Introduction

Recent genome-wide transcriptome analyses revealed that over 75% of the human genome is transcribed, among which about 95% are transcripts without coding capacity. Long noncoding RNAs (IncRNAs) emerge as a relatively new category of non-coding RNAs (≥ 500 nucleotides ¹). It is now widely accepted that lncRNAs can perform diverse regulatory roles in regulation of gene expression. lncRNAs are enriched in the nucleus and/or cytoplasm and possess the ability to interact with versatile biomolecules, including various proteins, chromosomal DNAs and RNAs. Therefore, in contrast to miRNAs, lncRNAs regulate gene expression and function by diverse mechanisms, such as functioning as scaffolds for transcriptional and chromatin-modifying complex assemblies, as enhancers or decoys regulating gene transcription, and as cis-acting or trans-acting regulators involved in gene expression and epigenetic regulation ^2–5. Importantly, lncRNAs are druggable targets, and identification of the functional importance of lncRNAs has unveiled new diagnostic and therapeutic opportunities for human diseases, such as cancer, cardiovascular diseases and genetic disorders ^2,6–13. Understanding the biological importance and clinical relevance of lncRNAs is at the forefront of RNA biology research. Nonetheless, only a small fraction of lncRNAs have well-established identifications. Most lncRNAs lack functional annotations, particularly with genetic evidence in vivo and convincing mechanistic studies.

Malat1 (metastasis associated lung adenocarcinoma transcript 1) is a highly evolutionarily conserved and abundant nuclear lncRNA. It is a 6.8 kilo-nucleotide RNA Polymerase II transcript from mouse chromosome 19 (8.9 kilo-nucleotide from human chromosome 11). As one of the first discovered lncRNAs, Malat1 was initially recognized as a gene showing specific upregulation in metastatic non-small-cell lung cancer cells¹⁴. Subsequent studies found a variety of associations between Malat1 and the growth and metastasis of different cancers, such as lung cancer, hepatocellular carcinoma, and breast cancer ^15,16. In physiological settings, although Malat1 was reported to be a key component of nuclear speckle and a regulator of alternative pre-mRNA splicing in in vitro studies, Malat1 knockout (KO) mice surprisingly appeared to have no defects in nuclear speckle assembly or pre-mRNA splicing ^17–21, raising questions about the extent to which Malat1 contributes to these processes in vivo. These unexpected discrepancies of in vitro and in vivo studies also draw increasing attention on the importance and necessity of in vivo studies to reveal lncRNA function. The homeostatic/physiological function of Malat1 in vivo has continued to be an enigma, because the absence of this abundant lncRNA in mice does not seem to exhibit abnormalities ^17–19. However, to the best of our knowledge, there are no studies that have comprehensively characterized bone phenotype of Malat1 KO mice.

Bone is a vital organ, which plays a fundamental role in providing structural support to the body, enabling movement, and protecting many other organs. Bone homeostasis is crucial to the quality of life and overall well-being. Bone homeostasis in adulthood is mainly maintained by an active bone remodeling process, which requires a delicate balance between osteoclast-mediated bone resorption and osteoblast-mediated bone formation. Bone tissues undergo constant remodeling, during which bone resorption and formation are usually coupled to ensure that osteoclast-generated resorption lacunae are filled with new bone produced by osteoblasts. This coordination helps maintain bone homeostasis and also provides a mechanism for adapting the skeleton to environmental changes and repairing bone damage. There exists a variety of crosstalk between two major bone cell types, bone-resorbing osteoclasts and boneforming osteoblasts, as well as other cells. The intricate cellular crosstalk coordinately couples the activities of different cells to maintain bone homeostasis during remodeling ^22,23. In pathological conditions, bone remodeling is often deregulated, which results in unbalanced bone resorption and formation. For example, excessive osteoclast formation accompanied by extensive bone resorption but with limited bone formation/repair often occurs in rheumatoid arthritis (RA), periodontitis and osteoporosis. On the other hand, excessive bone formation and/or defective bone resorption result in osteopetrosis or osteosclerosis. Notably, unbalanced activities between osteoclasts and osteoblasts in pathological bone remodeling synergize to aggravate the rate and extent of bone damage ^24–27. Although the importance of bone remodeling is evident to bone homeostasis and health, the coupling and crosstalk mechanisms are complex and far from well understood.

Given the vital importance of bone remodeling to skeletal health and overall well-being, we took advantage of rigorous genetic approaches using global and conditional Malat1 KO mice in this study. Contrary to previously held beliefs that Malat1 had no appreciable phenotype, we uncovered that Malat1 KO mice exhibit a significant osteoporotic bone phenotype characterized by reduced osteoblastic bone formation and enhanced osteoclastic bone resorption in vivo. Thus, Malat1 deletion uncoupled the normal bone remodeling process between osteoblasts and osteoclasts. Our data further demonstrate that Malat1 emerges as a novel regulator impacting multiple cell types, including osteoblasts, osteoclasts, and chondrocytes, through the β-catenin-OPG/Jagged1 pathway. This study discovered an important homeostatic function of Malat1, and identified this lncRNA as a previously unrecognized key bone remodeling regulator that controls both bone formation and resorption to maintain bone homeostasis, regeneration and health.

Results

Malat1 is a novel lncRNA regulator of bone homeostasis and remodeling

We first performed a series of experiments to comprehensively examine bone mass, bone resorption, bone formation, bone cell changes and bone remodeling by microCT (μCT), bone histology and dynamic histomorphometric analysis. Adult mice with Malat1 deficiency (Malat1^-/- on C57BL/6 background) do not show growth retardation or macroscopic differences (Suppl. Fig. 1a), but exhibit significantly reduced bone mass compared with wild type littermate controls (WT) (Fig.1a). μCT analysis showed markedly decreased trabecular bone volume (BV/TV), number (Tb.N.), bone mineral density (BMD) and connectivity density (Conn-Dens.), and increased trabecular spacing (Tb. Sp) in both male and female Malat1^-/- mice (Fig.1b and Suppl. Fig. 1b). Cortical bone appears normal (Suppl. Fig. 1c). Osteoclastic bone resorption was significantly enhanced in Malat1^-/- mice indicated by the elevated numbers and surfaces of osteoclasts (Fig. 1c, d). However, osteoblastic bone formation was not accordingly enhanced but instead was dramatically suppressed by Malat1 deficiency (Fig. 1e-h). Bone dynamic histomorphometric analysis by calcein double labeling showed that Malat1 absence notably inhibited both mineral apposition rate (MAR) and bone formation rate (BFR/BS) in Malat1^-/- mice (Fig. 1e, f). Furthermore, osteoblast parameters, such as osteoblast surfaces and numbers, as well as osteoid formed beneath osteoblasts were significantly reduced in Malat1^-/- mice (Fig.1g, h). These results indicate that the lack of Malat1 suppresses osteogenesis and bone formation in mice. Moreover, Malat1 deficiency uncouples osteoclastic bone resorption and osteoblastic bone formation, leading to markedly reduced bone mass. Our data thus identify Malat1 as a novel bone remodeling regulator in bone homeostasis.

