AK119ub1 levels at the promoters of 763 genes were significantly
   reduced in CKO OBs (Fig. [190]5a, b and Supplementary Data [191]4).
   Integrating these results with RNA-seq data revealed significant
   upregulation of the associated genes in CKO OBs (Fig. [192]5c). KEGG
   and GO analysis revealed that these genes are significantly enriched in
   the Wnt signaling pathway and are involved in skeletal system
   morphogenesis and development (Fig. [193]5d, e). Decreased H2AK119ub1
   enrichment and corresponding expression patterns in CKO OBs were
   observed at representative Wnt regulator genes, such as Wnt10b, Fzd10,
   with an irrelevant intergenic region as a negative control
   (Fig. [194]5f). Collectively, these data suggest that PRC1.1-mediated
   H2AK119ub1 accounts for the repression of Wnt/β-catenin target genes in
   OBs.

Fig. 5. Derepression of the Wnt/β-catenin target genes in KDM2B-inactive OBs
is associated with reduced H2AK119ub1 enrichment.

   [195]Fig. 5
   [196]Open in a new tab

   a Heatmaps represent the enrichment for H2AK119ub1 in chromatin of CON
   and CKO mice around the TSS (± 5 Kb) of H2AK119ub1 target genes. (n = 2
   per group). Colors represent CUT&Tag RPM, and rows  are ranked by
   CUT&Tag signals in CON. b Box plot showing peak intensity of CUT&Tag
   reads for H2AK119ub described in panel (a). (Wilcoxon matched-pairs
   signed rank test, two tailed, p < 0.0001) c Quantification of
   log[2]-transformed fold change of expression levels of H2AK119ub1
   decreased genes (CKO vs CON, n = 759) in promoter region of OBs (n = 3
   per group, Wilcoxon matched-pairs signed rank test, two tailed,
   p < 0.0001). The box plots in (b) and (c) indicate the median (centre
   line), the third and first quartiles (box limits) and 1.5 × IQR above
   and below the box (whiskers). (d, e) KEGG enrichment analysis and GO
   enrichment analysis for biological processes of genes described in (c),
   (n = 763, Fisher’s exact test, p < 0.05). Number of genes and
   statistical significance are shown. (f) The IGV view of decreased
   H2AK119ub1 enrichment at representative Wnt/β-catenin target genes and
   increased gene expression levels in control and CKO OBs. An intergenic
   region serves as a negative control. CON: Ocn ^Cre td-Tomato CKO:
   Kdm2b^fl/fl Ocn ^Cre td-Tomato. Source data of (b, c) are provided as a
   Source Data file.

iBP disassembles PRC1.1 by blocking the interaction between PCGF1 and BCOR

   Next, we investigated the possibility of pharmacologic targeting of
   PRC1.1 and the potential impact on bone formation. To develop specific
   PRC1.1 inhibitors, we examined the available structural data. We
   recently identified an electrostatic interaction formed between the
   linker region of BCOR and the positively charged patch on the F-box and
   LRRs of KDM2B^[197]39. However, this interaction surface is rather
   shallow, lacking a desirable binding pocket for inhibitor design. Here,
   we focused on the interaction between the BCORL1 PUFD domain and the
   RAWUL domain of PCGF1, which creates an extended interface for
   associating with KDM2B and is unique to the PRC1.1^[198]40. Based on
   previous studies, a pocket on PCGF1, which packs against two
   anti-parallel β strands on BCOR (Fig. [199]6a), is essential for
   PCGF1-BCOR complex formation^[200]41. Thereby, inhibitors targeting
   this pocket would potentially impair PCGF1 binding to BCOR/BCORL1,
   eventually disrupting the activity of PRC1.1. Based on this analysis,
   we performed high-throughput virtual screening of 231,000 compounds
   from ZINC database (Fig. [201]6b). Based on Docking Scores and visual
   inspection, 11 hits were selected for activity validation
   (Supplementary Fig. [202]4).

Fig. 6. Screening and validation of specific PRC1.1 inhibitors.

   [203]Fig. 6
   [204]Open in a new tab

   a Structural model of the binding interface between PCGF1-BCOR complex
   (PDB: 4HPL). The RAWUL domain of PCGF1 is shown as a surface (carbon in
   white, oxygen in red, nitrogen in blue, sulfur in yellow), while the
   PUFD domain of BCOR is depicted as a green cartoon. b Flowchart of
   screening and verification of PRC1.1 inhibitors. c AlphaScreen assay to
   show the inhibition of PCGF1^RAWUL L238/F242 interaction with BCOR^PUFD
   by selected compounds. d Split-GFP system construction. Representative
   images of each group of transfected 293 T cells are shown. Scale bar,
   100 μm. e Quantification of GFP/Hoechst fluorescence intensity in 293 T
   cells co-transfected with GFP1-10-PCGF1 and GFP11-BCOR, subsequent to
   treatment with candidate inhibitors. Statistical significance was
   assessed using One-Way ANOVA, two-tailed, error bars are presented as
   mean values ± SD (n = 3 per group). f Chemical structure of iBP (Left).
   Docking model of PCGF1^RAWUL (surface) and iBP (orange stick). g
   AlphaScreen determination of iBP’s IC50 value for inhibiting the
   interaction between PCGF1^RAWUL L238A/F242A and BCOR^PUFD. h Biolayer
   interferometry (BLI) analysis of iBP binding to PCGF1^RAWUL
   L238A/F242A. Representative data and 1:1 binding model fit shown as
   blue and red lines, respectively. i, j Co-IP assay examining how iBP
   treatment at various concentrations affects the interaction between
   PCGF1 and BCOR. k ChIP-qPCR analysis comparing H2AK119ub1, PCGF1 and
   RING1B enrichment at PRC1-target promoters and non-targeted gene bodies
   as control in control and iBP-treated 293 T cells. Statistical
   significance was assessed using Two-Way ANOVA, two-tailed, error bars
   are presented as mean values ± SD (n = 3 ~ 6 per group). Source data of
   (e, g–k) are provided as a Source Data file.

