https://orcid.org/0000-0002-2836-1081
Figures
Abstract
Osteogenesis depends on the self-renewal and differentiation of mesenchymal stem cells (MSCs). Emerging research underscores the regulatory functions of RNA methylation on bone homeostasis. Here, we show PCIF1, the N6,2′-O-dimethyladenosine (m6Am) methyltransferase, is essential for maintaining bone mass and promoting osteogenic differentiation of MSCs. Multiple complementary analyses—including GWAS, TWAS, and single-cell transcriptomics—collectively point to PCIF1 as a regulator of human bone mineral traits and early-stage mesenchymal differentiation. Global or MSC-specific Pcif1 deletion elicits osteoporotic pathology in mice, although myeloid cell-specific Pcif1 knockout does not induce femur bone alterations. Mechanistically, Pcif1 knockout decreases m6Am signals of Wnt-related genes (Wnt11, Fzd4, and Fgfr2) and accelerates mRNA degradation. This down-regulates active β-Catenin protein, and thus impairs osteogenic function of MSCs. Additionally, the WNT agonist attenuates the osteoporosis-like phenotype induced by Pcif1 deletion. These findings highlight the crucial role of PCIF1-mediated m6Am modification in regulating osteogenesis and suggest potential therapeutic implications for bone disorders.
Citation: Song W, Su K-J, Pan Z, Zhang Z, Yin Q, Lin W, et al. (2026) The m6Am methyltransferase PCIF1 promotes osteogenic differentiation of mesenchymal stem cells through stabilization of Wnt-related transcripts. PLoS Biol 24(4): e3003739. https://doi.org/10.1371/journal.pbio.3003739
Academic Editor: Jeff Coller, Johns Hopkins University, UNITED STATES OF AMERICA
Received: July 3, 2025; Accepted: March 20, 2026; Published: April 6, 2026
Copyright: © 2026 Song et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data are available in the main text or the supplementary materials. The RNA-seq data have been deposited into the NCBI database under the accession number GSE324917, and the scRNA-seq data of human head bone tissue have been deposited under accession number GSE324998. The custom code for single-cell analysis generated in this study is publicly available and archived at Zenodo (DOI: https://doi.org/10.5281/zenodo.19026412). We also downloaded the m6Am-Exo-Seq data of murine testes of Pcif1-HET and Pcif1-KO mice from GEO database (GSE151229).
Funding: This research was supported by grants from the National Natural Science Foundation of China (Grant No. 82125006 to QY; https://www.nsfc.gov.cn/), West China Hospital of Stomatology (Grant No. RCDWJS2025-2 to QY; https://hxkq.org/), and the National Institutes of Health (NIH) (R01AR069055, U19AG055373, and P20GM109036 to HWD, R01AG061917 to HS; https://www.nih.gov/). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Abbreviations: ALP, alkaline phosphatase; AOPI, acridine orange/propidium iodide; ARS, alizarin red staining; BFR/BS, bone formation rate; BMD, bone mineral density; eBMD, estimated bone mineral density; FBS, fetal bovine serum; GEM, gel bead-in-emulsion; GO, gene ontology; GWAS, genome-wide association studies; KEGG, Kyoto Encyclopedia of Genes and Genomes; LD, linkage disequilibrium; MAR, mineral apposition rate; MSCs, mesenchymal stem cells; NCBI, National Center for Biotechnology Information; OB, osteoblast precursors; PBS, phosphate-buffered saline; PCF11, Premature Cleavage Factor II; PCIF1, Phosphorylated CTD Interacting Factor 1; PCs, principal components; RBCs, red blood cells; RSPO2, R-spondin-2; scRNA-seq, single-cell RNA sequencing; THA, total hip arthroplasty; TRAP, tartrate-resistant acid phosphatase; TSS, transcription start sites; TWAS, transcriptome-wide association studies; UMIs, unique molecular identifiers; α-MEM, alpha minimum essential medium
Introduction
RNA epigenetics has been extensively implicated in the modulation of bone homeostasis, underscoring its pivotal role in maintaining skeletal integrity and orchestrating bone remodeling [1–5]. N6,2′-O-dimethyladenosine (m6Am) modification occurs on the first 2′-O-methyladenosine (Am) nucleotide at the 5′-end of capped mRNAs [6,7]. m6Am dynamics are synergistically regulated by the “writer” Phosphorylated CTD Interacting Factor 1 (PCIF1), the “eraser” Fat Mass and Obesity-associated Protein (FTO), and the “reader” Premature Cleavage Factor II (PCF11) [8–12]. m6Am modification has been closely associated with tumor cell aggressiveness, macrophage activation, and CD8+ T cell ferroptosis [13–17]. Currently, there is limited research on the effects of m6Am modification on bone health. Our team has first reported that Pcif1 deletion could regulate murine periodontal inflammation and bone loss by inhibiting macrophage function [16]. Mesenchymal stem cell (MSC)-specific knockout of Methyltransferase like 3 (Mettl3), methyltransferase responsible for N6-methyladenosine (m6A), results in skeletal disorders and aberrant adipogenic differentiation of MSCs [18]. Furthermore, FTO, a common demethyltransferase for m6Am and m6A, is involved in regulating bone mass and osteoblast functions [19]. Given the structural similarity and functional consistency between m6A and m6Am, we propose that m6Am modulation has similar effects on bone development.
Bone homeostasis is primarily a dynamic balance between osteoblast-mediated bone formation and osteoclast-mediated bone resorption [20–22]. MSCs have the capacity to differentiate into osteoblasts, chondrocytes, and adipocytes in response to changes in the microenvironment and cell–cell interactions [23,24]. During osteogenesis, MSCs convert to osteoblastic progenitors, migrate to the bone surface, and participate in osteoblast maturation as well as the formation of mineralized bone tissues [25]. This process is mediated by coordinated regulation of multiple signaling pathways and transcriptional factors [26]. MSC dysfunction leads to bone developmental diseases such as osseous dysplasia, osteoporosis, and osteogenesis imperfecta [27,28]. Dissecting the regulatory molecular mechanisms of MSC biological traits could provide valuable ideas for understanding the onset and progression of skeletal diseases.