Fig. 1

Malat1 acts as a cell-intrinsic regulator in osteoblasts to promote bone formation

Next, we examined the role of Malat1 in osteoblastic bone formation. We crossed Malat1^flox/flox mice with the mice expressing Cre under the control of Osteocalcin (Ocn) promoter to generate Malat1 conditional KO mice specifically in osteoblasts (Malat1^flox/floxOcnCre(+); hereafter referred to as Malat1^ΔOcn). Malat1 deficiency in osteoblasts significantly decreased trabecular bone volume (BV/TV), number (Tb.N.), thickness (Tb. Th), bone mineral density (BMD) and connectivity density (Conn-Dens.), and increased trabecular spacing (Tb. Sp) in Malat1^ΔOcn mice (Fig. 2a, b). There are no obvious changes in cortical bone phenotype (Suppl. Fig. 2). Furthermore, mineral apposition rate (MAR) and bone formation rate (BFR/BS) were both lower in Malat1^ΔOcn mice than the controls (Fig. 2c). In parallel, Malat1 deficiency in osteoblasts led to decreased osteoblast numbers and surfaces in vivo (Fig. 2d). These results indicate that Malat1^ΔOcn mice exhibit osteoporotic phenotype with reduced bone formation, which is consistent with Malat1^-/- mice. Therefore, Malat1 is a cell-intrinsic osteogenic regulator that promotes osteoblastic bone formation.

Fig. 2

Malatl binds to β-catenin and suppresses its transcriptional activity in osteoblasts

Given these findings, we sought to investigate the mechanisms underlying the regulation of osteoblastic bone formation by Malat1. β-catenin is a central transcriptional factor in canonical Wnt signaling pathway, and plays an important role in positively regulating osteoblast differentiation and function ^28–33. Upon stimulation, most notably from canonical Wnt ligands, β-catenin is stabilized and translocates into the nucleus, where it interacts with coactivators to activate target gene transcription. Previous reports observed a link between Malat1 and β-catenin signaling pathway in cancers ^34,35, but the underlying molecular mechanisms in terms of how Malat1 interacts with β-catenin and regulates its nuclear retention and transcriptional activity are unclear. In this study, we performed both Chromatin isolation by RNA purification (ChIRP) assay (Fig. 3a, b, Supplementary Fig. 3) and RNA immunoprecipitation (RIP) assay (Fig. 3c) to determine the interaction between Malat1 and β-catenin. The results obtained by these two approaches clearly show that Malat1 binds to β-catenin in osteoblasts (Fig. 3a-c). Next, we asked whether Malat1 regulates β-catenin nuclear translocation in response to Wnt3a. We did not find that Malat1 deficiency significantly affects nuclear localization of β-catenin stimulated by Wnt3a (Fig. 3d, Supplementary Fig. 4). We then performed Luciferase assay to examine whether Malat1 modulates the transcriptional activity of β-catenin. The results showed that Malat1 deficiency significantly reduced the transcriptional activity of β-catenin in response to Wnt3a stimulation (Fig. 3e). In line with this, the expression levels of β-catenin target genes, such as Axin2, Ccnd1, Lef1 and Myc, were substantially lower in Malat1^-/- cells than WT controls (Fig. 3f). Our findings indicate that Malat1 is a key regulator of Wnt/β-catenin signaling pathway that is important for osteoblasts. Since Malat1 is a nuclear lncRNA, these data also suggest that Malat1 acts as a scaffold to tether β-catenin in nuclei to implement its positive regulation of β-catenin activity.

Fig. 3

Malat1 inhibits osteoclastogenesis in a non-autonomous manner

As osteoclast formation is significantly increased in Malat1^-/- mice, we examined whether this osteoclast change is due to an intrinsic role of Malat1 in these cells. We first investigated in vitro osteoclast differentiation using bone marrow derived macrophages (BMMs) as osteoclast precursors. Surprisingly, there is no significant difference in osteoclast differentiation and mineral resorption between WT and Malat1^-/- BMM cultures (Fig. 4a, b), which is inconsistent with the in vivo data (Fig. 1c). Moreover, the expression of osteoclastogenic transcription factors and osteoclast marker genes was similar between WT and Malat1^-/- cultures (Fig. 4c, d). We also generated Malat1^flox^floxLysMCre(+) (Malat1^ΔM/ΔM) mice, a myeloid lineage osteoclast specific Malat1 conditional KO mouse line. There is no significant difference in bone mass between Malat1^ΔM/ΔMmice and their littermate control LysMCre(+) mice (Fig. 4e, f). These results demonstrate that Malat1 expressed in osteoclastic lineage cells does not affect osteoclastogenesis. Therefore, Malat1 inhibits osteoclastogenesis in a non-autonomous manner in vivo.