   To determine the inhibitory potency of these hit compounds on the
   PCGF1-BCOR interaction in vitro, we conducted a bead-based AlphaScreen
   assay with recombinant PCGF1^RAWUL protein and BCOR^PUFD. As the
   purification of wild type PCGF1^RAWUL protein turned out to be rather
   challenging, eventually we obtained purified PCGF1^RAWUL mutant
   (PCGF1^RAWUL L238A/F242A) by mutating hydrophobic residues (L238 and
   F242) on the surface. In the structure of PCGF1^RAWUL/BCOR^PUFD
   complex, these two residues are away from the predicted inhibitor
   binding pocket (Supplementary Fig. [205]5), and are surrounding the Leu
   cage of BCOR^PUFD that is less important for the complex
   formation^[206]41. AlphaScreen assay showed that, among these selected
   compounds, Compound 1, 6, 8, 9 exhibited more than 50% inhibition rate
   for the interaction between PCGF1^RAWUL L238A/F242A and BCOR^PUFD, with
   Compound 8 (C8) achieving an inhibition rate about 95% at concentration
   of 200 μM (Fig. [207]6c).

   Meanwhile, we assessed the cellular efficacy of these candidate
   compounds using a Split-GFP system (Fig. [208]6b), which relies on the
   self-association of GFP10 and GFP11 with GFP1-9 to reconstitute a
   functional GFP upon interaction of the fused proteins^[209]42. We fused
   PCGF1 and BCOR into plasmids with GFP1-10 and GFP11 fragments,
   respectively, and co-transfected the plasmids into 293 T cells
   (Fig. [210]6d). High-content imaging and analysis revealed that among
   the compounds tested, C6 and C8 significantly downregulated GFP
   fluorescence intensity, with C8 exhibiting the most potent efficacy
   (Supplementary Fig. [211]6a and [212]e). Based on these in vitro and
   intracellular screenings, we selected C8 as the optimal candidate and
   named it as “iBP”, the inhibitor of BCOR-PCGF1 (Fig. [213]6f and
   Supplementary Fig. [214]6b). The IC50 of iBP, as determined by
   AlphaScreen, was 12.21 ± 0.88 μM (Fig. [215]6g). In addition, affinity
   measurements revealed that iBP bound to PCGF1^RAWUL L238A/F242A with a
   dissociation constant (Kd) of 6.42 ± 0.75 μM (Fig. [216]6h).

   High-performance liquid chromatography (HPLC) analysis showed that
   92.2% and 86.9% of iBP remained after 12 and 24 h (hrs) respectively in
   cell culture media, indicating iBP’s inherent stability (Supplementary
   Fig. [217]6c). According to the luminescence cell viability assay in
   293 cells, the IC50 of iBP was 42.91 μM (Supplementary Fig. [218]6d).
   To determine a safe and effective concentration of iBP, we conducted
   Co-Immunoprecipitation (Co-IP) assays in 293 T cells. Titrating iBP to
   5 and 10 μM, we found that 10 μM was sufficient to inhibit the
   interaction between PCGF1 and BCOR, without disrupting PRC2 or PRC1.4
   integrity (Fig. [219]6i, j and Supplementary Fig. [220]7a, b). WB
   analysis showed a reduction of total H2AK119ub1 levels in 293 T cells
   following treatment with 10 μM iBP (Supplementary Fig. [221]7c).
   ChIP-qPCR analysis further demonstrated a significant reduction in the
   enrichment of PCGF1, RING1B and H2AK119ub1 levels specifically at
   target promoter regions following iBP treatment (Fig. [222]6k). These
   findings confirm iBP as a potent and selective PRC1.1 inhibitor.

iBP treatment enhances OB functions and reverses bone loss in OVX models

   To delineate the functional impact of iBP treatment on aged OBs, we
   conducted RNA-seq analysis. Strikingly, 1786 genes upregulated in aged
   OBs exhibited significant enrichment in osteogenic morphogenesis,
   skeletal development, and Wnt/β-catenin signaling pathways
   (Fig. [223]7a, b). Notably, 386 genes downregulated during aging showed
   marked upregulation in both CKO and iBP-treated aged OBs compared to
   age-matched controls (Fig. [224]7c). These findings collectively
   demonstrate that genetic ablation of KDM2B or pharmacological
   inactivation of PRC1.1 rescues the transcriptional signature of bone
   aging.

Fig. 7. iBP treatment enhances OB functions and reverses bone loss in OVX
models.

   [225]Fig. 7
   [226]Open in a new tab

   a Volcano plot of DEGs in OBs (n = 3 CON and n = 3 CON-iBP).
   Upregulated genes are labeled in red and downregulated genes in blue
   (Wald test, p < 0.05; |Fold change| >1.5). The x-axis shows the log[2]
   (fold change) in gene expression between CKO and control OBs, and the
   y-axis shows the statistical significance of the differences. OBs were
   collected after 14 days of culture by removing bone fragments. b GO
   enrichment analysis for biological processes and KEGG enrichment
   analysis of upDEGs in (a), (CKO-iBP vs CON, n = 1786, Fisher’s exact
   test, p < 0.05). Number of genes and statistical significance are
   shown. c Boxplots showing log[2]-transformed fold change of expression
   levels of downDEGs in aged OBs (OLD vs YOUNG, n = 386) following KDM2B
   ablation or iBP treatment (n = 3 per group, Wilcoxon matched-pairs
   signed rank test, two tailed, p < 0.0001). The boxplots indicate the
   median (centre line), the third and first quartiles (box limits) and
   1.5 × IQR above and below the box (whiskers). d, e Images of 3D
   reconstruction of trabecula bone of OVX mice treated with iBP or PBS,
   with sham group as controls. CT analysis of the distal femur
   metaphysis. Statistical significance was assessed using One-Way ANOVA,
   two-tailed, error bars are presented as mean values ± SD, n = 5 per
   group. f H&E staining of femur of the designated groups of mice. g IHC
   staining of WNT2, WNT3A and β-catenin in femurs of the designated
   groups of mice. Scale bar, 1 mm. OVX, Ovariectomy. Source data of (e,
   f, g) are provided as a Source Data file.