Here, we demonstrate PCIF1 is required for maintenance of bone mass and osteogenic functions of MSCs by regulating WNT signaling. Multi-omics approaches suggest PCIF1 is associated with osteoporosis pathogenesis and might control the lineage commitment of MSCs. Whole-body or MSC-specific Pcif1 deletion decreases trabecular bone mass and impairs osteogenic differentiation of MSCs. Mechanistically, MSC osteogenic potential is regulated through the m6Am-Wnt signaling, featured by the accelerated degradation rates of Wnt-related mRNA transcripts and decreased active β-Catenin level by Pcif1 deficiency. Thus, these results highlight the significance of m6Am on bone formation, with potential strategies for bone repair and remodeling.
Results
Genetic and transcriptomic evidence links PCIF1 to osteoporosis risk and early MSC differentiation
To investigate the impact of PCIF1 gene on skeletal homeostasis, we first conducted a gene-based association study involving 11 published genome-wide association studies (GWAS) focused on BMD across various skeletal sites, alongside traits such as fracture susceptibility, osteoclast activity, and estrogen levels (S1 Table). A total of 31 unique SNPs showed associations with PCIF1 across the analyzed traits (S1 File). Four significant associations for PCIF1 were identified in gene-based analyses of bone-related traits (Fig 1A and S1 File). The strongest associations were detected for estimated bone mineral density (eBMD) (p = 2.0 × 10⁻⁸, rs8119032), skull bone mineral density (BMD) (p = 0.034, rs183625591), and osteoclast-related eQTL traits (p = 0.023, rs8114050), suggesting a consistent role of PCIF1 in bone regulation.
(A) GWAS analysis for PCIF1 gene. a: P-value for gene-based association test; b: Female samples only; %: the number of SNPs. Skull BMD: Head skull BMD; Osteoclast eQTL. TB-BMD: Total body BMD. (B) TWAS analysis for PCIF1 gene identified by (Gene-based Integrative Fine-mapping through conditional TWAS; GIFT) method. (C) Flowchart of the scRNA-seq pipeline. This schematic diagram was Created in BioRender. Song, W. (2025) https://BioRender.com/w178r3g and https://BioRender.com/vhpjizr. (D) UMAP projection of all cell types identified from femoral head scRNA-seq data. Major immune, vascular, and skeletal cell populations were annotated, including neutrophils, macrophages, NK/T cells, endothelial cells, red blood cells (RBC), vascular smooth muscle cells (VSMCs), and stromal osteolineage populations: mesenchymal stem cells (MSC), fibroblast-like MSCs (MSC_Fibro), osteoblast precursors (OB), and mature osteoblasts (Mature OB). (E) PCIF1 expression mapped onto the UMAP projection. (F) Violin plot showing PCIF1 expression among human femoral head-derived bone-resident cell types. (G) CytoTRACE analysis of osteolineage cells. Differentiation potential predicted by CytoTRACE demonstrates a gradual loss of stemness from MSC and MSC_Fibro populations toward mature osteoblasts. (H) Monocle 2 pseudotime trajectory of osteoblast differentiation. Cells are arranged along a pseudotime axis showing a continuous transition from MSC_Fibro through OB to Mature OB.
We examined genetic associations of key m6Am regulators (S1 File). The m6Am eraser FTO was strongly associated with fracture risk and BMD, consistent with previous reports [19,29]. The m6Am reader PCF11 also showed associations with estradiol levels and BMD. Given PCIF1’s central role as the m6Am writer and its impact on murine bone physiology, we therefore focused this study on elucidating PCIF1’s function in bone homeostasis.
In the transcriptome-wide association studies (TWAS), PCIF1 was found to have 139 associations with various diseases (S3 File). Notably, it was identified as associated with osteopenia using the GIFT method (p = 2.77 × 10⁻⁷) and showed significant associations with M65-M68 disorders of the synovium and tendon (p = 6.11 × 10⁻⁸) and spinal injury (p = 1.74 × 10⁻⁷). Additional associations included joint disorders (p = 5.02 × 10⁻⁸), M25 joint disorders (p = 9.69 × 10⁻⁹), and gouty arthropathy (p = 5.33 × 10⁻⁷), suggesting a potential role for PCIF1 in osteoporosis, joint diseases, and musculoskeletal health (Fig 1B).
To assess the cellular context and transcriptional pattern of PCIF1 in human bone, we performed single-cell RNA sequencing (scRNA-seq) on femoral head-derived cells collected from patients undergoing total hip arthroplasty for osteoarthritis (Fig 1C). Following stringent quality control, 40,821 high-quality cells were retained for downstream analysis, with a median of 1,030 genes and 2,228 unique molecular identifiers (UMIs) detected per cell. Unsupervised graph-based clustering partitioned the transcriptomes into 11 distinct clusters (Fig 1D), encompassing both hematopoietic and non-hematopoietic populations. These included B cells, T/NK cells, neutrophils, macrophages, red blood cells (RBCs), vascular smooth muscle cells, endothelial cells, and mesenchymal lineage cells such as MSCs, fibroblast-like MSCs (MSC_Fibro), osteoblast precursors (OB), and mature osteoblasts (Mature OB). The key cell lineage marker genes are presented in S1 Fig.
Feature plot analysis revealed that PCIF1 expression was predominantly localized to the MSC cluster, with substantially reduced expression in downstream osteolineage cells (Fig 1E and 1F). This spatial restriction highlights PCIF1 as a candidate regulator of mesenchymal stemness and suggests a potential role in maintaining progenitor identity and transcriptional readiness for osteogenic differentiation.
CytoTRACE analysis further identified MSCs as the most primitive population within the mesenchymal lineage, exhibiting the highest scores across all clusters—indicative of a stem-like, undifferentiated state (Fig 1G). MSC_Fibro cells showed intermediate scores, consistent with their known transitional and multipotent roles [30]. In contrast, OB and Mature OB populations largely displayed transcriptomic signatures of terminal differentiation, though subsets retained residual developmental plasticity.