Fig. 4

Malat1 couples osteoblast-osteoclast crosstalk via β-catenin-OPG axis

It is intriguing how Malat1 regulates osteoclastogenesis. We first looked at osteoclast phenotype in Malat1^ΔOcn mice, and found that Malat1 deletion in osteoblasts significantly enhanced osteoclast numbers and surfaces in vivo (Fig. 5a). This finding indicates that Malat1 affects osteoclast formation through crosstalk between osteoblasts and osteoclasts. To further explore the underlying mechanisms of this crosstalk, we took advantage of a well-established co-culture system ³⁶ with primary osteoblasts in a transwell and bone marrow cells on the bottom of the dish. RANKL and M-CSF secreted by osteoblasts induced osteoclast differentiation on the bottom (Fig. 5b). Interestingly, we found that more osteoclasts formed in the co-cultures with Malat1^-/- osteoblasts than WT osteoblasts (Fig. 5b). The results of this co-culture system recapitulate the in vivo enhanced osteoclast phenotype, and indicates that certain soluble factors secreted from osteoblasts are highly likely involved in the Malat1 regulation of osteoclastogenesis. Since we found that Malat1 binds to β-catenin to regulate its target genes, we primarily searched β-catenin target genes that function as secreted factors. Osteoprotegerin (OPG), encoded by Tnfrsf11b, is a β-catenin target ^37,38. OPG is a secreted factor that acts as a RANKL decoy receptor to block RANKL activity, thereby suppressing osteoclast formation and bone resorption ³⁹. We found that OPG expression was markedly lower in the Malat1^-/- osteoblasts than the WT control cells (Fig. 5c). RANKL-OPG axis is a typical bone remodeling mediator through impacting osteoclastogenesis, and RANKL/OPG ratio is usually an indicator for osteoclastogenic/bone resorption potential and status ⁴⁰. Our results showed that OPG but not RANKL was down-regulated in Malat1^-/- osteoblasts, leading to a higher RANKL/OPG ratio in Malat1^-/- osteoblasts than WT controls (Fig. 5d, e). The elevated RANKL/OPG ratio in Malat1^-/- osteoblasts favors enhanced osteoclastogenesis, which is aligned with the enhanced osteoclast formation and decreased bone mass in Malat1^-/- mice. We next examined the OPG expression in vivo. Recent literature demonstrates that locally produced OPG in bone, but not the serum OPG, is a critical inhibitor of osteoclast formation and bone resorption ⁴¹. Indeed, the OPG level in serum in Malat1^-/- mice is similar to that in WT controls (Fig. 5f). However, OPG level in bone marrow was drastically decreased in Malat1^-/- mice compared to WT mice (Fig. 5g). We further tested the importance of the decreased OPG level in the Malat1 regulation of osteoclasts. We applied two well-established ex vivo culture systems to closely recapitulate in vivo conditions. In the whole bone marrow cultures with M-CSF and RANKL (Fig. 5h), Malat1 deletion in Malat1^-/- bone marrow enabled more osteoclastogenesis than WT bone marrow, which is consistent with the in vivo findings. When recombinant OPG was added to the bone marrow cultures, the osteoclast formation in both WT and Malat1^-/- cultures was inhibited as expected, but the osteoclast formation in Malat1^-/- bone marrow with additional OPG became similar to that in the WT control cultures without additional OPG (Figure 5h column 1 vs 4). In another co-culture system (Fig. 5i), WT or Malat1^-/- osteoblasts were co-cultured with WT bone marrow in the presence of 1,25(OH)2-vitD3 and PGE2. The RANKL secreted from osteoblasts induces osteoclast formation in this co-culture system. Furthermore, because these cultures included the same WT bone marrow cells but different osteoblasts from WT or Malat1^-/- mice, the results directly reflected osteoblastic Malat1 effects on osteoclast formation. In this co-culture system, we observed a similar phenotype as that in Fig. 5h. Extra OPG can lead to osteoclast formation in Malat1^-/- osteoblasts and WT bone marrow cocultures comparable to that in WT osteoblasts and WT bone marrow cocultures without extra OPG (Fig. 5i column 1 vs 4). These results collectively support that the decreased OPG level in Malat1 deficient osteoblasts plays a significant role in the excessive osteoclast formation in Malat1^-/-’ mice (Fig. 1c, d) and Malat1^ΔOcn mice (Fig. 5a), as illustrated in Fig. 7.

Fig. 5

Malat1 positively regulates the production of β-catenin targets, OPG and Jagged1, from chondrocytes

As OPG is a key osteoclastic inhibitor, we asked whether other cells, in addition to osteoblasts, in bone marrow could also produce OPG, which we have established is regulated by Malat1. We analyzed a bone scRNAseq dataset (GSE128423) ⁴². Except for osteoblasts as expected, we surprisingly found that chondrocytes express a high level of OPG in the analyzed cells from bone (Fig. 6a-d, Supplementary Fig. 5). We then isolated articular cartilage from mouse knees and digested it to obtain primary chondrocytes, which highly express chondrocyte marker genes, such as Sox9, Acan and Col2a1 (Fig. 6e, Supplementary Fig. 6b). These cells were also Alcian blue positive (Supplementary Fig. 6a). We confirmed OPG expression in chondrocytes isolated from mice (Fig. 6g). Moreover, OPG expression level in Malat1^-/- chondrocytes was approximately 50% less compared to that in WT cells (Fig. 6f, g). RANKL (encoded by Tnfsf11) is nearly undetectable in chondrocytes (Fig. 6f). The chondrocyte marker genes, including Sox9, Acan and Col2a1, were not affected by Malat1 (Fig. 6e). These data show that chondrocytes are a previously unrecognized cellular source of OPG in bone, which is critically regulated by Malat1.

Fig. 6

We next examined whether Malat1 also modulates β-catenin activity in chondrocytes. In addition to OPG, we observed that the expression of other β-catenin target genes, such as Tcf7, Lef1 and Jag1, was approximately 50% lower in Malat1^-/- chondrocytes than that in WT cells (Fig. 6h). This suggests that, similar to osteoblasts, Malat1 positively regulates β-catenin activity in chondrocytes. The decreased expression of Jagged1 in Malat1^-/- chondrocytes drew our attention, as Jagged1 is not only a Notch ligand ⁴³ but also an inhibitor of osteoclastogenesis ^44,45. Following this observation, we evaluated the protein expression levels of Jagged1, and found that Malat1 deficiency drastically decreased Jagged1 protein expression in chondrocytes (Fig. 6i). These results collectively indicate that the decreased OPG and Jagged1 production in Malat1^-/- chondrocytes also contributes to the enhanced osteoclast formation and low bone mass in Malat1^-/- mice. Thus, Malat1 links the activities of chondrocytes and osteoclasts through β-catenin-OPG/Jagged1 axis (Fig. 7).

Fig. 7:

Malat1 promotes bone regeneration in fracture healing

We next examined whether Malat1 impacts osteogenesis in pathological conditions. In a femoral midshaft fracture model ⁴⁶, we found that Malat1 deficiency in both Malat1^-/- mice and Malat1^ΔOcn mice significantly disturbed the bone regeneration at 21 days post-fracture, indicated by significantly decreased bone mass in the newly formed callus (Fig. 8). These results indicate that Malat1 in osteoblasts plays an important role in enhancing bone regeneration in fracture healing.

Fig. 8:

Discussion

Malat1 stands out as one of the most abundant and evolutionarily conserved long noncoding RNAs (lncRNAs). Initially, the lack of evident defects in Malat1 knockout (KO) mice led to the assumption that Malat1 might not be essential for development and physiological processes. However, previous phenotyping studies did not encompass the examination of bone, a vital yet often overlooked organ. Despite its appearance as a hard tissue, bone undergoes continuous remodeling involving various cell types to maintain its mass and function. In this study, we investigated the function of Malat1 in bone. Our results revealed that Malat1 acts as a crucial player through the concerted actions of multiple bone cells via a new molecular network in the intricate regulation of bone homeostasis and remodeling under physiological conditions (Fig. 7). In pathological conditions, such as fracture healing, Malat1 is an important regulator that promotes bone regeneration. These findings exemplify the unprecedented key role of lncRNAs, like Malat1, in tissue homeostasis and remodeling, shedding light on their broader significance in orchestrating cellular and molecular networks for maintaining and adapting tissue homeostasis and regeneration.

During the submission of our manuscript, another group ⁴⁷ reported that Malat1 KO mice show reduced bone mass, which is the only similar phenotype as observed in our study. However, there are many novel findings that are distinct in our study. In addition to global Malat1 KO mice, we generated both osteoblast and osteoclast specific Malat1 conditional KO mouse lines. Using the osteoblast-specific conditional KO mice (Malat1^ΔOcn), we demonstrate that Malat1 promotes bone formation in osteoblasts. Using the myeloid osteoclast-specific mice (Malat1^ΔM/ΔM), we demonstrate that Malat1 does not directly affect osteoclastogenesis, but suppresses osteoclastogenesis in a non-autonomous manner in vivo by integrating crosstalk between multiple cell types, including osteoblasts, osteoclasts and chondrocytes. We further found that Malat1 binds to β-catenin and functions through the β-catenin-OPG/Jagged1 pathway in osteoblasts and chondrocytes to enhance bone formation and suppress bone resorption (Fig. 7). Moreover, Malat1 significantly promotes bone regeneration in fracture healing. Taken together, these findings, including the cellular and molecular mechanisms as well as the healing model, are novel and not reported before.

