Identification of an intronic enhancer regulating RANKL expression in osteocytic cells

The bony skeleton is continuously renewed throughout adult life by the bone remodeling process, in which old or damaged bone is removed by osteoclasts via largely unknown mechanisms. Osteocytes regulate bone remodeling by producing the osteoclast differentiation factor RANKL (encoded by the TNFSF11 gene). However, the precise mechanisms underlying RANKL expression in osteocytes are still elusive. Here, we explored the epigenomic landscape of osteocytic cells and identified a hitherto-undescribed osteocytic cell-specific intronic enhancer in the TNFSF11 gene locus. Bioinformatics analyses showed that transcription factors involved in cell death and senescence act on this intronic enhancer region. Single-cell transcriptomic data analysis demonstrated that cell death signaling increased RANKL expression in osteocytic cells. Genetic deletion of the intronic enhancer led to a high-bone-mass phenotype with decreased levels of RANKL in osteocytic cells and osteoclastogenesis in the adult stage, while RANKL expression was not affected in osteoblasts or lymphocytes. These data suggest that osteocytes may utilize a specialized regulatory element to facilitate osteoclast formation at the bone surface to be resorbed by linking signals from cellular senescence/death and RANKL expression.

Bone remodeling begins with the removal of old or damaged bone by osteoclasts, which is followed by the formation of new bone by osteoblasts. 1,4Although this is one of the fundamental concepts in bone biology, the molecular signals driving osteoclastogenesis on the surface of bone that need to be renewed are still largely unknown.Due to their matrix-embedded and long-lived nature, osteocytes contribute to the sensing of mechanical loading and the detection of microdamage, thus playing a key role in the response to bone tissue senescence. 1 Osteocyte death is associated with various conditions, such as microdamage, disuse, estrogen loss, glucocorticoid treatment, inflammatory diseases and aging. 1 A previous study has shown that bone tissue that contains dying osteocytes displays an increased RANKL expression level. 13Dying osteocytes have been shown to release adenosine triphosphate (ATP) to stimulate RANKL expression in neighboring bystander osteocytes. 14Since phagocytes cannot reach and engulf apoptotic osteocytes embedded in the bone matrix, apoptotic osteocytes undergo secondary necrosis and release damage-associated molecular patterns (DAMPs), which can promote RANKL-induced osteoclastogenesis. 15The senescenceassociated transcription factor (TF) GATA4 stimulates RANKL expression in osteocytes, leading to bone loss in adult mice. 16,17These studies suggest that dying and senescent osteocytes may produce RANKL to induce osteoclast formation at the sites at which renewal is needed.However, it remains unknown how signals from cell death and senescence are integrated into the genomic region of RANKL and thus stimulate its expression in osteocytes.
In this study, we analyzed the chromatin landscape of various RANKL-producing cell types and identified an intronic RANKL enhancer specifically activated in osteocytic cells, including osteocytes and late-stage osteoblasts.TF binding site analysis suggested that the TFs associated with cell death and senescence activate the intronic enhancer element.Mice lacking the intronic enhancer region exhibited decreased RANKL expression specifically in osteocytic cells, leading to a high-bone-mass phenotype in the adult stage.These data provide a molecular basis for RANKL regulation in osteocytic cells and may have therapeutic implications for various skeletal disorders.