To reconstruct the osteogenic differentiation continuum, we applied Monocle 2 pseudotime analysis, which revealed a continuous, unbranched trajectory extending from MSCs to mature osteoblasts (Fig 1H). Cells at the earliest pseudotime points overlapped with those exhibiting the highest CytoTRACE scores, reinforcing their identity as early progenitors. MSC_Fibro and OB cells occupied intermediate states, while Mature OBs were enriched at the terminus of the trajectory. Notably, PCIF1 expression was highest at early pseudotime stages and progressively declined along the osteogenic trajectory (Fig 1H), further supporting its functional involvement in early-stage mesenchymal differentiation.
Global deletion of Pcif1 decreases bone mass
To discern the role of Pcif1 in osteogenesis, we first assessed the bone characteristics using whole-body Pcif1 knockout mice (Pcif1−/−) and control littermates (Pcif1+/+). Micro-computed tomography (micro-CT) analysis revealed a reduction in the trabecular bone mass of the distal femur of male Pcif1−/− mice, as indicated by decreased BMD and bone volume (BV/TV) (Fig 2A and 2B). Extended research showed Pcif1 deletion reduced trabecular number (Tb.N) and increased trabecular separation (Tb.Sp), while midshaft cortical thickness (Ct.Th) was unaffected (Fig 2B). Von Kossa staining also confirmed the impaired mineralization of distal femoral metaphysis (Fig 2C). Histomorphometric analyses suggested that Pcif1−/− mice exhibited a significant reduction in osteoblast numbers (N.Ob/B.Pm) and a marginal decrease in osteoclast numbers (N.Oc/B.Pm) (Fig 2D and 2E). Loss of Pcif1 contributed to the decreased mineral apposition rate (MAR) and bone formation rate (BFR/BS) as assessed by calcein double-labeling (Fig 2F and 2G). Female Pcif1−/− mice exhibited a bone loss phenotype similar to that observed in male mice (S2 Fig).
(A) Representative Micro-CT images of femurs from 6-week-old male Pcif1−/− knockout mice and Pcif1+/+ littermates. Scale bar, 200 μm. (B) Quantitative analyses of distal trabecular bone and midshaft cortical thickness of femurs (n = 8). (C) Representative images of Von Kossa staining of undecalcified femoral sections. Scale bar, 500 μm. (D) Representative TRAP images of trabecular bone from the femoral metaphysis. Scale bar, 50 μm. (E) Quantitation of osteoblast and osteoclast numbers of trabecular bone (n = 8). (F) Representative images demonstrating calcein double-labeling patterns in trabecular bone from the femoral metaphysis. Scale bar, 20 μm. (G) Quantitation of mineralization apposition rate (MAR) and bone formation rate (BFR/BS) of trabecular bone (n = 8). (H) Representative images and quantitative analyses of ALP and ARS staining of MSCs isolated from Pcif1−/− knockout mice and Pcif1+/+ controls (n = 3–4). (I) qRT-PCR analysis of osteogenic markers in MSCs cultured with osteogenic medium for 7 days in vitro (n = 3). The data underlying panels B, E, G, H, and I can be found in S1 Data (Sheet Fig 2).
Next, we sought to elucidate the role of Pcif1 in osteogenic differentiation of MSCs in vitro. Pcif1 deficiency markedly compromised the intensity of alkaline phosphatase (ALP) and alizarin red staining (ARS) (Fig 2H). The expression of osteogenic markers, including Runx2, Sp7, Col1a1, Alp, and Bglap, was also reduced (Fig 2I).
Conditional deletion of Pcif1 in MSCs leads to osteopenia
To determine whether the low bone mass of Pcif1−/− mice specifically results from osteogenic defects, we generated the MSC-specific Pcif1 knockout mice model (Prrx1-Cre;Pcif1fl/fl mice). Prrx1-Cre;Pcif1fl/fl mice were born alive at an expected mendelian ratio. Micro-CT revealed Prrx1-Cre;Pcif1fl/fl mice had significantly reduced bone mass compared to Pcif1fl/fl controls at both 3 and 6 weeks of age (Fig 3A and 3B). Quantitative results showed that ablation of Pcif1 in Prrx1+ MSCs resulted in a reduction in BMD, BV/TV, Tb.N, and trabecular thickness (Tb.Th), along with an increase in Tb.Sp (Fig 3B). Ct.Th remained unchanged (Fig 3B). Pcif1 deletion diminished the mineralized trabecular bone (Fig 3C), as well as MAR and BFR/BS of femoral trabecular bone (Fig 3F and 3G). Although the number of osteoclasts was slightly decreased, we also observed a further weakened state of osteoblasts in Prrx1-Cre;Pcif1fl/fl mice (Fig 3D and 3E).
(A) Representative Micro-CT images of femurs from 3-week-old or 6-week-old male Prrx1-Cre;Pcif1fl/fl mice and Pcif1fl/fl littermates. Scale bar, 200 μm. (B) Quantitative measurements of trabecular bone parameters in the distal femur and cortical bone thickness at the femoral midshaft (n = 8). (C) Representative images demonstrating Von Kossa staining of undecalcified femoral sections. Scale bar, 500 μm. (D) Representative TRAP images of trabecular bone at the distal end of femurs. Scale bar, 50 μm. (E) Quantitation of osteoblast and osteoclast numbers of trabecular bone (n = 8). (F) Representative images showing calcein double-labeling of trabecular bone. Scale bar, 20 μm. (G) Quantitation of MAR and BFR/BS of trabecular bone from the femoral metaphysis (n = 8). (H) Representative ALP and ARS images with quantification of MSCs isolated from Prrx1-Cre;Pcif1fl/fl mice and Pcif1fl/fl controls (n = 3–4). (I) qRT-PCR analysis of osteogenic gene expression in MSCs following 7 days of culture in osteogenic medium in vitro (n = 3). The data underlying panels B, E, G, H, and I can be found in S1 Data (Sheet Fig 3).