Fine-tuning in biology is essential for maintaining homeostasis ^48,49. Bone must adapt to various environmental changes, ensure optimal function, and respond effectively and precisely to internal and external signals. In this study, we found that Malat1 is a central player in finetuning bone homeostasis and remodeling. Malat1 promotes osteoblastic bone formation. It also orchestrates the β-catenin pathway to activate the downstream target genes, OPG and Jagged1, which are potent osteoclastogenic inhibitors. Interestingly, Malat1 does not directly affect osteoclastogenesis in osteoclast cell lineage. The regulation of OPG and Jagged1 by Malat1 occurs in osteoblasts and chondrocytes. Malat1 alters the expression of OPG and Jagged1 in these cells, thereby impacting osteoclastogenesis in a non-autonomous manner. Therefore, Malat1 impacts not only one cell type but fine-tunes a complex cellular and molecular network in the skeletal system. In this network, OPG and Jagged1 are newly identified targets regulated by Malat1. Although previous reports observed a link between Malat1 and β-catenin signaling pathway in cancers ^34,35, the underlying molecular mechanisms were unclear. Specifically, it is poorly understood how Malat1 interacts with β-catenin and regulates its nuclear retention and transcriptional activity. Our studies elucidated these questions and uncovered the molecular basis of the Malat1- β-catenin pathway. With consideration that lncRNAs can utilize diverse molecular mechanisms, Malat1 may also bind to additional regulators. Future studies may involve ChIRP assays followed by mass spectrometric analysis to identify additional potential Malat1-bound targets.

While studies on lncRNAs using in vivo genetic approaches are increasing, many lncRNA investigations still rely heavily on in vitro methods, such as RNAi techniques for a knockdown in cell cultures. While a knockdown model has the advantage of facilitating high-throughput screens, emerging evidence has revealed significant non-specific or off-target effects in vitro. This has caused subsequent mischaracterizations of lncRNA functions and their respective mechanisms. This is particularly concerning for abundant nuclear lncRNAs, such as Malat1. Additional concerns include the transient knockdown effects in vitro versus the stable knockout effects in vivo. The definitive evidence of functionality comes from genetic knockouts, enabling the examination of in vivo function and minimizing the risk of off-target effects. Controversial reports exist between in vitro and in vivo studies regarding Malat1’s function in nuclear speckles and cancers ^{17–21,50–53}. Recently, some in vitro studies suggested that Malat1 acts as a miRNA sponge ⁵⁴. However, miRNAs are primarily located in the cytoplasm and lncRNA-miRNA interaction occurs almost exclusively in this cellular compartment. Given Malat1’s nuclear localization, it is unlikely to function as an ‘miRNA sponge’. At least, the location of the potential interaction between Malat1 and miRNAs, as well as the Malat1 copies in extra-nuclear compartments, need to be definitively verified using robust approaches. Current reports on Malat1’s functions in bone cells have similar concerns with their in vitro approaches ^51–53. Overall, the distinct results obtained from in vitro knockdown systems and in vivo Malat1 knockout mouse models provide a clear example of the limitations of in vitro knockdown systems for annotating lncRNA function in vivo. Therefore, discussions in the lncRNA field have emphasized the robustness of approaches, and highlighted the critical need to apply genetic methods aptly to confidently characterize lncRNA functions ^17,19,50. This is especially relevant in in vivo contexts such as development, metabolism, homeostasis, and tissue remodeling. The Malat1^-/- mice ¹⁷ used in this study were generated by inserting beta-galactosidase gene followed by a polyadenylation signal immediately downstream from the transcriptional start site of Malat1. This strategy preserves DNA regulatory elements at the Malat1 locus, thereby avoiding confusion in result interpretation arising from the loss of these elements.

A challenge in lncRNA field is the unclear structures of most lncRNAs. The size of lncRNAs typically spans from 1kb to over 100kb. These molecules exhibit intricate secondary and tertiary structures that become increasingly dynamic and complex based on their interactions with various proteins. The structural flexibility of lncRNAs in diverse cellular contexts poses a more significant challenge in deciphering the relationship between their structure and function ¹. For instance, in a model of estimating lncRNA structures, approximately half of Malat1 nucleotides were found to adopt various structures, including nearly 200 helices, many pseudoknots, structured tetraloops, internal loops, and intricate intramolecular long-range interactions featuring multiway junctions ⁵⁵. To unravel these complexities, breakthrough techniques are required to understand the functions of each and all of the dynamic structures of lncRNAs.

Many skeletal diseases involve defects in bone remodeling, a process that includes multiple cell types, cellular crosstalk, and complex molecular pathways. Treatment efficacy is usually insufficient by targeting only one bone cell type. This is clearly evident in diseases such as RA and periodontitis, in which standard antiresorptive therapies, such as bisphosphonates that inhibit osteoclast activity, are not able to effectively restore bone formation ^24–27. Therefore, there is a clinical need for new or complementary therapies that can target multiple bone cell types, such as osteoblasts and osteoclasts. These innovative therapeutic strategies go beyond targeting a single cell type, aiming instead to restore healthy bone remodeling and significantly enhance treatment efficacy. Malat1, identified in this study, emerges as a crucial fine-tuner that integrates multiple bone cells and a molecular network to maintain bone homeostasis. Thus, Malat1 and its mediated cellular and molecular mechanisms not only hold significant implications for our understanding of conditions related to bone health but also offer a potential avenue for developing more efficient treatments for bone diseases.

Materials and Methods

Animals

Malat1^{-/- 17} and Malat1^flox/flox mice ¹⁹ were described previously. Sex- and age-matched Malat1^-/- mice and their littermates WT (Malat1^+/+) mice were used for experiments. We generated mice with osteoblast-specific deletion of Malat1 by crossing Malat1^flox/flox mice with osteocalcin cre mice (The Jackson Laboratory, Stock No: 019509). Sex- and age-matched Malat1^flox/flox ;Ocn-Cre mice (referred to as Malat1^ΔOcn) and their littermates Malat1^flox/flox mice as the controls (referred to as the Malat1^f/f) were used for experiments. We generated mice with myeloid/macrophage-specific deletion of Malat1 by crossing the Malat1^flox/flox mice with LysM Cre mice (The Jackson Laboratory, Stock No: 004781). Sex- and age-matched Malat1^flox/flox; LysM-Cre mice (referred to as the Malat1^ΔM/ΔM mice) and their littermates with Malat1^+/+; LysMcre (+) genotype as WT controls (hereafter referred to as Control) were used for experiments.