Identification of an intronic enhancer of RANKL in osteocytic cells
9][20][21] To gain insight into the mechanisms underlying RANKL expression in osteocytes, we analyzed publicly available epigenomic datasets of osteocytic cells, chondrocytes, fibroblasts and lymphocytes (GSE188004, GSE175159, GSE187956, GSE188124, GSE128642 and ENCODE project dataset ENCFF021TWR). 22Chromatin immunoprecipitation followed by sequencing (ChIP-seq) along with data on antibody binding to acetylated histone H3 Lys27 (H3K27ac) identified an intronic region in the TNFSF11 gene locus with an enrichment of the active enhancer marker H3K27ac in osteocytic cells but not in other cell types (Fig. 1a).The binding of CCCTC-binding factor (CTCF), which often occurs in active enhancers, 23,24 was observed in the intronic region in osteocytic cells (Fig. 1a).These data suggest that the intronic region functions as an active enhancer of RANKL in osteocytic cells.
TF-binding motif analysis for the intronic enhancer region (GRCh38: 42, 594, 792-42, 595, 392) and TF ChIP-seq data analysis suggested the binding of GATA4, C/EBPβ, C/EBPα and CREB to this intronic enhancer (Fig. 1b).The binding sites of these TFs are located in close proximity to one another (Fig. 1b).STRING network analysis predicted potential protein-protein associations among the TFs (Fig. 1c).Functional enrichment analysis of these TFs revealed enrichment of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway terms associated with cell senescence and death, such as "cellular senescence", "p53 signaling pathway" and "apoptosis" (Fig. 1d).These results suggest that signals from cell death and senescence may stimulate the intronic enhancer in osteocytes.profiles in human osteocytic cells (GSE188004, GSE175159, GSE187956), 22 chondrocytes (GSE188124), 22 fibroblasts (GSE128642), 56 and lymphocytes (ENCFF021TWR) and RNA Pol II ChIA-PET results (ENCFF674MPM) in WTC11 cells within the TNFSF11 gene locus.The blue and red shaded areas indicate the RANKL promoter region and intronic enhancer region, respectively.Genetic variants associated with bone mineral density (P < 5 × 10 -6 ) identified in previous studies [25][26][27][28] that are located in the RANKL promoter-intronic enhancer loop are indicated in red letters.b Schematic depicting the enriched transcription factor-binding motifs and the predicted binding sites within the intronic enhancer region (upper).ChIP-seq profiles of C/EBPβ (ENCFF679XUJ), C/EBPα (ENCFF988QGQ), CREB1 (ENCFF894ALC) and GATA4 (ENCFF213XUO) (lower).c,d Protein-protein interaction network (c) and functional enrichment analysis (d) of the predicted TFs using the STRING database 58 (https://string-db.org) Dying osteocytic cells highly express RANKL To test whether cell death signaling stimulates RANKL expression in osteocytes, we analyzed single-cell RNA sequencing (scRNA-seq) data for normal and dying osteocytic cells (GSE154719). 29This scRNA-seq dataset consists of tdTomato-positive cells isolated from the long bones of Dmp1-Cre; Sp7 +/+ ;tdTomato (control) and Dmp1-Cre;Sp7 flox/flox ;tdTomato (Sp7 OcyKO ) mice. 29Osteocyte death was reported to be elevated in the Sp7 OcyKO mice based on terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining and immunostaining for activated caspase-3. 29s Dmp1-Cre is reported to be expressed not only in osteocytes but also in other cells, such as mature osteoblasts, 5 these scRNAseq data contained heterogeneous cell populations.Unsupervised clustering analysis identified an osteocytic cell cluster characterized by high expression levels of the osteocyte marker genes Sost, Ackr3, Fbln7, Dmp1, Irx5 and Dkk1 in both control (normal osteocytic cells, Cluster 8) and Sp7 OcyKO mice (dying osteocytic cells, Cluster 9) (Fig. 2a, b).We performed data integration and confirmed that Sp7 expression was depleted in the osteocytic cells (Cluster 9 in the integrated UMAP) derived from Sp7 OcyKO data (Fig. 2c and Fig. S1A, B), and the Gene Ontology (GO) terms associated with apoptosis were enriched in the Sp7-deleted osteocytic cells (Fig. 2d).Motif gene-set enrichment analysis showed that dying osteocytic cells highly expressed genes regulated by factors associated with CEBP, the well-known transcriptional regulator of cellular senescence and apoptosis [30][31][32] (Fig. 2e).We found that dying osteocytic cells displayed increased expression levels of Tnfsf11 as well as the cell senescence/death-related TFs Cebpb, Cebpa and Trp53 while maintaining the expression of osteocyte markers such as Sost, Ackr3, Fbln7, Dmp1, Irx5 and Dkk1 (Fig. 2f).There was no difference between normal and dying osteocytic cells in the expression level of osteoprotegerin (OPG, encoded by Tnfrsf11b) (Fig. S1C, D).As reported in previous studies, 33,34 RANKL expression was also detected in bone marrow adipo-progenitor cells (Cluster 4, characterized by high expression levels of Lpl, Kcnk2, Cxcl12 and Igfbp5) (Fig. S1A), but Sp7 deficiency did not affect RANKL expression in adipo-progenitor cells (Fig. S1C, D).
These data suggest that the cell death signal stimulates RANKL expression in osteocytic cells, probably via regulation of the intronic enhancer.
Generation of intronic enhancer-knockout mice Intronic elements often undergo negative selection pressure during the course of evolution, so intron genomic sequences are much less conserved than exons. 23,35However, the RANKL intronic enhancer sequence is highly conserved (nearly 90%) among humans, mice, cows and dogs, suggesting that it plays an important role in mammalian homeostasis (Fig. 3a).The murine homologous region of the intronic enhancer also displays enrichment of the H3K27ac binding signal in osteocytic cells (GSE54784), 36 suggesting a conserved function as an active enhancer of RANKL in osteocytic cells (Fig. 3b).Indeed, this murine homologous region has been shown to display reproducible gene enhancer activity in an in vivo enhancer assay (VISTA Enhancer Browser: 37 http://enhancer.lbl.gov).
To examine the physiological relevance of this intronic element in vivo, we generated mice lacking the homologous region (GRCm39 chr14: 78, 525, 012-78, 526, 007; 996 bp deletion) of the intronic enhancer using CRISPR-Cas9-mediated genome editing technology.The successful generation of intronic enhancer-KO mice (hereafter called "IN-KO" mice) was confirmed by Sanger DNA sequencing (Fig. 3b).The IN-KO mice were born at the expected Mendelian frequency and displayed normal tooth eruption and a body size similar to that of littermate wild-type controls (Fig. 3c).
Strikingly, IN-KO mice displayed a severe high-bone-mass phenotype at the age of 12 weeks (Fig. 3d).Microcomputed tomography (μCT) analyses showed that IN-KO mice had an increased bone volume per tissue volume and trabecular number along with decreased trabecular separation and trabecular spacing (Fig. 3e).In contrast, there was no difference between the wild-type and IN-KO mice in bone mass at the age of 3 weeks (Fig. 3f, g).This adult stage-restricted high bone mass phenotype of the IN-KO mice is similar to that observed in Tnfsf11-floxed mice crossed with Dmp1-Cre and Sost-Cre mice. 2,3,5tronic enhancer-deficient mice exhibited decreased osteoclastogenesis and RANKL expression in osteocytic cells To understand the cellular basis underlying the high-bone-mass phenotype of IN-KO mice, we performed histological analysis using toluidine blue and tartrate-resistant acid phosphatase (TRAP) staining on bone sections of the wild-type and IN-KO mice.Consistent with the μCT data, toluidine blue staining indicated that the 12-week-old IN-KO mice had a high bone mass phenotype along with an increased trabecular number and thickness (Fig. 4a).Dynamic bone histomorphometry showed that the osteoclast surface per bone surface was significantly decreased in IN-KO mice (Fig. 4b), indicating impaired osteoclastogenesis in these animals.There was no difference between the wild-type and IN-KO mice in osteoblast surface per bone surface (Fig. 4b).At 3 weeks of age, both the osteoclastic and osteoblastic parameters were normal in IN-KO mice (Fig. 4c, d).
To test whether RANKL expression was decreased in osteocytes in IN-KO mice, we collected osteocyte-enriched bone fractions by using a previously established method. 5The mRNA and protein expression levels of RANKL were significantly decreased in the osteocyte-enriched bone fractions of IN-KO mice (Fig. 5a, b).However, there was no significant difference between the wildtype and IN-KO mice in terms of the RANKL mRNA and protein expression levels in primary osteoblasts or lymphocytes (Fig. 5c-f).These data indicate that the intronic enhancer regulates RANKL expression specifically in osteocytic cells.
Collectively, these data demonstrate that the intronic enhancer controls RANKL expression in osteocytic cells in vivo, thus critically contributing to physiological bone remodeling in the adult stage.