   We also tested iBP’s effects on osteogenic differentiation in MC3T3-E1
   cells. After 14 days in osteogenic medium (OM), we observed a
   significant upregulation in the expression of osteogenic marker genes
   and Wnt/β-catenin target genes, which was further enhanced by iBP
   treatment (Supplementary Fig. [227]7d, e). Additionally, transfecting
   MC3T3-E1 cells with the TOP-FLASH reporter plasmid, containing six
   TCF-binding motifs upstream of a luciferase gene, revealed that iBP
   treatment significantly activated the Wnt reporter. This activation was
   reversed by LGK974 (Supplementary Fig. [228]7f). These data demonstrate
   that iBP treatment enhances OB functions as well as KDM2B ablation.

   To evaluate the drug’s toxicity and metabolic profile in vivo,
   pharmacokinetic (PK) studies were conducted. Following an
   intraperitoneal administration of 20 mg/kg iBP to C57/BL6 mice, the
   maximum concentration (C[max]) of iBP reached 3209 ng/mL (9.66 μM) at
   0.083 h (T[max]) (Supplementary Fig. [229]8a), approaching the
   effective concentration of 10 μM noted in 293 T cells. Moreover, iBP
   exhibited acceptable plasma exposure (AUC^0-t, 1973 h × ng/mL) and
   half-life (T1/2, 1.05 h). Importantly, iBP was well tolerated by the
   tested mice, indicating its suitability for in vivo efficacy studies.
   Therefore, a 20 mg/kg dose was selected for further experiments.

   iBP was administered intraperitoneally to OVX mice for one month
   (20 mg/kg daily). During this period, OVX mice exhibited weight gain
   compared to the control sham group, which was modestly restricted by
   iBP treatment (Supplementary Fig. [230]8b). Histological examination of
   heart, liver, spleen, lung, and kidney tissues showed no apparent
   abnormalities (Supplementary Fig. [231]8c). Bone metrology analysis
   following μCT scanning revealed that iBP treatment markedly reversed
   the loss of trabecular bone in OVX mice, as evidenced by the recovery
   of BV/TV, Tb.Th, and Tb.N, with no alterations in Tb.Sp (Fig. [232]7d,
   e). Increased trabecular bone was also confirmed by sagittal plane H&E
   staining (Fig. [233]7f). Immunohistochemistry (IHC) assay demonstrated
   that the expression levels of WNT2, WNT3a and β-catenin were increased
   in iBP-treated femur (Fig. [234]7g). These data indicate that iBP
   treatment effectively activates Wnt signaling and prevents
   osteoporosis.

iBP treatment accelerates bone healing

   Then we continued to examine whether iBP could foster bone repair in
   bone trauma models. To enhance local repair efficiency, we formulated
   an iBP-loaded hydrogel for targeted and controlled drug delivery. F127,
   a thermosensitive hydrogel known for its temperature responsiveness,
   biodegradability, and excellent biocompatibility^[235]43, was selected
   for this purpose. It solidifies at room temperature (RT) (25°), as
   shown in Supplementary Fig. [236]8d.

   Analysis of particle size and surface charge revealed iBP particles to
   be around 162 nm in diameter (Fig. [237]8a and Supplementary
   Fig. [238]8e). Scanning electron microscopy showed the hydrogel voids
   to be approximately 10 μm in diameter (Fig. [239]8b), implying that the
   nanosized iBP would result in an initial burst release. Gelatin, known
   to reduce the burst release of F127, is clinically utilized for bone
   wound hemostasis^[240]44,[241]45, making it a promising candidate for
   iBP delivery. We cut gelatin sponges to a specific size
   (2 × 2 × 10 mm^3) (Supplementary Fig. [242]8f) and injected cold (4 °C)
   liquid iBP-loaded F127 into them before quickly transferring them to a
   37 °C incubator to facilitate solidification within the sponge
   (Fig. [243]8c). Based on the UV absorbance spectrum of iBP, which peaks
   at 300 nm, we established a standard curve (Supplementary Fig. [244]8g,
   h) and found that a 20% F127@gelatin sponge loaded with 2 mg/ml iBP
   exhibited a sustained release over 14 days, with an average drug
   release concentration of 20 μM (Fig. [245]8d). Thus, we successfully
   developed a sustained-release delivery system for iBP.

Fig. 8. iBP treatment accelerates bone healing.

   [246]Fig. 8
   [247]Open in a new tab

   a Images obtained by transmission scanning microscope of iBP. b Images
   of scanning electron microscope of F127. c Schematic diagram of
   construction of iBP loaded gelatin sponge -F127 hydrogel. d Statistical
   analysis of sustained release efficiency of iBP for 14 days. e
   Representative images of scanned sections and sagittal plane views of
   femur defects in the designated treatment groups for two weeks. Red
   triangles indicate the new callus at the sites of circular bone defect.
   f BV/TV quantification of the osteotylus in (e). Statistical
   significance was assessed using One-Way ANOVA, two-tailed, error bars
   are presented as mean values ± SD, n = 3 per group, Mock: A model
   representing femur defects, F127-Gelatin: femur defects area treated
   with F127-Gelatin for two weeks, iBP@F127-Gelatin: femur defects area
   treated with iBP@F127-Gelatin for two weeks. c created with Biorender
   ([248]https://biorender.com/j89n671). Source data of (a, b, d, f) are
   provided as a Source Data file.