After sacrifice, the bones were fixed by 4% formaldehyde and subjected to μCT analysis, sectioning, TRAP staining and histological analysis. μCT analysis of femoral trabecular bones and cortical midshaft was conducted to evaluate bone volume and 3D bone architecture using a Scanco μCT-35 scanner (SCANCO Medical) according to the manufacturer’s instructions and the American Society of Bone and Mineral Research (ASBMR) guidelines ⁵⁶.

For dynamic histomorphometric measures of bone formation ⁵⁷, calcein (25 mg/kg, Sigma) was injected into mice intraperitoneally at 5 and 2 days before sacrifice to obtain double labeling of newly formed bones. The non-decalcified tibia bones were embedded in methyl methacrylate. 5 mm thick sections were sliced using a microtome (Leica RM2255, Leica Microsystems, Germany). For static histomorphometric measures of osteoblast parameters, non-decalcified sections of the tibiae were stained using toluidine blue or Masson-Goldner staining kit (MilliporeSigma). The Osteomeasure software was used for bone histomorphometry using standard procedures according to the program’s instruction.

All mice were housed in a 12h:12h light/dark cycle with food and water ad libitum. All animal procedures were performed according to the approved protocol (2016—0001 and 0004) by the Institutional Animal Care and Use Committee (IACUC) of Hospital for Special Surgery and Weill Cornell Medical College.

Bone fracture model

Bone fracture was performed as described previously with slight modifications ⁴⁶. 13-week-old mice were anesthetized with 2.5% isoflurane via inhalation. The mice received meloxicam (subcutaneous injection, 2 mg/kg) and buprenorphine (subcutaneous injection, 0.5mg/kg) as analgesia prior to surgery. The surgical site was shaved and sterilized using iodine and 70% ethanol. An incision was made over the anterolateral femur. 0.1 ml bupivacaine (5mg/ml) was injected into the tissue adjacent to the incision line. A 25-gauge needle was inserted into the femoral canal through the patellar groove. The mid-diaphysis of the femur was transected using dental drill with a burr (19007-05, Fine science tools). The needle was then trimmed from the distal end to prevent it from projecting into the femoral joint space. The left part of the needle remained in femur to stabilize the fracture. The muscle was repositioned over the injury site and stitched using absorbable sutures. The skin was closed with wound clips, which were removed 2 weeks post-surgery. Animals were administered 0.5mg/kg buprenorphine every 12 hours and 2.0 mg/kg meloxicam every 24 hours subcutaneously for analgesia up to 3 days after the surgery. Mice were sacrificed 21 days post-fracture. All surgical procedures were performed according to the approved protocol (2016—0001/4) by the Institutional Animal Care and Use Committee (IACUC) of Hospital for Special Surgery and Weill Cornell Medical College.

Reagents

Murine M-CSF (Cat# 315–02), recombinant human sRANK Ligand (Cat# 310–01), recombinant human OPG (Cat# 450-14) and murine Wnt3a (Cat# 315-20) were purchased from PeproTech. Leukocyte Acid Phosphatase (TRAP) Kit (Cat# 387A) and Mouse Osteoprotegerin ELISA Kit (Cat# RAB0493) were obtained from MilliporeSigma. The plasmid of M50 Super 8x TOPFlash (Cat# 12456) was purchased from Addgene. pRL-Tk control plasmid (Cat# E2241) was purchased from Promega. Fetal Bovine Serum (FBS, Cat #S11550) was obtained from Atlanta Biologicals. β-glycerophosphate disodium salt hydrate (Cat# G9422), L-ascorbic acid (Cat# A4544), prostaglandin E2 (PGE2) (Cat# P0409) and 1α,25-Dihydroxyvitamin D3 (VitD3) (Cat# D1530) were obtained from MilliporeSigma.

Cell culture

For osteoclastogenesis, bone marrow cells were obtained from age and sex-matched WT and Malat1^–’ littermates and cultured in aMEM supplemented with 10% FBS and 2.4 mM glutamine (25030081, Thermo Fisher Scientific) and 1% Penicillin—Streptomycin along with CMG14–12 supernatant ⁵⁸, serving as the condition medium (CM) which contained a concentration equivalent to 20 ng/ml of rM-CSF and was utilized as the M-CSF source. After 3-day culture, the cells were washed in 1x PBS and the attached cells were scraped, and seeded into plates at a density of 4.5 × 10⁴/cm² with CM. Next day, the cells were induced for osteoclastogenesis with 40 ng/ml RANKL and CM for the indicated days shown in figures. Multi-nucleated osteoclasts were stained using Leukocyte Acid Phosphatase (TRAP) Kit (Cat# 387A, MilliporeSigma) according to the manufacturer’s instructions.

Primary osteoblastic cells were isolated from the calvaria of new-born (0–3d) mice by enzymatic digestion in 10% FBS αMEM with 0.1% collagenase (LS004177, Worthington) and 0.2% dispase (17105041, Thermo Fisher Scientific) as described ⁵⁷. The cells were used immediately or cultured to expand for 6 days for experiments indicated in relevant figures.

For the co-cultures of primary osteoblasts and bone marrow cells in transwell plates, primary calvarial osteoblasts isolated from WT and Malat1^-/- newborn mice were seeded in the upper chamber (2 × 10⁴ cells/96-well) in αMEM medium with 10% FBS. After primary osteoblasts reached 60–70% confluence, bone marrow cells harvested from the femur and tibia of 6-week-old mice were plated to the bottom chamber (1 × 106 cells/24-well), together with 1 μM Prostaglandin E2 (PGE2) (P0409, MilliporeSigma) and 10nM 1α,25-Dihydroxyvitamin D3 (VitD3) (D1530, MilliporeSigma). Culture media were exchanged every two days for 12 days. TRAP staining was performed and multinucleated osteoclasts were counted.

For the direct co-cultures of primary osteoblasts and bone marrow cells, primary osteoblasts isolated from WT and Malatl^-/- newborn mice were seeded in 48-well plates at a density of 2 × 104 cells/well in αMEM supplemented with 10% FBS. After osteoblasts reached 60–70% confluence, murine bone marrow cells isolated from the femur and tibia of 6-week-old mice were plated at a density of 2 × 105 cells/well on the top the osteoblasts in the presence of 1 μM PGE2 and 10nM VitD3. Culture media were exchanged every two days. On the sixth day, TRAP staining was performed and multinucleated osteoclasts were counted.

Preparation and culture of primary mouse articular chondrocytes were performed as previously described with slight modification ⁵⁹. In brief, femoral condyles and tibial plateau were carefully dissected from 5-day-old WT and Malat1^-/- mice. Soft tissues were removed under a microscope. After washing with PBS for two times, the femoral condyles and tibial plateau were first digested in digestion buffer (αMEM containing 10% FBS, 1 mg/ml collagenase II, and 2 mg/ml Dispase I) in 37 °C incubator for 45 min. Then, the femoral condyles and tibial plateau were transferred to fresh digestion buffer and digested overnight in 37 °C incubator. The next day, the isolated chondrocytes were filtered through a 70 μm cell strainer and spun down at 1600 rpm for 5 min. The chondrocytes were directly used for RNA extraction, seeded for Alican blue staining (A5268, MilliporeSigma) in 24-well plates, or for immunofluorescence staining of Aggrecan (A8536, ABclonal) in 96-well plates.