DISCUSSION
In this study, we identified an intronic enhancer specifically involved in RANKL regulation in osteocytic cells.This intronic enhancer was enriched with binding motifs for GATA, CEBP and CREB TFs, and ChIP-seq data analysis indicated that C/EBPβ, C/ EBPα, CREB and GATA4 directly bind to this element.1][32] CREB is a downstream factor of ATP-induced calcium signaling, 38 which is suggested to be activated in osteocytes surrounded by dying cells.,16,40,41 As osteocytes are the longest-living cell type in bone tissue, cell death/senescence in osteocytes may reflect places of old or damaged bone that need to be replaced by osteoclasts.We hypothesize that the intronic enhancer may function as a network hub integrating signals from cell death and senescence to induce RANKL expression in osteocytes, telling osteoclasts where to resorb.Since the high-bone-mass phenotype and the reduction in osteoclast number in the IN-KO mice were milder than those in previously reported osteocyte-specific RANKL KO mice, 2,3,5 it will be important to examine the possibility that other enhancers also contribute to RANKL expression in osteocytes.Further studies are needed to provide a complete picture of the molecular mechanisms underlying RANKL expression in osteocytes.
Consistent with previous reports, 33,34 RANKL expression was also detected in bone marrow adipo-progenitor cells (Fig. S1A, C, D).The importance of RANKL in adipo-progenitor cells has been proposed by studies using Tnfsf11-floxed Adipoq-Cre mice. 33,34owever, the Adipoq-Cre system has been shown to target almost all CXCL12-abundant reticular (CAR) cells. 42Since CAR cells can give rise to osteoblasts and osteocytes, [43][44][45] it is technically difficult to exclude the possibility that the phenotype of Tnfsf11-floxed Adipoq-Cre mice is influenced by the deletion of RANKL in osteoblasts and osteocytes.Further studies are required to clarify the relative contributions of osteocytes and adipo-progenitor cells as sources of RANKL in bone metabolism and to elucidate the role of the intronic RANKL enhancer in other cell types, including adipoprogenitors.
The importance of osteocyte RANKL in bone remodeling has been demonstrated by studies using Tnfsf11-floxed mice crossed with Dmp1-Cre and Sost-Cre mice. 2,3,5,7,8,11,16Although Dmp1-Cre can delete targeted gene expression in cell types other than osteocytes under certain conditions, the importance of osteocyte RANKL is supported by the findings of a study using Sost-Cre mice, in which Cre recombination occurred only in osteocytes, not in osteoblasts or lining cells. 5A previous study using the Dmp1-Cre; DTR system showed that diphtheria toxin administration markedly increased RANKL expression in bone tissue. 137][48] Given that osteocytes are the major sources of RANKL and the most abundant cells in bone, it has long been enigmatic why osteocyte ablation results in RANKL augmentation and osteoclast activation.Our data suggest that cell death signaling stimulates the intronic enhancer to promote RANKL expression in osteocytes, providing a key missing link that may help explain the conflicting observations in previous reports.Since osteocyte death and RANKL derived from osteocytes have been shown to be involved in bone loss under various pathological conditions, such as osteoporosis, arthritis and osteonecrosis, it will be necessary to examine the pathological significance of the intronic enhancer in future studies.RANKL is a multifunctional cytokine that plays a central role in the bone and immune systems. 6,20RANKL activity is maintained under tight local control, and the main source of RANKL varies among different biological processes. 620][21]49 In contrast to other cell types, osteocytic cells utilize regulatory elements located in the intronic region.Intriguingly, the importance of intronic enhancers has also been reported in other genes essential for vertebrate homeostasis, including Foxp3 in regulatory T cells, 50 immunoglobulin heavy chain genes in B cells 51 and musashi1 in neural stem/progenitor cells. 52We suspect that the regulatory element inside the intronic region may exert an advantage in precise control of gene expression by limiting the accessibility of TFs to the open DNA regions and thus avoiding their leaky activity.It is of considerable interest to further investigate why it is necessary for osteocytes to utilize the intronic element to express RANKL and thus ensure skeletal homeostasis.
In summary, bone remodeling is controlled by an osteocytic cell-specific enhancer located in the intronic region of the TNFSF11 gene (encoding RANKL).This study highlights the cell typedependent and context-dependent machinery controlling RANKL expression, providing key insights into the regulatory mechanisms underlying vertebrate homeostasis.