   With this system, we evaluated the healing effects of iBP. Notably,
   after 14 days of healing, mice that received local iBP treatment
   exhibited a significantly larger amount of new callus formation
   compared to mice in the mock or hydrogel-only control groups, as
   revealed by μCT scanning (Fig. [249]8e, f). These findings indicate
   that iBP treatment promotes new bone formation in vivo.

Discussion

   Epigenetic regulation plays critical roles in tissue homeostasis,
   extending beyond lineage differentiation and development. In this
   study, we provide genetic evidence that KDM2B inactivation in OBs
   enhances bone homeostasis by derepressing Wnt/β-catenin target genes
   due to impaired PRC1.1 activity. Based on this finding, we have
   developed a specific PRC1.1 inhibitor, iBP, which enhances osteogenic
   functions in mouse models of osteoporosis and acute trauma
   (Supplementary Fig. [250]9).

   The processes of cell signal transduction and epigenetic regulation are
   intricately linked in OB differentiation, bone formation, and
   homeostasis. Current strategies to increase bone density, such as
   administering Wnt ligands, Vitamin D, Parathyroid Hormone (PTH), or
   Bone Morphogenetic Proteins (BMPs), rely heavily on extracellular
   factors^[251]46. However, the effectiveness of these approaches depends
   on the systemic or local microenvironment, which influences chromatin
   states and thus responsiveness to signaling cascades. This underscores
   the importance of identifying chromatin regulators that shape the
   chromatin landscape, facilitating osteogenic lineage commitment and
   bone remodeling.

   The biochemistry and epigenomic regulation of PRCs have been well
   characterized. In stem cells, PRCs are known to preserve their
   differentiation potential by silencing key developmental genes, keeping
   them in a poised state ready to respond to signaling
   cues^[252]15,[253]20,[254]22. For example, the Wnt10b pathway promotes
   mesenchymal stem cell (MSC) differentiation towards
   OBs^[255]47,[256]48, a process that PRCs restrict. Accordingly, EZH2
   inhibition stimulates OB differentiation^[257]32,[258]49. However, the
   composition and chromatin binding of PRCs vary dynamically across
   different developmental stages^[259]22,[260]23,[261]36,[262]50,[263]51,
   during aging^[264]52,[265]53, disease progression or under drug
   treatment^[266]54–[267]57. The context-dependent regulation of PRCs and
   their functions will be an interesting direction for future research.
   Our unexpected observation of increased PRC1.1 activity in aging bones
   raises intriguing questions about potential upstream modulators. While
   our study delineates the functional consequences of PRC1.1 activation
   in aged OBs, the molecular triggers driving its enhanced activity
   remain incompletely resolved. We postulate that unidentified signaling
   cascades or post-translational modifications of KDM2B-PRC1 components
   may fine-tune its chromatin-binding affinity or enzymatic robustness in
   aging contexts.

   Detailed phenotypic analyses in mouse models reveal that PRC1.1
   inactivation not only enhances OB function but also promotes OB
   differentiation. This increase in OB differentiation is supported by
   the elevated OB counts in CKO mice and the activation of Wnt/β-catenin
   and osteogenic genes in PRC1.1-inactive pre-OB cells. Consistently,
   BCOR deficiency in MSCs enhances osteo-dentinogenic differentiation.
   BCOR loss-of-function mutations are associated with
   oculo-facio-cardio-dental (OFCD) syndrome, which includes
   characteristic dental and neural crest defects^[268]58. In contrast,
   the inactivation of PR-DUB complex as evidenced by Asxl1 deletion
   causes gain of H2AK119ub1 activity and significant OB dysfunction and
   bone loss^[269]59. These evidence highlight a critical role of
   H2AK119ub1 activity in the epigenetic regulation of OB lineage
   differentiation and functions. Additionally, KDM2B inactivation also
   enhances OC activity (Supplementary Figs. [270]1f, g and [271]2c, d),
   suggesting a new equilibrium in which bone formation and resorption are
   heightened upon PRC1.1 inhibition.

   Our transcriptomic and epigenomic analyses indicate that excessive
   PRC1.1 accumulation raises the activation threshold for Wnt/β-catenin
   target genes, a crucial regulatory mechanism for bone homeostasis.
   PRC1.1 inactivation lowers this threshold, enhancing the
   transcriptional response to Wnt signaling, particularly under bone-loss
   conditions associated with PRC1.1 gain, such as aging. Additionally,
   PRC1.1 inhibition may increase chromatin accessibility to various
   stimuli, indicating that combining PRC1.1 targeting with extracellular
   osteogenic factors could provide synergistic therapeutic benefits.

   Existing PRC1 inhibitors like PRT-4165^[272]60, RB-3^[273]61, etc.
   target RING1B’s E3 ligase activity or the stability of key partner
   proteins, broadly affecting PRC1 complexes. In contrast, iBP
   selectively targets PRC1.1, potentially minimizing side effects. In our
   OVX mouse models, it showed no significant toxicity over one month of
   treatment. Our study provided the core structure of iBP, which could
   function as a lead compound for future optimization to improve its
   bioavailability. Additional long-term studies in aged mice with varying
   doses are necessary to confirm its safety and efficacy. Beyond bone
   disease, iBP could offer insights into PRC1.1’s role in transcriptional
   regulation and might be applicable to other conditions, such as
   PRC1.1-overactive cancers^[274]62,[275]63.

   However, our study has some limitations. The factors driving PRC1.1
   gain during aging and its occurrence in other pathological conditions
   are still unknown. Additionally, the detailed regulatory mechanisms by
   which PRC1.1 influences OB differentiation and OC activity require
   further elucidation. Moreover, clinical applicability will depend on
   assessing PRC1.1 expression patterns and relevant epigenomic profiles
   in human bone deterioration disorders.