Similarly, MSCs from Prrx1-Cre;Pcif1fl/fl mice showed limited osteogenic differentiation compared to Pcif1fl/fl controls, as evidenced by ALP and ARS staining (Fig 3H), as well as qRT-PCR tests (Fig 3I).
LysM-driven Pcif1 knockout does not change the bone mass
Bone homeostasis relies on the equilibrium between osteogenesis and osteoclastogenesis [31–33]. Tartrate-resistant acid phosphatase (TRAP) staining showed reduced bone resorption occurring in femoral trabecular bone of Pcif1−/− mice (Fig 2D and 2E). Although the change in osteoclast numbers was smaller than that in osteoblast numbers (Fig 2E), we investigated whether Pcif1 deletion could directly induce osteoclast impairment in vivo by generating myeloid cell-specific Pcif1 knockout mice (LysM-Cre;Pcif1fl/fl mice) [16]. Femurs from 6-week-old mice exhibited comparable bone parameters between LysM-Cre;Pcif1fl/fl mice and Pcif1fl/fl littermates, regardless of sex, as supported by BMD, BV/TV, Tb.N, Tb.Th, Tb.Sp, and Ct.Th (Figs 4A, 4B, and S3). Histomorphometric measurements did not show notable difference by LysM-driven deletion of Pcif1 (Fig 4C–4E). Although in-vitro induction of BMDMs derived from LysM-Cre;Pcif1fl/fl mice displayed decreased osteoclast numbers and sizes (Fig 4F), the expression of osteoclast markers did not differ between two groups (Fig 4G).
(A) Representative Micro-CT images of femurs from 6-week-old male LysM-Cre;Pcif1fl/fl mice and Pcif1fl/fl littermates. Scale bar, 200 μm. (B) Quantitation of bone parameters (n = 8). (C) Representative Von Kossa images of undecalcified femoral sections. Scale bar, 500 μm. (D) Representative TRAP images of distal trabecular bone. Scale bar, 50 μm. (E) Quantitation of osteoclast and osteoblast numbers of trabecular bone (n = 8). (F) Representative in-vitro TRAP images and osteoclast quantification from BMDMs of LysM-Cre;Pcif1fl/fl mice and Pcif1fl/fl controls (n = 3). Scale bar, 200 μm. (G) qRT-PCR analysis of osteoclast markers in vitro (n = 3). The data underlying panels B, E, F, and G can be found in S1 Data (Sheet Fig 4).
PCIF1-mediated m6Am modification regulates osteogenic differentiation of MSCs through WNT signaling
To uncover the potential mechanism of PCIF1 in regulating osteogenesis, we performed bulk RNA-seq using MSCs from Prrx1-Cre;Pcif1fl/fl mice and Pcif1fl/fl littermates after 7 days of osteoblast induction. A total of 370 down-regulated genes and 162 up-regulated genes were identified (|log2FoldChange| ≥ 1, padj ＜ 0.05) (S4A Fig). Gene ontology (GO) analysis revealed that ossification, osteoblast differentiation, and canonical Wnt signaling pathway were the most significantly down-regulated gene sets in Pcif1-deficient cells (Fig 5A). Kyoto Encyclopedia of Genes and Genomes (KEGG) assessment suggested Pcif1 deficiency may mainly affect Wnt signaling pathway, calcium signaling pathway, and cGMP-PKG signaling pathway (Fig 5B). We found that Pcif1 knockout reduced the expression of osteogenic marker genes, such as Dlx5, Alpl, Sp7, Bmp4, and Bmp5, as well as Wnt-related genes, including Wnt11, Wnt4, Fgfr2, Fzd4, Snai2, Lgr5, Pth1r, and Rspo2 (Fig 5C). Western blot confirmed a decrease in WNT11, FZD4, and FGFR2 protein levels by Pcif1 deficiency (Fig 5D). Additionally, we examined expression levels of Wnt-related transcripts in MSCs under basal conditions. Pcif1 deletion markedly reduced Wnt11 and Fzd4 expression in undifferentiated MSCs, though Fgfr2 remained unchanged (S4B Fig).
(A) GO analysis of differentially expressed genes by RNA-seq of MSCs from Prrx1-Cre;Pcif1fl/fl mice and Pcif1fl/fl controls after 7 days of osteogenic induction. (B) KEGG analysis of significantly down-regulated genes. (C) Heatmap of representative down-regulated genes associated with ossification and Wnt signaling. (D) Western blot showing expression of Wnt-related proteins. (E) IGV images showing down-regulated m6Am peaks of Wnt11, Fzd4, and Fgfr2 mRNAs by Pcif1 deletion. (F–H) RNA decay analyses of Wnt11, Fzd4, and Fgfr2 in MSCs after 7-day osteogenic induction in vitro (n = 3). The data underlying panels F, G, and H can be found in S1 Data (Sheet Fig 5).
By integrating our transcriptome profiles, the m6Am database (GSE151229), and the Wnt signaling pathway gene set (mmu04310) (S4C Fig), we proposed that PCIF1-mediated m6Am modification was prevalent in the canonical and non-canonical WNT signaling pathways. Specifically, we prioritized several genes for a detailed analysis, focusing on those with m6Am peaks present at the transcription start sites (TSS) (Fig 5E). Wnt family member 11 (Wnt11) has been recognized as one of the predominant non-canonical WNT ligands [34,35]. Frizzled class receptor 4 (Fzd4) primarily participates in the canonical WNT signaling pathway, activating the β-Catenin-dependent signaling cascade, thereby initiating the expression of genes associated with bone development [36–38]. Fibroblast growth factor receptor 2 (Fgfr2) and the WNT signaling pathway act synergistically, particularly in skeletal and craniofacial development [39–41]. Loss of Pcif1 decreased m6Am modification levels on the Wnt11, Fzd4, and Fgfr2 mRNAs (Fig 5E). In addition, Pcif1 knockout reduced the expression of active β-Catenin, but no significant difference was observed in β-Catenin levels (Fig 5D).