Mice
Mice were housed under specific pathogen-free conditions, and all experiments were performed with the approval of the Institutional Review Board at The University of Tokyo.C57BL/6 J (B6) mice were purchased from CLEA Japan.Intronic enhancer-knockout (IN-KO) mice were generated by CRISPR-Cas9-mediated genome editing technology on the C57BL/6 J background.Single-guide (sg) RNA targeting the sequences of the intronic enhancer region (5′-CCTGGACAGATCTTAGAGTG-3′ and 5′-AAGTTGGGCCCACTTATGGG-3′) and hCas9 mRNA were prepared as described in a previous study. 21The primers for the detection of the IN-KO allele were as follows: forward, 5′-TTGGTTGGAAGTGTTGAAATCCC-3′ and reverse, 5′-CAACCAAGTTTTCCCTTGTAGTATCC-3′.The PCR product of the IN-KO allele (175 base pairs) was further sequenced using the forward primer to confirm that the founder progenies harbored intronic region deletion (996 base-pair deletion corresponding to GRCm39 chr14: 78 525 012-78 526 007).Founder progenies were mated with C57BL/6 J WT mice to generate heterozygous offspring and subsequently intercrossed to generate the wild-type and IN-KO littermate progeny.Age-and sex-matched littermates were used for all experiments unless otherwise noted.