   In summary, our genetic mouse models and chemical screenings reveal a
   promising epigenetic therapeutic strategy to increase bone density. iBP
   may serve as a candidate for treating other conditions characterized by
   aberrant PRC1.1 function, offering a novel approach to address bone
   loss and related diseases.

Methods

Mice

   All C57BL/6 mice were maintained in specific pathogen-free barrier
   facilities and housed at 20–22 °C with 12 h:12 h light: dark cycles at
   50–60% humidity, animals were used in accordance with protocols
   approved by the institutional animal care and user committee (Approval
   No. TMUaMEC 2020008) at the Tianjin Medical University
   (SYXK(JIN)−2020-0010). The details of mice age and sex are provided in
   the main text and figure legends. Kdm2b^fl/fl mice were reported
   previously^[276]18. Ocn-Cre mice were purchased from Cyagen.
   Ai14(Rosa-CAG-LSL-tdTomato) mice were a gift from Dr. Zhiyong Liu,
   Institute of Neuroscience, CAS. Kdm2b^fl/fl mice were cross-bred with
   Ocn-Cre mice to specifically delete KDM2B-CxxC domain OBs (Kdm2b^fl/fl
   Ocn-Cre). The genotyping primers are listed in Supplementary
   Table [277]1. Age- and sex-matched littermates Ocn-Cre mice were used
   as control.

Isolation and culture of OBs

   To isolate OBs, femurs from 3-month-old and 18-month-old C57BL/6 mice
   were cleaned of surrounding tissue. After the bone marrow was flushed
   out with phosphate-buffered saline (PBS), the bone tissue was cut into
   small pieces. These bone fragments were then digested overnight at
   37 °C using Type II collagenase (C8150, Solarbio) and subsequently
   cultured in α-MEM medium supplemented with 10% FBS (LONSERA, Suzhou
   Shuangru Biotechnology) and 1% penicillin-streptomycin (P/S) (Sigma) in
   a humidified incubator at 37 °C with 5% CO[2]. Adherent OBs were
   collected after 14 days of culture by removing the bone
   fragments^[278]64. For tdTomato-reporter mice, OBs were processed
   similarly and subsequently sorted using flow cytometry based on
   tdTomato expression, which was used as a marker for OBs.

Culture and osteogenic induction of MC3T3-E1 cells

   MC3T3-E1 cells are grown in Dulbecco’s Modified Eagle’s Medium (DMEM,
   Gibco) medium (GM) supplemented with 10% FBS, 10 mM HEPES (Sigma), 1%
   P/S. The cells were maintained in a humidified incubator at 37 °C with
   an atmosphere of 5% CO[2]. The culture medium was changed every 2 and 3
   days to ensure cell growth and proper maintenance. For osteogenic
   differentiation, the MC3T3-E1 cells were induced with an osteogenic
   medium (OM), which consists of 10% FBS in DMEM, 50 μM ascorbic acid,
   10 mM β-glycerol phosphate, and 100 nM dexamethasone.

WB

   For the extraction of total proteins from OBs, cells were lysed in RIPA
   buffer supplemented with a phosphatase and protease inhibitor cocktail
   (PIC, HY-K0010, MCE). For bone tissues, femurs were isolated from the
   surrounding soft tissues, and the mid-shaft portions were carefully
   dissected and ground into a fine powder under liquid nitrogen and a
   mortar and pestle. These powders were then lysed in RIPA buffer
   containing the same inhibitor cocktail for 30 min (min) at 4 °C. After
   centrifugation at 12,000 × g for 15 min at 4 °C to remove cellular
   debris, the supernatant containing the total protein extract was
   collected, and the proteins were denatured and separated by SDS-PAGE on
   a 10–12% polyacrylamide gel, depending on the protein size. The
   resolved proteins were transferred onto polyvinylidene fluoride (PVDF)
   membranes at 300 mA for 2 h. The membranes were blocked with 5% non-fat
   dry milk in Tris-buffered saline with 0.1% Tween-20 (TBST) for 1 h at
   RT to reduce non-specific binding. After blocking, membranes were then
   probed with specific primary antibodies which are listed in
   Supplementary Table [279]2. Following incubation with appropriate
   secondary antibodies, the membranes were washed three times for 5 min
   each with TBST. The blots were visualized using the NcmECL High Reagent
   (P2300, New Cell & Molecular Biotech) and captured with a Uvitec
   Cambridge Imaging System. The protein bands were quantified using
   ImageJ software by normalizing to an internal control.

RNA-Seq

   The library preparation and sequencing were conducted by Novogene,
   using an Illumina Novaseq platform to generate 150 bp paired-end reads.
   The raw sequencing data were first cleaned to remove low-quality reads.
   Clean reads were aligned to the mm39 reference genome with Hisat2
   (v2.2.1), and Samtools (v1.19) was used to convert sam files to bam
   files. Deeptools (v3.5.4) was employed to generate bigwig files, with
   visualization done via Integrative Genomics Viewer. Gene counts were
   calculated by Stringtie (v2.2.1), and differential expression analysis
   was performed using the R package DESeq2 (v1.34.0). Genes were
   considered significantly different if they met the criteria of P < 0.05
   and |log[2]FoldChange| > 0.585. Volcano plots and heatmaps were
   generated with R packages ggplot2 (v3.4.3) and pheatmap (v1.0.12).
   Pathway enrichment analysis was performed using Metascape, and selected
   pathways were visualized with R packages ggplot2 (v3.4.3)^[280]65.