To further assess the effect of m6Am modulation on the aforementioned mRNA transcripts, we examined their decay rates. Following a 7-day osteoblastic induction period, Pcif1-deficient cells and Pcif1-wildtype cells were treated with actinomycin D. After 2, 4, or 6 hours of treatment, Wnt11 mRNA in Pcif1-deleted cells exhibited greater degradation compared to that in the control cells (Fig 5F). The mRNA levels of Fzd4 and Fgfr2 exhibited similar trends (Fig 5G and 5H).
Functional verification of candidate m6Am-modified genes
To examine whether these Wnt-related genes are regulated by m6Am modifications in MSCs, we predicted the m6Am distribution using the m6AmPred database [42]. Mouse Wnt11 transcript revealed a high likelihood of m6Am modification within the first 200 nt downstream of the TSS, particularly at position 56, 59, 65, 177, and 194 (Fig 6A). We further performed MeRIP-qPCR to assess m6Am enrichment on the Wnt11 mRNA using two primer sets (Primer 1, 0:+125 distance; Primer 2, +125:+250 distance). Primer 1 detected a prominent m6Am peak near the 5′-UTR, and the m6Am level was markedly reduced upon Pcif1 deficiency (Fig 6B). In contrast, Primer 2 showed low enrichment in the corresponding region (Fig 6B). Collectively, these results indicated that m6Am modification is present on the adenosine located downstream of the Wnt11 TSS.
(A) The calculated likelihood ratio of m6Am modification sites on mouse Wnt11 mRNA predicted by m6AmPred database. (B) MeRIP-qPCR analysis of m6Am enrichment at two predicted sites within Wnt11 mRNA in MSCs. (C) The calculated likelihood ratio of m6Am modification sites on mouse Fzd4 mRNA predicted by m6AmPred database. (D) Detection of the m6Am level in Fzd4 mRNA by MeRIP-qPCR. (E) The calculated likelihood ratio of m6Am modification sites on mouse Fgfr2 mRNA predicted by m6AmPred database. (F) MeRIP-qPCR–based quantification of m6Am enrichment at predicted sites in Fgfr2 mRNA. (G–I) Relative luciferase activity of WNT11 (G), FZD4 (H), and FGFR2 (I) in 293T cells co-transfected with PCIF1 siRNA and PCIF1 overexpression plasmid. The data underlying panels B, D, F, G, H, and I can be found in S1 Data (Sheet Fig 6).
As for Fzd4, potential m6Am modifications were modeled within the first 100 nucleotides (position 53, 55, 56, and 77) (Fig 6C). Similarly, by examining this region, we confirmed that loss of Pcif1 in MSCs led to an over 50% reduction of m6Am level on Fzd4 mRNA (Fig 6D). While the +400:+500 sequence of Fgfr2 transcript was predicted to have a much higher probability than the 0:+150 region (Fig 6E), MeRIP-qPCR results revealed a completely opposite conclusion. The 0:+150 region (Primer 1), rather than the +400:+500 region (Primer 2), was identified as the m6Am-modified site (Fig 6F).
To directly assess the role of m6Am in regulating WNT11 mRNA stability, we generated a dual-luciferase WNT11 reporter plasmid, a wild-type PCIF1 overexpression plasmid (PCIF1-WT), and a catalytic-dead PCIF1 mutant plasmid (PCIF1-MUT) [13,15]. Overexpression of wild-type PCIF1 restored the compromised WNT11 expression induced by PCIF1 knockdown (Fig 6G). Catalytically inactive PCIF1 could not increase the WNT11 level in the siPCIF1 group (Fig 6G). Likewise, similar results were also observed for the expression of FZD4 and FGFR2 (Fig 6H and 6I), suggesting that the catalytic activity of PCIF1 is essential for mRNA stability of these genes.
WNT agonist mitigates bone loss induced by Pcif1 deletion
R-spondin-2 (RSPO2) has been widely identified as an activator of canonical WNT signaling [43–46]. We treated MSCs from Prrx1-Cre;Pcif1fl/fl mice and Pcif1fl/fl littermates with recombinant RSPO2 (rhRSPO2) (100ng/mL). After 7 days of osteoblast induction, rhRSPO2 treatment enhanced the ALP activity and mineralization of Pcif1-wildtype cells (Fig 7A). In detail, rhRSPO2 partially restored the impaired osteogenic capacity of Pcif1-deleted cells, although the recovery did not reach the original level observed in Pcif1-wildtype cells (Fig 7A). qRT-PCR results showed the decreased expression of Runx2, Alp, and Bglap by Pcif1 deficiency was elevated by rhRSPO2 (Fig 7B). Besides, the compromised expression of WNT11, FZD4, FGFR2, and active β-Catenin in Pcif1-knockout cells were up-regulated by rhRSPO2 treatment (Fig 7C).
(A) Representative images and measurements of ALP and ARS staining of MSCs (n = 3–4). (B) qRT-PCR analysis of osteogenic markers in MSCs after induction for 7 days (n = 3). (C) Western blot showing expression of Wnt-related proteins. (D) Representative Micro-CT images of distal femur metaphysis and midshaft cortical bone after administration of WNT agonist. Scale bar, 200 μm. (E) Quantitation of bone volume (n = 6). (F) Schematic diagram of the manuscript created using selected icons from BioRender. Song, W. (2025) https://BioRender.com/lihviy1. The data underlying panels A, B, and E can be found in S1 Data (Sheet Fig 7).
To test the effect in vivo, we performed weekly intraperitoneal injection of rhRSPO2 (4 mg/kg) to 3-week-old Prrx1-Cre;Pcif1fl/fl male mice for five weeks. Notably, the bone deformities caused by Pcif1 deficiency were alleviated (Fig 7D). The trabecular parameters of Prrx1-Cre;Pcif1fl/fl mice, including BMD, BV/TV, Tb.N, and Tb.Th, demonstrated notable improvement following rhRSPO2 injection (Fig 7E). As well, the increased Tb.Sp by Pcif1 deletion was effectively ameliorated by rhRSPO2 treatment (Fig 7E).