Analysis of the bone phenotype
For microcomputed tomography analysis, femurs were isolated from wild-type and IN-KO mice and fixed with 70% ethanol.
Computational analyses for scRNA-seq data The scRNA-seq dataset (GSE154719) 29 was downloaded from the Gene Expression Omnibus (GEO) database.The samples were prepared from digested bone matrix of Dmp1-Cre;Sp7  29 The viable tdTomato + live cells (collected from the final three collagenase fractions) were sorted and subjected to scRNA-seq.We retrieved the data and performed downstream analysis using the Seurat R package (v4.2.1), including primary analysis, quality control, normalization and scaling, clustering, gene marker identification and visualization of gene expression.Primary filtering removed genes expressed in fewer than 3 cells and cells expressing fewer than 200 genes.Cells with more than 10% mitochondrial reads and 6 000 nFeature_RNA were also removed.After filtering, per-cell counts were normalized (scale factor 10 000 by default), the 2 000 most variable genes using the VST method were identified, and the expression levels of these genes were scaled prior to principal component analysis.The first twenty principal components were used for UMAP projection, and the cells were clustered at a resolution of 1 through a standard graph-based clustering approach.Minor groups of contaminated cells, such as hematopoietic cells (Ptprc + ), perivascular cells (Ptprc -Pecam1 + ) and red blood cells (Slc4a1 + ), were removed, and 21 671 cells from Dmp1-Cre;Sp7 +/+ samples and 5 098 cells from Dmp1-Cre;Sp7 flox/flox samples were used for further downstream analysis.An osteocytic cell cluster was defined by the expression of wellestablished osteocyte marker genes such as Sost, Ackr3, Fbln7, Dmp1, Irx5 and Dkk1.We integrated the two datasets using the scRNA-seq integration method described in Seurat 60 (https:// satijalab.org/seurat/articles/integration_introduction.html),resulting in a total of 26 769 cells available for downstream analysis.We used the FindMarkers function to identify genes that were differentially expressed in normal osteocytic cells (derived from Dmp1-Cre;Sp7 +/+ samples) and dying osteocytic cells (derived from Dmp1-Cre;Sp7 flox/flox samples).Metascape (https:// metascape.org) 61was used to perform the enrichment analyses, including the GO biological process, canonical pathway, Reactome gene set, KEGG pathway, WikiPathways, PANTHER pathway and transcription factor target (miscellaneous) analyses.Violin plots were used to visualize the expression of selected genes in normal and dying osteocytic cells.

Statistics
The data were analyzed using GraphPad Prism software v9.