CUT&Tag

   CUT&Tag assay was performed using the Hyperactive Universal CUT&Tag
   Assay Kit (TD904, Vazyme) as previously described^[281]36. Briefly,
   100,000 cells were harvested, washed, and resuspended in wash buffer
   before incubation with pre-activated concanavalin A beads. Primary
   anti-H2AK119ub1 antibody (1:1000, 8240S, CST) and secondary antibodies
   were added sequentially, followed by incubation with Dig-wash buffer
   and pA/G-Tnp Pro was added to initiate tagmentation. Once tagmentation
   was complete, DNA was extracted from the beads, and an equal amount of
   E. coli DNA was added as a spike-in control. After DNA extraction and
   PCR amplification, sequencing was performed on an Illumina Novaseq6000
   platform (Annoroad Gene Technology, Beijing). Raw data underwent
   quality evaluation with Fastqc (v0.12.1) and Multiqc (v1.19). Adapters
   were removed with Trim-galore (v0.6.10). Clean reads were aligned to
   the mm39 reference genome and λDNA (TD904, Vazyme), normalized based on
   scale-factor, and duplicates removed with Sambamba (v1.0). Peak calling
   was conducted with Macs2 (v2.2.9.1), and bigwig files were generated
   with Deeptools (v3.5.4). Peaks were annotated with R packages
   ChIPseeker (v1.30.3), and pathway enrichment analysis was performed
   using Metascape ([282]https://metascape.org/gp/index.html).
   Visualization of selected pathway were generated by R packages ggplot2
   (v3.4.3)^[283]65.

µCT

   To observe and evaluate bone microstructure, µCT analysis was
   performed. Briefly, femurs were harvested from mice, fixed overnight in
   4% paraformaldehyde (PFA), and scanned with an isotropic voxel size of
   8 µm and a peak tube voltage of 55 kV and current of 100 μA (SKYSCAN
   1276; Bruker). Three-dimensional images were reconstructed and regions
   of interest (ROI) were analyzed using NReconServer, CTAn, and CTvox
   softwares (GE). Trabecular bone parameters were measured from 100
   consecutive slices of the distal metaphysis, devoid of epiphyseal
   structures.

Calcein double labeling and three-point bending test

   For calcein double labeling, mice were injected with calcein at
   10 mg/kg on day 13 and day 3 prior to femur harvest. Femurs were fixed,
   embedded in methylmethacrylate, and sectioned at 5 μm. Images were
   acquired using a fluorescence confocal microscope, and bone formation
   rates (BFR/BS and BFR/TV) were calculated from fluorochrome double
   labels at periosteal and endocortical surfaces using OsteoMeasure
   software. The three-point bending test was conducted on fresh femurs to
   assess structural and material strength using an Instron
   electromechanical tester (Instron 3367). Load-displacement curves were
   recorded until fracture, and the maximum force at failure was
   calculated.

H&E staining, TRAP staining and IHC

   Samples including femurs, heart, liver, spleen, lung, and kidney were
   collected and sectioned at 5 μm for H&E staining. Paraffin sections
   were dewaxed, stained with hematoxylin and eosin, dehydrated, and
   sealed^[284]66. For IHC, slides were stained with specific antibodies
   which are listed in Supplementary Table [285]2. Sections were
   deparaffinized, rehydrated, and antigen retrieval was performed with 3%
   hydrogen peroxide. After blocking with 3% BSA, primary and secondary
   antibodies were applied, and DAB substrate was used to develop signals.
   For TRAP staining, sections were deparaffinized, rehydrated, and
   stained with a TRAP staining kit (G1050, Servicebio). Images were
   scanned at ×100 magnification (Pannoramic 250 FLASH, 3DHISTECH), and
   OB/OC counts were performed using ImageJ software.

IF staining

   Femurs were fixed in 4% PFA and decalcified in EDTA solution before
   embedding in OCT compound and sectioning at 8 μm. Sections were stained
   with an anti-β-Catenin antibody. For OBs, cells were fixed in 4% PFA,
   permeabilized with 0.2% Triton X-100, and blocked with 3% BSA^[286]67.
   Sections were then incubated with primary antibodies: RUNX2 (1:100,
   12556S, Cell Signaling Technology), ALP (1:100, AB_2838191, Affinity
   Biosciences Cat# DF6225) and the sections were incubated with
   corresponding secondary antibodies conjugated to fluorescent dyes. DAPI
   were used to stain the nuclei. Images were captured using a
   fluorescence microscope.

ChIP-PCR analysis

   Cells (1 × 10^6) were used for each immunoprecipitation following a
   previously described protocol^[287]36. In brief, cells were crosslinked
   with 1% formaldehyde at RT in medium for 10 min, after which glycine
   was added to a final concentration of 0.125 M to quench the
   crosslinking and incubate at RT for 5 min. Cells were washed twice with
   PBS and lysed with SDS buffer (100 mM NaCl, 50 mM pH 8.1 Tris-HCl,
   5 mM pH 8.0 EDTA, 0.02% NaN[3], 0.5% SDS, PIC) at RT for 5 min. Cell
   lysates were harvested, and samples were thawed in a water bath to
   ensure complete dissolution of SDS. After centrifugation at 4 °C
   1200 × g for 6 min, the supernatant was discarded and 1 mL of
   prechilled IP buffer (a mixture of SDS buffer and Triton Dilution
   Buffer in a 2:1 ratio, supplemented with PIC) was added. Samples were
   then sonicated to produce 0.2–0.5 Kb DNA fragments, centrifuge
   sonicated chromatin 13,000 × g for 30 min.

   Equal amount of lysates was taken, and the volume was adjusted to 1 mL
   with IP buffer. Primary antibodies as listed in Supplementary
   Table [288]2 were added for IP overnight at 4 °C with rotation. Protein
   A/G beads were then added and incubated on a rotating wheel at 4 °C for
   2 h. Beads were washed once with 1 mL of 150 mM wash buffer (containing
   1% TritonX-100, 0.1% SDS, 150 mM NaCl, 2 mM EDTA pH 8.0, 20 mM Tris-HCl
   pH 8.0), and then twice with 500 mM wash buffer (containing 1%
   TritonX-100, 0.1% SDS, 500 mM NaCl, 2 mM EDTA pH 8.0, 20 mM Tris-HCl pH
   8.0). Finally, 120 μL of de-crosslinking buffer was added to both input
   and IP samples and incubated at 65 °C overnight at 1200 × g to elute
   the complexes from beads. QIAGEN PCR purification kit was used for
   ChIP-DNA purification, and the samples were quantified by real-time PCR
   using primers listed in Supplementary Table [289]1.