Discussion
Emerging evidence has elaborated the role of m6Am modification in tumorigenesis, anti-tumor immunity, anti-infection responses, and various biological processes [15–17,47–49]. Since PCIF1 has been identified as the methyltransferase for editing m6Am peaks [50,51], it has become a prominent research topic in the field of epigenetic regulation. Existing studies have mainly focuses on the immunological functions of m6Am modification, with limited attention given to its role in bone elongation. Moreover, m6A and m6Am exert similar functions in bone physiology, probably because of their structural similarity. Our previous work has unveiled that m6A methyltransferase METTL3 modulates MSC differentiation, tooth root formation, calvarial ossification, cartilage development and periodontal inflammation [2,5,18,52,53]. In contrast, current research on m6Am has predominantly centered on its immunological functions. Our group remains the only one to provide evidence that m6Am regulates periodontal bone loss, a biological effect that parallels the role of METTL3-mediated m6A modification in periodontal inflammation-associated bone loss. In this study, we further demonstrate that m6Am-Wnt signaling regulates femoral bone mass, expanding the known scope of m6Am function in skeleton development, and suggesting its potential as a therapeutic target for osteoporosis.
We used a combination of human gene-based association genetic studies from large population-based cohorts and scRNA-seq data to investigate the role of PCIF1 in bone-related traits. GWAS and TWAS studies identified the strong associations between PCIF1 and eBMD or bone disorders. Notably, to the best of our knowledge, previous genome-wide and transcriptome-wide studies have not directly implicated PCIF1 in bone-related traits. Our current gene-based results revealed that PCIF1 harbors multiple SNPs associated with bone phenotypes, with the lead SNP rs8119032 showing genome-wide significance (p = 2.0 × 10⁻⁸). Annotation of the lead PCIF1-associated SNPs (rs8119032, rs18362559, and rs8114050) using Ensembl showed that they lie in intronic or proximal intergenic regions relative to PCIF1, a genomic context where many regulatory elements reside. Although such noncoding variants frequently act through regulatory mechanisms, none of the lead SNPs appeared as significant cis-eQTLs for PCIF1 in GTEx v8 single-tissue datasets. This suggests that any regulatory effects may be tissue-specific, modest in magnitude, or restricted to bone-related or disease-associated cellular contexts that are not represented in GTEx. Thus, while our genetic analyses robustly implicate the PCIF1 locus in bone traits, the precise regulatory elements and mechanisms linking these noncoding variants to PCIF1 activity remain to be determined by future functional studies. In line with our previous report [16], PCIF1 also showed significant correlation with osteoclasts, indicating that it may act as a broader regulator of bone homeostasis. These findings nominate PCIF1 as a previously unrecognized genetic contributor to bone health and provide functional insight into its role.
Our scRNA-seq analysis of femoral head-derived bone-resident cells revealed that PCIF1 is selectively expressed in MSCs and progressively declines along the osteogenic differentiation trajectory, suggesting a role in early osteolineage programming. This expression pattern implies that PCIF1 may contribute to the maintenance of MSC identity or the regulation of early differentiation events. The use of femoral head tissue from total hip arthroplasty remains a practical and ethically appropriate approach for obtaining human bone cells. Since our scRNA-seq data were derived from femoral head tissue of patients with osteoarthritis, we acknowledge that disease-associated inflammatory factors might influence the transcriptional profiles or composition of bone-resident MSCs. Although evidence suggests that osteoarthritis has limited effects on deep trabecular bone [30], the potential impact of disease-related changes on PCIF1 expression and MSC heterogeneity cannot be fully excluded. We note that our findings may not fully represent the transcriptional landscape of healthy bone. Nevertheless, this dataset provides a valuable foundation for future comparative studies aimed at validating PCIF1’s role in normal bone development and homeostasis. The elevated expression of PCIF1 in undifferentiated MSCs suggests that PCIF1-mediated m6Am methylation may help sustain gene expression programs essential for stemness and inhibit premature lineage commitment.
Notably, cells with high PCIF1 expression also exhibited gene expression profiles characteristic of early-stage progenitors, while more differentiated osteoblasts showed minimal PCIF1 activity. This graded expression pattern supports the idea that PCIF1 may act as a molecular checkpoint in osteogenic differentiation, helping to regulate the transition from stemness to commitment. Similar functions for RNA methylation machinery have been described in other stem cell systems, such as the role of METTL3-mediated m6A methylation in promoting differentiation in embryonic stem cells [54]. While PCIF1 catalyzes the related m6Am modification, its expression pattern in MSCs suggests it may similarly modulate stemness and lineage progression in skeletal tissue. Our findings extend the regulatory role of RNA methylation in cell fate decisions to the context of bone development. Based on these data, we hypothesized that PCIF1 was strongly correlated with bone mass, and might regulate bone homeostasis via osteoblast–osteoclast balance.
To verify this hypothesis, we performed extensive studies using several Pcif1-knockout mouse models. We first observed global Pcif1 ablation mice had shorter size and reduced trabecular bone measurements, while cortical bone thickness was unaffected by Pcif1 deficiency. Notably, loss of Pcif1 led to a decrease in osteoblast and osteoclast numbers, with the more pronounced downregulation of osteoblasts compared with osteoclasts. The observed reduction in osteoclasts may reflect altered RANKL/OPG signaling secondary to compromised osteoblast function, as osteoblasts are a major source of RANKL and OPG [55,56]. Hence, we introduced Prrx1-Cre and LysM-Cre for conditional Pcif1 knockout in MSC and myeloid cells, respectively, to further explore whether PCIF1 targets at osteogenesis or osteoclastogenesis. Prrx1-driven Pcif1 knockout induced delayed bone formation in murine femurs, resembling the phenotype observed in systemic knockout mice. However, LysM-driven Pcif1 deletion did not cause changes of trabecular volumetric BMD, though in-vitro culture showed compromised osteoclast differentiation. We propose that the effects of Pcif1 on osteoclasts may be masked or compensated by other factors in vivo, and unable to produce measurable changes. Moreover, osteoclast formation is mainly driven by RANKL in vitro, and could not fully recapitulate the complex microenvironment found in vivo. In this paper, we prioritized MSCs as the principal cell population affected by Pcif1 deficiency during skeleton development.