MC3T3-E1 osteoblasts were purchased from ATCC and cultured in αMEM with 10% FBS and 1% Penicillin—Streptomycin (15140122, Thermo Fisher Scientific).

Production of Wnt3a conditioned medium

The L Wnt-3A (CRL-2647, ATCC) cells and the control L cell line (CRL-2648, ATCC) were generously gifted by Dr. Joe Zhou (Weill Cornell Medicine). We followed ATCC handling information to maintain and culture the cells. Briefly, the cells were maintained in DMEM medium supplemented with 10% FBS, 1% Penicillin—Streptomycin and 0.4 mg/ml G-418. To produce the Wnt3a conditioned medium (Wnt3a CM) and control medium (L CM), the cultured cells (80–90% confluence) from one 10cm dish were split at a 1:10 ratio and seeded into 10 cm tissue culture dishes. After an initial 4-day culture (approximately to confluency) in DMEM medium containing 10% FBS and 1% Penicillin—Streptomycin, the first batch of medium was harvested. Subsequently, 10 ml of fresh medium was added, and the cells were cultured for another 3 days to collect the second batch of conditioned medium. The two batches of conditioned media were combined at a 1:1 ratio, filtered using a 0.22 μm filter, and stored at -80°C. Wnt3a activity was confirmed by nuclear translocation of β-catenin.

Immunofluorescence staining

Primary osteoblasts or chondrocytes were seeded into 96-well plates at a density of 1 × 10⁴ cells/well. Primary osteoblasts were serum starved for 16 h, followed by treatment with 50% Lor 50% Wnt3a CM for 1h. After washing with PBS, the cells were fixed with 4% paraformaldehyde in PBS for 20 min at room temperature, permeabilized with 0.5% Triton X-100 in PBS for 15 min at room temperature, and blocked for 1 h with 1% BSA in PBS (blocking buffer). The cells were then incubated with primary antibodies that were diluted in blocking buffer for overnight at 4 °C. After washing three times with PBS, the cells were incubated with secondary antibodies for 1 h at room temperature. After three times of washing with PBS, the cells were mounted with the ProLong™ Gold Antifade Mountant with DAPI (P36941, ThermoFisher Scientific). A Zeiss microscope was used to take images. Primary antibodies include anti-β-catenin antibody (8480s, Cell Signaling Technology, 1:100) and anti-Aggrecan (A8536, ABclonal,1:100) in this study. The goat anti-Rabbit Alexa Fluor™ 488 (A-11008, ThermoFisher Scientific, diluted at 1:500 with blocking buffer) was used as the secondary antibody for these primary antibodies.

Immunoblot analysis

Total cellular extracts were obtained using lysis buffer containing 150 mM Tris-HCl (pH 6.8), 6% SDS, 30% glycerol, and 0.03% Bromophenol Blue, with 10% 2-Mercaptoethanol added immediately before harvesting cells. Cell lysates were fractionated on 7.5% SDS-PAGE, transferred to Immobilon-P membranes (0.45 μm, Millipore), and incubated with specific antibodies. Western Lightning Plus-ECL (PerkinElmer) was used for detection. β-catenin antibody (9562, 1:1000) and Jag1 antibody (70109, 1:1000) were obtained from Cell Signaling Technology. Nfatc1 antibody (556602, 1:1000) was obtained from BD Biosciences; Blimp1 (sc-47732, 1:1000), c-Fos (sc-52, 1:1000), OPG/Osteoprotegerin (sc-390518, 1:1000) and p38α (sc-535, 1:3000) antibodies were purchased from Santa Cruz Biotechnology.

Cytoplasmic and nuclear extraction

The cultured primary osteoblastic cells were subjected to serum starvation in αMEM with 2% FBS for 16h. Subsequently, the cells were treated with either 50% L- or 50% Wnt3a conditioned medium (CM) for 1 hour. The cells were collected and lysed with Buffer A (10mM Hepes PH7.9, 1.5mM MgCl2, 10mM KCl) supplemented with fresh protease inhibitor cocktail (11836170001, MilliporeSigma). After incubation on ice for 15 min, 0.2 % NP-40 was added. The mixture was vortexed and incubated on ice for 2 min. The cellular lysate was then centrifuged at 10,000 g for 3 minutes at 4°C. The supernatant was used as cytoplasmic fraction. The nuclei pellets were washed two times using 1 ml of Buffer A. After the second wash, the buffer was completely removed. The nuclei pellet was then lysed using Buffer C (20mM Hepes pH7.9, 1.5mM MgCl2, 420mM NaCl, 0.2mM EDTA, 25% Glycerol) with fresh protease inhibitor cocktail for 30 minutes on ice, vortexing every 5 minutes. The nuclear lysate was centrifuged at 12,000 x g at 4°C for 15 minutes, and the supernatant was collected as the nuclear extract. GAPDH (sc-25778) and TBP1(sc-204) were purchased from Santa Cruz Biotechnology and used as the cytoplasmic and nuclear markers, respectively. β-catenin antibody (9562, 1:1000) was obtained from Cell Signaling Technology.