Fig. 2
Fig. 2 Dying osteocytic cells highly express RANKL.a,b Uniform manifold approximation and projection (UMAP) plots of the scRNA-seq (GSE154719) 29 on tdTomato + cells derived from bone tissues of Dmp1-Cre;Sp7 +/+ ;tdTomato + mice (n = 21 671 cells) (a) and Dmp1-Cre;Sp7 flox/ flox ;tdTomato + mice (n = 5 098 cells) (b).The dot plots in the right panels show the expression of selected osteocyte marker genes in the identified subclusters.c UMAP plot of the integrated data (n = 26 769 cells) and feature plots showing the expression of osteocyte marker genes at the single-cell level.The arrowheads indicate Osteocytic Cell Cluster 9. d,e Functional enrichment analysis (d) and regulatory target gene set analysis (e) comparing normal osteocytes and dying osteocytic cells.f Violin plots showing the expression of selected genes in dying and normal osteocytic cells

Fig. 3
Fig.3Deletion of the intronic enhancer leads to an age-dependent high-bone-mass phenotype.a Sequence similarity analysis of the intronic enhancer among human, mouse, cow and dog genomes using the ECR Browser 59 (https://ecrbrowser.dcode.org)with 90% ECR similarity across a 300-base-pair ECR length.b H3K27ac ChIP-seq profile in murine osteocytic cells (GSE54784).36The nucleic acid sequence of the intronic enhancer in mice is shown in the lower panel.The sequence region colored pink denotes the deletion region obtained with the CRISPR/Cas9 method.c Representative macroscopic images of more than three wild-type (WT) and intronic enhancer-knockout (IN-KO) mice (female, 12 weeks).d, e Representative micro-CT images (d) and micro-CT parameters (e) of the femur in WT and IN-KO mice at the age of 12 weeks (n = 6 male WT and n = 6 male IN-KO mice; n = 7 female WT and n = 4 female IN-KO mice).f, g Representative micro-CT images (f) and micro-CT parameters (g) of the femur in WT and IN-KO mice at the age of 3 weeks (n = 5 female WT and n = 9 female IN-KO mice).Micro-CT scale bars: 1 mm.The data are expressed as the mean ± SEM.P values were determined by two-way ANOVA followed by Tukey's post hoc test (e) and two-tailed t test (g)

Fig. 4
Fig. 4 Intronic enhancer deletion inhibits osteoclastogenesis.a Toluidine blue and TRAP staining of the proximal tibias of WT and IN-KO mice at the age of 12 weeks.The data are representative of more than 3 independent experiments.b Histological analysis of the proximal tibias of WT and IN-KO mice at the age of 12 weeks (WT, n = 15; IN-KO, n = 14).c Toluidine blue and TRAP staining of the proximal tibias of WT and IN-KO mice at the age of 3 weeks.The data are representative of more than 3 independent experiments.d Histological analysis of the proximal tibias of WT and IN-KO mice at the age of 3 weeks (WT, n = 5; IN-KO, n = 9).Histology scale bars: 100 μm.The data are expressed as the mean ± SEM.P values were determined by two-tailed t test (b and d)

Fig. 5
Fig. 5 Intronic enhancer deletion decreases RANKL expression in osteocytic cells but not in other cell types.a Tnfsf11 mRNA expression levels in osteocyte-enriched femur (n = 5 per group) and calvaria (n = 3 per group) bone fractions.b RANKL concentration measured by ELISA in the lysate of osteocyte-enriched femurs and tibias (n = 3 per group).c, d Tnfsf11 mRNA expression levels in primary osteoblasts (n = 3 and n = 4 in the untreated and vitamin D 3 -treated groups) (c) and lymphocytes (n = 3 per group) (d).e Flow cytometric profiles for RANKL expression in CD4 T cells (e, upper panel) and CD8 T cells (e, lower panel).The T cells were stimulated with plate-bound anti-CD3ε (10 μg•mL −1 ) for 20 h.The shadow indicates nonstimulated cells.The graphs show the frequency of RANKL-positive T cells.f RANKL concentration measured by ELISA in the lysate of calvarial primary osteoblasts stimulated with PGE 2 , 1,25(OH) 2 D 3 and TNF-α (n = 3 per group).The data are expressed as the mean ± SEM.P values were determined by two-tailed t test 4.1 and R software v4.2.0.The statistical tests, n values, replicate experiments and P values are all indicated in the figures and/or legends.P values were calculated using Student's t test and ANOVA with Tukey's post hoc test.