RNA isolation and qRT-PCR

   Total RNA was isolated using kit (M050, New Cell & Molecular Biotech)
   following the manufacturers’ instruction and reverse transcribed using
   HiScript III All-in-one RT SuperMix (R333-01, Vazyme). The resulting
   cDNA was used for PCR which was conducted in 10-μl reactions using SYBR
   Green Master Mix (Q711-03, Vazyme) with the LightCycler480 system. The
   relative expression of target genes was calculated using method of 2
   −ΔΔCT and calculated after normalization to the expression levels of
   rp0 in each sample. Primers used in qRT-PCR was listed in Supplementary
   Table [290]1.

Dual luciferase assay

   To detect relative luciferase activity of TOP reporter, 3T3-E1 cells
   were co-transfected with 100 ng TOP-FLASH plasmid harboring six
   TCF-binding motifs (Millipore), and 2 ng of the renilla luciferase
   control vector pGL4.74 (Promega). Following a 7-day incubation in OM,
   cell lysates were prepared using the lysis buffer from the TransDetect®
   Double-Luciferase Reporter Assay Kit (FR201, TransGen Biotech, China).
   Luciferase activity was subsequently quantified using a GloMax 20/20
   luminometer (Promega)^[291]68.

High-throughput screening with AutoDock Vina

   Compounds from ZINC database were subject to docking by using AutoDock
   Vina. Structure of PCGF1^RAWUL was taken from crystal structure of
   PCGF1^RAWUL –BCOR^PUFD complex (PDB code 4HPL). pdbqt file of
   PCGF1^RAWUL was generated usinig MGL Tools. PCGF1^RAWUL protein was
   prepared by adding hydrogen atoms and Kolloman charges. The pocket on
   PCGF1^RAWUL, that packs against two anti-parallel β strands on
   BCOR^PUFD, was defined as a potential ligand binding site. The grid box
   was generated using grid-box option from MGL Tools. Each ligand was
   docked using exhaustiveness value of 16, and energy_range value of 0.1.

Protein purification

   The cDNA encoding PCGF1^RAWUL L238A/F242A mutant (amino acids 166-255)
   was cloned into pGEX 6P-1 vector with N-terminal GST and hexa-histidine
   tag followed by PreScission Protease cleavage site. The cDNA encoding
   BCOR^PUFD (residues 1580-1696) was cloned into pET-28a vector with
   N-terminal hexa-histidine tag. Both PCGF1^RAWUL L238A/F242A mutant and
   BCOR^PUFD were expressed in E. coli BL21 (DE3) strain, respectively.
   Cultures were grown at 37 °C to OD[600] of 0.6–0.8 before induction
   with 0.5 mM IPTG, and incubated for an additional 20 hrs at 16 °C to
   promote protein expression.

   For the purification of PCGF1^RAWUL L238A/F242A mutant, cells were
   harvested and resuspended in lysis buffer (20 mM Tris pH 8.0, 1 M NaCl,
   7 mM β-mercaptoethanol, 5% glycerol). The recombinant protein was
   purified using Ni^2+ affinity chromatography, followed by cleavage of
   the GST and His tags with PreScission Protease. The mutant PCGF1 ^RAWUL
   was further refined by additional Ni^2+ affinity chromatography and gel
   filtration on a Superdex 75 (16/60) column pre-equilibrated with a
   buffer comprising 20 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, and
   0.5 mM TCEP. For the purification of His-BCOR^PUFD, cells were
   harvested and resuspended in lysis buffer (20 mM Tris pH 8.0, 300 mM
   NaCl, 7 mM β-mercaptoethanol, 5% glycerol). His-BCOR^PUFD was purified
   using a Ni^2+ affinity column with standard protocol, and followed by
   gel filtration on a Superdex 75 (16/60) column pre-equilibrated with a
   buffer comprising 20 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, and
   0.5 mM TCEP.

AlphaScreen assay

   The PCGF1^RAWUL L238A/F242A mutant was biotinylated using Biotinylation
   kit (G-MM-IGT, Genemore, Shanghai, China) according to manufacture’s
   instructions. A concentration of 200 nM biotinylated PCGF1^RAWUL
   L238A/F242A mutant was mixed with inhibitor at indicated concentration
   in the buffer containing 50 mM MOPS pH 7.4, 0.05 mM CHAPS, 50 mM NaF
   and 0.1 mg/mL BSA, and incubated for 15 min at room temperature. After
   adding 7.5 µg/mL Nickerl chelate beads, 7.5 µg/mL Streptavidin beads
   and 200 nM His-BCOR, the mixture was then incubated for additional
   1.5 h at 20 °C. Finally, the mixture was transferred to 384-well plate
   and analyzed by a EnVision 2105 (PerkinElmer). Data were plotted using
   GraphPad Prism software.

Biolayer interferometry assay (BLI)

   Biolayer interferometry equipment (Gator Bio) was used to determine the
   binding affinity between iBP and the PCGF1^RAWUL L238A/F242A mutant.
   Biotinylated PCGF1^RAWUL L238A/F242A at concentration of 50 μg/mL was
   immobilized on a SA XT biosensor. To remove non-specific bound, the
   biosensors were washed with assay buffer (50 mM MOPS pH 7.4, 0.05 mM
   CHAPS, 50 mM NaF, 0.1 mg/mL BSA and 0.5% DMSO). After obtaining a
   baseline reading in assay buffer, the biosensors were dipped into
   reference well or wells containing the various concentration of iBP for
   5 min. Then, the biosensors were washed with assay buffer for 2 min.
   Binding kinetics were analyzed using 1:1 binding model with on-board
   software. Data were plotted using GraphPad Prism software.