WNT signaling facilitates MSC renewal and osteoblast differentiation and inhibits osteoblast apoptosis [57–59]. Accumulating reports have confirmed the crucial role of WNT signaling in skeletal development and bone regeneration [60–62]. Our RNA-seq of osteogenic MSCs results showed the most significant downregulation pathways including ossification and Wnt signaling. Wnt-related genes exhibited widespread downregulation, suggesting PCIF1 could broadly affect WNT pathway rather than targeting a single gene or a limited subset of genes. By combing analyses of a published m6Am-exo-seq database, we noticed the TSS regions of Wnt11, Fzd4, and Fgfr2 mRNA had PCIF1-mediated m6Am peaks. WNT11 is a potent ligand of non-canonical WNT signaling [63,64], regulating osteoblast maturation and bone formation through RSPO2-LGR5 pathway [65,66]. Overexpression of FZD4 restores the inhibited bone formation by mechanical unloading [67]. Inducible FGFR2 activation enhances the osteogenesis by activating Wnt/β‐catenin signaling after bone marrow ablation [68]. FGFR2 mutations cause lethally systemic skeletal malformation in human subjects [69]. Our previous research has reported that FGFR2 ligand FGF2 controls BMSC fate by modulating rDNA chromatin architecture through phase separation [70]. Existing literature has illustrated that PCIF1-mediated m6Am regulation controls mRNA stability [8,14,16,47]. Here, we found Pcif1 deficiency accelerated decay rates of Wnt11, Fzd4, and Fgfr2 mRNAs, indicating that PCIF1 manipulates WNT signaling by stabilizing mRNAs. We validated the down-regulated expression of WNT11, FZD4, and FGFR2 proteins, as well as the decreased levels of active β-Catenin.
Given the epitranscriptomic landscape is highly cell-type and context-specific, we integrated m6AmPred predictions with MeRIP-qPCR for further validation. We subjected the Pcif1-deleted MSCs and WT cells to m6Am enrichment, and designed the primers according to the prediction scores by m6AmPred. Mouse Wnt11 and Fzd4 transcripts showed significant m6Am distribution within the first ~150 nt downstream of TSS, which were consistent with previous reports [10,51]. The pronounced loss of these signals in the Pcif1-deleted cells strongly supported that the detected modification corresponds to m6Am. Interestingly, mouse Fgfr2 transcript got higher score in the+400:+500 region, whereas MeRIP-qPCR results identified the m6Am peaks within the 0:+150 region. This might be ascribed to the limitations of the prediction model, the complexity of the sequence context, and cell-type-specific TSS selection. Dual-luciferase reporter assays also showed that exogenous expression of PCIF1 with m6Am catalytic activity was able to restore the expression of WNT11, FZD4, and FGFR2 in PCIF1-knockdown cells. A limitation is that, although m6Am was detected, its precise modification sites remain to be determined. Further investigations are needed to elucidate the broader landscape of m6Am-mediated regulation.
We next evaluated whether the WNT activator RSPO2 could function as potential therapeutic intervention for Pcif1 knockout-induced osteoporosis. Prior research has identified RSPO2 mutations in fetuses exhibiting severe limb disorders, emphasizing its critical role in skeletal development [71]. Conditional Rspo2 ablation in Osteocalcin+ cells could lead to decreased body size and femoral trabecular bone in either male or female mice [72]. Overexpression of Rspo2 promotes BMP-induced osteoblast mineralization in vitro [65]. In vitro and in vivo functional studies revealed administration of rhRSPO2 partially rescued the bone formation disruption and the restrained osteoblast differentiation of MSCs. The partial rescue suggests that PCIF1 may affect other signaling cascades related to bone biology. For example, RNA-seq results indicates that Pcif1 deletion also downregulates the Calcium signaling and cGMP–PKG pathways. Besides, Pcif1 deletion reduces the expression of multiple genes related to osteoblast differentiation and mineralization, e.g., Bmp4, Bmp5, Alpl, and Sp7. Moreover, RSPO2 efficacy is likely influenced by the abundance of downstream receptors and WNT ligands [46]. It is also possible that some developmental defects are not fully reversible when established during the early postnatal period in Pcif1-deficient mice.
It should be noted that osteoporosis in humans is predominantly an age-related condition, whereas our study primarily employed young mouse models (3- and 6-week-old). Future studies using aged or osteoporotic animal models will be essential to determine whether PCIF1-mediated m6Am modification similarly contributes to age-associated bone loss.
In sum, our findings demonstrate PCIF1 as a candidate associated with osteoporosis-related traits. Pcif1 deletion impairs trabecular bone formation by suppressing osteogenic differentiation. Mechanically, PCIF1-mediated m6Am modulation enhances the mRNA stability of Wnt-related genes. Our findings unveil a pivotal role of PCIF1-mediated m6Am modulation in skeletal homeostasis, highlighting its potential for therapeutic intervention.
Methods
Genome-wide association studies (GWAS)
Data were obtained from the GEnetic Factors for Osteoporosis (GEFOS) consortium (http://www.gefos.org/). BMD assessments in studies like GEFOS2, GEFOS Life Course, and the Louisiana Osteoporosis Study (LOS) utilized dual-energy X-ray absorptiometry (DXA). The UK Biobank studies (UKBB), including UKBB2017 and UKBB2018, estimated BMD using quantitative heel ultrasound (eBMD). Fracture data were sourced from GEFOS2 ALLFX and the UKBB studies, identified through hospital-based diagnoses (ICD10 codes) or self-reported incidents via questionnaires. Participants reported fractures occurring within the past 5 years, with exclusions for fractures of the skull, face, hands, and feet, as well as pathological fractures due to malignancy, atypical femoral fractures, periprosthetic fractures, or healed fractures. For detailed information on the GWAS datasets, please refer to S1 Table.