Reverse transcription and real-time PCR

DNA-free RNAs were isolated from cells with the RNeasy MiniKit (74104, Qiagen) with DNase treatment, and total RNA was reverse-transcribed with random hexamers using the RevertAid RT Kit (K1691, Thermo Fisher Scientific) according to the manufacturer’s instructions. Real-time PCR was done in triplicate with the QuantStudio 5 Real-time PCR system (A28138, Applied Biosystems) and Fast SYBR^® Green Master Mix (4385612, Thermo Fisher Scientific) with 500 nM primers. mRNA amounts were normalized relative to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA. The mouse primers for real-time PCR were as follows: Gapdh: 5’-ATCAAGAAGGTGGTGAAGCA-3’ and 5’-AGACAACCTGGTCCTCAGTGT-3’; Tnfrsf11b: 5’-CGGAAACAGAGAAGCCACGCAA-3’ and 5’-CTGTCCACCAAAACACTCAGCC-3’; Tnfsf11: 5’-CAGCATCGCTCTGTTCCTGTA-3’ and 5’-CTGCGTTTTCATGGAGTCTCA-3’; Axin2: 5’-ATGCAAAAGCCACCCAAAGG-3’ and 5’-TGCATTCCGTTTTGGCAAGG-3’; Ccnd1: 5’-GCGTACCCTGACACCAATCTC-3’ and 5’-CTCCTCTTCGCACTTCTGCTC-3’; Lef1: 5’-TGTTTATCCCATCACGGGTGG-3’ and 5’-CATGGAAGTGTCGCCTGACAG-3’; c-myc: 5’-CAGCGACTCTGAAGAAGAGCA-3’ and 5’-TTGTGCTGGTGAGTGGAGAC-3’; Jag1: 5’-TGCCTGCCGAACCCCTGTCATAAT-3’ and 5’-CCGATACCAGTTGTCTCCGTCCAC-3’; Tcf7: AACTGGCCCGCAAGGAAAG and CTCCGGGTAAGTACCGAATGC;Nfatc1: 5’-CCCGTCACATTCTGGTCCAT-3’ and 5’-CAAGTAACCGTGTAGCTCCACAA-3’; Acp5: 5’-ACGGCTACTTGCGGTTTC-3’ and 5’-TCCTTGGGAGGCTGGTC-3’; Ctsk: 5’-AAGATATTGGTGGCTTTGG-3’ and 5’-ATCGCTGCGTCCCTCT-3’; Dc-stamp: 5’-TTTGCCGCTGTGGACTATCTGC-3’ and 5’-AGACGTGGTTTAGGAATGCAGCTC-3’; Blimp1: 5’-TTCTTGTGTGGTATTGTCGGGACTT-3’ and 5’-TTGGGGACACTCTTTGGGTAGAGTT-3’; Atp6v0d2: 5’-GAAGCTGTCAACATTGCAGA-3’ and 5’-TCACCGTGATCCTTGCAGAAT-3’; Malat1: 5’-AGCAGGCATTGTGGAGAGGA-3’ and 5’-ATGTTGCCGACCTCAAGGAA-3’; Col2a1: CGATCACAGAAGACCTCCCG and GCGGTTGGAAAGTGTTTGGG; Sox9: AAGCTCTGGAGGCTGCTGAACGAG and CGGCCTCCGCTTGTCCGTTCT; Acan: GGTCACTGTTACCGCCACTT and CCCCTTCGATAGTCCTGTCA.

RNA immunoprecipitation (RIP) assay

Thirty million MC3T3-E1 cells were collected, centrifuged, and washed with PBS. The cells were then spun down at 800xg for 4 min at room temperature. The cell pellet was resuspended in 1% formaldehyde in PBS (28906, Thermo Fisher Scientific) and crosslinked for 10 min at room temperature on an end-to-end rotator. 1.25M glycine at 1/10 volume of 1% formaldehyde solution was used to quench the cross-linking reaction at room temperature for 5 min. The cells were spun down at 2000 x g for 5 min and the pellet was washed with chilled PBS once, followed by centrifugation at 2000 x g for 5 min. The cell pellets were resuspended in 1.1 ml immunoprecipitation (IP) lysis buffer (50 mM HEPES at pH 7.5, 0.4 M NaCl, 1 mM EDTA, 1 mM DTT, 0.5% Triton X-100, 10% glycerol) containing 1 mM PMSF (78830, MilliporeSigma), protease inhibitor cocktail, and RNase inhibitor (100 U/ml Superase-in, AM2694, Thermo Fisher Scientific). The cell lysate was then sonicated using the Bioruptor® Pico sonication device (Diagenode, NJ, USA) until the liquid became clear (sonication cycle: 30 sec ON, 30 sec OFF). The lysates were centrifuged for 10 min at 14,000 x g at room temperature. 50 μl of the supernatant was used as the input control. The remaining supernatant was precleared using protein A/G agarose (sc-2003, Santa Cruz Biotechnology). The precleared cell lysate was split into two tubes evenly and incubated with 5μg of β-catenin antibody (9562, Cell signaling technology) or normal rabbit IgG (2729S, Cell signaling technology) at 4 °C overnight with rotation. 30 μl of washed protein A/G beads were then added into each sample and incubated at 4 °C for 1 h with rotation. The beads were washed with 900 μl of IP lysis buffer for 3min/each time for 5 times and collected by centrifugation for 3 min at 400 x g at room temperature. After the last wash, 100μl of RIP buffer (50 mM HEPES at pH 7.5, 0.1 M NaCl, 5 mM EDTA, 10 mM DTT, 0.5% Triton X-100, 10% glycerol, 1% SDS) with 1μl RNase inhibitor was added to each sample. 50 μl of RIP buffer was added to the input control sample. All samples were incubated at 70 °C for 1 h to reverse crosslinking. 100 μl of supernatant were then collected by spinning down the beads at 400 x g for 1 min at room temperature. The supernatant was used for RNA extraction using the RNeasy Mini Kit (Qiagen) with DNase I treatment. RNA was eluted using 12ul RNase-free H2O and reversed to complementary DNA using One-step cDNA synthesis kit (Thermo, Revert Aid RT kit, k1691). The qPCR was then performed using Malat1 primers.

Chromatin Isolation by RNA Purification (ChIRP) assay

Chromatin Isolation by RNA Purification (ChIRP) was performed as described previously with slight modifications ⁶⁰. We designed and synthesized an antisense oligonucleotide probe of murine Malat1. The design was based on the high sequence homology to a human MALAT1 probe’s sequence ⁶¹ (Supplementary Fig. 3). The murine Malat1 probe’s sequence is 5’-GTCTTTCCTGCCTTAAAGTTAATTTCG/iSp18//3’-BiotinTEG (INTEGRATED DNA Technologies). We also synthesized a GFP probe (sequence: 5’ TATCACCTTCAAACTTGACTTC/ iSp18-3’-Biotin TEG) as the negative control. 30 million MC3T3-E1 cells were collected, washed in PBS and centrifuged at 800xg for 4 min at room temperature. The cell pellet was resuspended in 4% formaldehyde in PBS and crosslinked for 30 min at room temperature on an end-to-end rotator. 1.25M glycine at 1/10 volume of 4% formaldehyde solution was used to quench the crosslinking reaction at room temperature for 5 min. After washing with chilled PBS once, the cells were resuspended in 1 ml lysis buffer (50 mM Tris-Cl pH 7.0, 10 mM EDTA, 1% SDS) supplement with 1 mM PMSF, protease inhibitors, and RNase inhibitor. The cell lysate was sonicated using the Bioruptor until the lysate became clear (sonication cycle: 30 sec ON, 30 sec OFF). The lysate was centrifuged for 10 min at 14,000 x g at 4 °C. The sonicated lysate was then precleared with magnetic streptavidin beads (65001, Thermo Fisher Scientific) at 37 °C for 30 min with slow rotation. 2% volume of precleared lysate was saved for the input. The remaining lysate was split into two new tubes evenly and incubated with 100pmol of the Malat1 probe or the negative control probe in the hybridization buffer (750 mM NaCl, 1% SDS, 50 mM Tris-Cl pH 7.0, 1 mM EDTA, 15% formamide containing 1 mM PMSF, protease inhibitor cocktail (11836170001, MilliporeSigma), and RNase inhibitor (100 U/ml Superase-in, AM2694, Thermo Fisher Scientific)) at 37 °C for overnight with slow rotation. Magnetic streptavidin beads were then added to the samples and incubated at 37 °C for 30 min with slow rotation. DynaMag-15 magnetic strip was used to separate beads from the liquids. After 5 times of washing using the wash buffer (2x NaCl and Sodium citrate (SSC, 15557044, Thermo Fisher Scientific), 0.5% SDS) with PMSF and proteinase inhibitor cocktail, the beads were isolated. The bound proteins were eluted by boiling the beads in 30ul of 3x blue juice buffer (150 mM Tris-HCl (pH 6.8), 6% SDS, 30% glycerol, and 0.03% Bromophenol Blue, with fresh 10% 2-Mercaptoethanol) and subjected to immunoblot analysis.