Stability assay of iBP

   iBP (Vitas-M, STL373868) stability (400 μM) was assessed using HPLC
   following incubation in DMEM supplemented with 10% FBS at 37 °C for 0,
   12, or 24 h. To facilitate the precipitation of proteins from FBS, the
   incubation mixture was diluted with a fourfold volume of acetonitrile
   and centrifuged. The resulting supernatant was then subjected to HPLC
   analysis to determine the remaining iBP concentration.

HPLC

   HPLC analysis for iBP was conducted using a Shimadzu LC-20AT system
   equipped with an SPD-20A UV–VIS detector. The chromatographic
   separation was achieved on a 4.6 × 150 mm Agilent Eclipse XDB-C18 5 µm
   column. The mobile phase consisted of solvent A (water with 0.1%
   trifluoroacetic acid) and solvent B (acetonitrile) at a flow rate of
   1.0 mL/min. The gradient elution program was as follows: 10% B
   (0–2 min), 10–100% B (2–16 min), 100% B (16–18 min), 100–80 % B
   (18–19 min) and 80% B (19–20 min).

Mouse PK Study

   A standard pharmacokinetic study was conducted on iBP using C57BL/6J
   mice (n = 3). The intraperitoneal injection formulation consisted of a
   4 mg/mL solution of iBP prepared with a ratio of 10% DMSO, 40% PEG300,
   5% Tween 80, and 45% Saline. The mice were administered iBP
   intraperitoneally at a dosage of 20 mg/kg, and plasma samples were
   collected at intervals of 5 min, 15 min, 30 min, 1, 2, 4, 8, and 24 h
   post-injection. The harvested supernatant was diluted with water at a
   1:2 ratio, and 2 µL of the diluted supernatant was subjected to
   LC/MS/MS for quantitative analysis.

Scanning electron microscope (SEM) and transmission electron microscopy (TEM)

   The surface structure of the F127 hydrogel (20% W/V) was examined using
   SEM^[292]69. Briefly, samples were freeze-dried using an LGJ-12A freeze
   dryer (China), gold-coated, and imaged with a Zeiss Gemini 300 SEM. The
   morphology of iBP was observed under a HT7700 TEM (Hitachi High-Tech,
   Tokyo, Japan). The size and size distribution of iBP particles were
   further analyzed using a Zetasizer Nano ZS automatic particle size
   detector (Malvern, UK).

In vitro drug release assay

   The release profile of iBP from the F127 hydrogel-gelatin sponge was
   assessed using a dialysis membrane with a molecular weight cut-off of
   3500 Da. Sponges loaded with either 2 mg/mL (100 μL) or 4 mg/mL
   (100 μL) of iBP (measuring 2 × 2 × 10 mm^3) were placed in 100 mL of
   PBS at 37 °C with continuous stirring at 100 × g. At each time point,
   1 mL of the solution was removed and replaced with 1 mL of PBS to
   maintain a constant volume. The concentration of iBP in the solution
   was determined using a standard curve established at OD[300] nm,
   recorded with a U-3310 spectrophotometer (Hitachi High-Tech, Tokyo,
   Japan).

Mouse model construction and In vivo drug administration

   For establishing OVX models, mice were randomly assigned to either the
   OVX or Sham group. Under isoflurane inhalation anesthesia, a lateral
   dorsal skin incision was made to expose the ovaries. In the OVX group,
   the ovaries were removed after ligating with bilateral sutures, while
   in the Sham group, the ovaries were exposed but not removed. Penicillin
   was administered postoperatively to prevent infection, and the muscle
   and skin incisions were closed.

   To evaluate the therapeutic efficacy of iBP in countering bone loss,
   3-month-old virgin female mice were subjected to either ovariectomy or
   sham surgery (n = 4 per group). A subset of ovariectomized mice
   received daily intraperitoneal injections of iBP at a dose of 20 mg/kg,
   starting two days post-surgery and continuing for four weeks. A
   comparative group of ovariectomized mice received PBS injections under
   the same conditions.

   To evaluate the pharmacological effect of iBP on bone healing, a
   unilateral circular bone defect was created in the right femur of CON
   and CKO mice. Using a low-speed dental drill with a 1 mm diameter bit,
   a defect was made in the femur. A gelatin sponge, dimensioned at
   2 × 2 × 10 mm^3 and preloaded with 0.2 mg of iBP in F127 hydrogel, was
   positioned onto the defect site and secured with sutures. Two weeks
   later, the healing progress in both treated and untreated groups was
   monitored using μCT imaging.

Statistical analysis

   Statistical analysis One- or two-way ANOVA followed by Sidak’s test for
   multiple com parisons (with repeated measures for time series data) was
   used in all studies, as noted in figure legends. For comparison between
   two groups, Student’s t tests were performed. All tests used the
   software GraphPad Prism (GraphPad). The exact p-value is provided in
   the corresponding figure. For data presented without statistics, the
   experiment was repeated at least twice to ensure reproducibility,
   unless otherwise stated.

Reporting summary

   Further information on research design is available in the [293]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [294]Supplementary Information^ (4.6MB, pdf)
   [295]Peer Review file^ (875KB, pdf)
   [296]41467_2025_59638_MOESM3_ESM.pdf^ (90.3KB, pdf)

   Description of Additional Supplementary Files
   [297]Supplementary Data 1^ (108.5KB, xlsx)
   [298]Supplementary Data 2^ (332.7KB, csv)
   [299]Supplementary Data 3^ (461KB, xlsx)
   [300]Supplementary Data 4^ (435.6KB, csv)
   [301]Reporting Summary^ (137.2KB, pdf)

Source data

   [302]Source Data^ (11.1MB, zip)

Acknowledgements