Luciferase reporter assay

3.5x10⁴ of primary calvarial osteoblasts were plated in 48-well plates. Next day, 480 ng of the M50 Super 8x TOPFlash plasmid (12456, Addgene) and 20ng of pRL-Tk control plasmid (E2241, Promega) were transfected per well using Lipofectamine 3000 (L3000001, Invitrogen) according to the manufacturer’s instructions. After 24h transfection, the culture medium was replaced with fresh αMEM medium supplemented with 10%FBS. 48h post transfection, the cells were serum starved for 1h and then treated with 20% L- or 20% Wnt3a CM for 16h. Firefly and Renilla luciferase activities were measured using a Dual-Luciferase Reporter Assay system (E1910, Promega) on a Gen5 Microplate Reader (BioTek) according to the manufacturer’s instructions. Firefly luciferase activity was normalized to Renilla luciferase activity.

Single cell RNAseq (scRNAseq) Analysis

Single cell RNAseq datasets for bone (GSM3674239, GSM3674240, GSM3674241, GSM3674242) and bone marrow (GSM3674243, GSM3674244, GSM3674246) were downloaded from GSE128423 ⁴². Seurat package ⁶² was applied for downstream analysis. Briefly, genes expressed in fewer than 3 cells and cells with less than 500 genes were filtered out. Cells with over 10% mitochondrial reads and exceeding 6,000 nFeature_RNA were excluded. After NormalizeData, the top 5000 variable genes were selected based on dispersion method using FindVariableGenes function of Seurat package. Subsequently, data scaling was performed using the ScaleData function. All datasets were integrated based on the identified anchors. The first 30 principal components for both UMAP (Uniform Manifold Approximation and Projection) and the subsequent application of a graph-based clustering approach were used, with resolution at 0.1. The Clustree function was executed to understand how the structure of clusters changes across different resolutions. The FindAllMarkers function was utilized with parameters set to prioritizing positive markers expressed in at least 10% of cells within a cluster and exhibiting a log-fold change threshold of 0.25. The following gene markers were also included for cluster annotation: Acan, Col2a1 and Sox9 for chondrocytes ⁴²; Acta2, Myh11 and Mcam for pericytes ⁴²; Adipoq, Lepr, Cxcl12, Cebpa, Kitl and Lpl for Adipoq-lineage progenitors ⁶³; Prrx1, Col1a1, Ibsp and Bglap for osteoblast lineage ⁶³; Pecam1 and Cdh5 for endothelial cells ⁶³; Cd19 for B cells ⁶³; S100a8 for neutrophils ⁶³; Mpz, Mbp and Plp1 for Schwann cells ⁶⁴; Gypa, Alas2, Snca, Hbb-bs, Hbb-bt, Car1, Car2, Klf1, Gata1 and Gata2 for Erythroid cells ^65–68; Pf4, Itga2b and Fli1 for Megakaryocytes ⁶⁸; Cd68, Lyz2, Ly6c2 and Sell for Monocyte-Macrophage lineage^63,69. We utilized Seurat’s FeaturePlot to visualize gene expression in individual cells. DotPlot was employed to illustrate the percentage of cells in a cluster expressing a specific gene and to visualize the average scaled expression level of each gene. The distribution of normalized gene expression levels across all cells in each cluster was visualized using VlnPlot function in Seurat. R version 4.3.2 and Seurat 5.0.1 were used in the study.

Bone marrow supernatant collection

Bone marrow from 12-week-old WT and Malat1^-/- littermate mice was harvested from the femur and tibia (with both ends cut open) by flash centrifugation. The cell pellets were resuspended in 0.2 ml of chilled PBS containing protease inhibitor cocktail (11836170001, MilliporeSigma) and incubated on ice for 30 min. The suspension was centrifuged at 1,500 rpm for 15 min at 4°C and the supernatant was collected for immunoblot analysis.

ELISA

Mouse serum OPG was measured by using Mouse Osteoprotegerin ELISA Kit (MilliporeSigma) according to the manufacturer’s instruction.

Statistical analysis

Statistical analysis was performed using Graphpad Prism® software. Two-tailed Student’s t test was applied when there were only two groups of samples. In the case of more than two groups of samples, one-way ANOVA will be used with one condition, and two-way ANOVA was used with more than one condition. ANOVA analysis was followed by post hoc Bonferroni’s correction for multiple comparisons. p < 0.05 was taken as statistically significant. Data are presented as the mean ±SD as indicated in the figure legends.

Data availability

All data supporting the findings of this study are available within the paper, its Supplementary Information, and source data file. The sequence dataset from GSE128423 ⁴² was reanalyzed.

Code availability

Customized computational scripts of analyzing OPG data were deposited in Zenodo (https://zenodo.org/records/10421692) and GitHub (https://github.com/RugeC/scRNA-seq_OPG_Zhao-lab).

Supplementary Materials

Supplementary Fig. 1

Supplementary Fig. 2

Supplementary Fig. 3

Supplementary Fig. 4

Supplementary Fig. 5

Supplementary Fig. 6

Acknowledgements

We thank Ziyu Chen for mouse breeding and technical assistance, Matthew B. Greenblatt for sharing microCT scanner. We are grateful to the lab members from Dr. Baohong Zhao’s laboratory for their helpful discussions and assistance, and Peyton Carpen for reading the manuscript and valuable discussion. This work was supported by grants from the National Institutes of Health (AR071463 and AR078212 to BZ) and by support for the Rosensweig Genomics Center at the Hospital for Special Surgery from The Tow Foundation. KVP laboratory is supported by grants from NIGMS/NIH (GM132458) and NIA/NIH (AG065748). The content of this manuscript is solely the responsibilities of the authors and does not necessarily represent the official views of the NIH.

Author Contributions

Y.Q. and J.S. designed and performed experiments, analyzed and curated data, prepared figures, wrote and contributed to manuscript preparation, prepared source data file and store raw data. C.X., C.N., S.N., M.E. and Z.D. assisted with experiments. R.C. performed scRNAseq analysis, assisted with uCT and histomorphometry analysis and contributed to manuscript preparation. K.V.P., M.F.E., and S.N. provided mice, discussed data and contributed to manuscript preparation. W. R. mentored R.C., discussed data of fracture model and contributed to manuscript preparation. B.Z. conceived, oversaw the project and wrote the manuscript. All authors reviewed, provided input on the manuscript and approved submission.