We conducted gene-based association tests using the R package “snpsettest,” which aggregates SNP effects while accounting for linkage disequilibrium (LD) between markers [73]. This approach incorporates LD among SNPs within each gene and employs permutation-based simulation to compute gene-based p-values. We evaluated association signals from all SNPs within a 50kb window before and after the gene for SNP mapping. To infer pairwise LD correlations among SNPs, we utilized 1000 Genomes data [74] (available at https://www.cog-genomics.org/plink/2.0/resources) aligned to human reference genomes such as Hg18, Hg19, or Hg38, selected based on the GWAS studies. Selection criteria considered race and sex to accurately estimate LD. Additionally, we utilized the UCSC Genome Browser’s UCSC Overlap Lift (UCSC’s USCS overlift) tool to convert coordinates from the Hg19 human reference genome to the Hg18 reference genome. A gene with p-values less than 0.05 was considered significant.
To assess the regulatory relevance of variants identified from the gene-based association analyses, we performed systematic variant annotation and eQTL mapping. Lead and nominally significant SNPs (p ≤ 0.05) for PCIF1 were examined. Variant annotation was conducted using the R package biomaRt [75], which queried the Ensembl database [76] to retrieve genomic coordinates, variant consequences, and gene annotations for the top SNPs identified from the gene-based analyses. The annotation was performed according to the genome reference version used in the corresponding GWAS datasets (S1 Table). This process enabled classification of variants as coding or noncoding and identification of potential regulatory regions proximal to each gene.
Significant single-tissue eQTL data were obtained from the GTEx v8 database [77] using the R package gtexr [78] [https://ropensci.r-universe.dev/gtexr]. Gene symbols were first mapped to their corresponding Ensembl identifiers, and all significant variant–gene–tissue associations were retrieved across GTEx tissues. These eQTLs were then intersected with the top SNPs with matched rsID identified from the gene-based analyses to identify overlapping regulatory variants. For each overlap, we recorded the corresponding tissue, the pre-computed eQTL association p-value, and normalized effect size (NES) as provided by the original eQTL analysis [77].
Transcriptome-wide association studies (TWAS)
To identify causal genes and variants, webTWAS integrates seven statistical models from three widely used TWAS software packages: PrediXcan [79]/S-PrediXcan [80], TWAS-FUSION (best-TWAS, BLUP, LASSO, Top1) [81], and UTMOST (joint tissue GBJ model) [82]. Each gene is analyzed across 47 GTEx tissues, generating 47 × 6 + 1 association results, with UTMOST combining results into a single score. The 1000 Genomes Project LD matrix is applied across all models, and a Bonferroni-corrected significance threshold accounts for multiple testing.
Given the tissue-specific nature of gene expression, webTWAS employs tissue-specific enrichment analysis based on TSEA-DB [83] to identify disease-relevant tissues. We used webTWAS to examine the “PCIF1” gene in relation to osteoporosis-related traits, including osteoporosis, osteopenia, fractures, and other bone density disorders. This analysis helped identify genes potentially involved in bone health and osteoporosis pathogenesis.
Subjects and Sample Collection for snRNA-seq analysis
Femoral head tissues were collected from 14 patients undergoing total hip arthroplasty (THA) at Tulane Lakeside Hospital (Metairie, LA, USA) between 2023 and 2024. The sample included individuals aged 42–78 years (mean ± SD: 62.8 ± 11.4), consisting of 6 females, 7 males, and 1 individual with missing sex information. The majority were White (64.3%), followed by African American (28.6%) and Hispanic (7.1%). Detailed sample characteristics are provided in S2 and S3 Tables. THA is a routine orthopedic procedure that provides access to high-quality human bone tissue, which is typically discarded during surgery. Although all participants were diagnosed with osteoarthritis—a condition highly prevalent in the aging population that primarily affects the joint surface—it has limited impact on the underlying trabecular bone, as described in previous studies [30]. Thus, femoral head specimens obtained through THA serve as a practical and ethically appropriate source of bone tissue for investigating age-related skeletal biology. Patients with chronic conditions known to affect bone metabolism—including hepatic or renal dysfunction, thyroid or parathyroid disorders, diabetes, malignancies, malabsorption syndromes, hematologic diseases, or a history of pathological fractures—were excluded. The study protocol was approved by the Tulane University Institutional Review Board (Tulane IRB #: 2022-530), and written informed consent was obtained from all participants. All procedures involving human participants were conducted in accordance with the principles of the Declaration of Helsinki.
Isolation of osteolineage cells from human femoral head tissue
Upon surgical removal, femoral heads were immediately placed in pre-chilled 1× phosphate-buffered saline (PBS) at 4 °C and transported to the Tulane Center for Biomedical Informatica and Genomics laboratory on ice. All subsequent procedures were performed under a biosafety level 2 cabinet, with instruments and surfaces sterilized using 75% ethanol. Residual soft tissue was removed, and bone specimens were cut into approximately 5 mm3 fragments. These fragments were thoroughly rinsed with PBS until the supernatant was free of blood and then temporarily stored in ice-cold alpha minimum essential medium (α-MEM). For enzymatic digestion, bone fragments were incubated in a digestion buffer containing 0.1% collagenase type II (Worthington), DNase I (Sigma), 10% fetal bovine serum (FBS), and antibiotics at 37 °C for 3 hours in a water bath with gentle mixing every 30 min. The resulting cell suspension was filtered through a 40 μm cell strainer and centrifuged. RBCs were lysed using RBC lysis buffer (e.g., BioLegend), followed by a second centrifugation. Cell viability and concentration were assessed using acridine orange/propidium iodide (AOPI) staining and quantified with an automated cell counter. Only samples containing more than 1 × 10⁶ total cells with viability greater than 80% were selected for downstream sorting.
To enrich for osteolineage cells, hematopoietic and immune-derived cells were removed using magnetic-activated cell sorting (MACS; Miltenyi Biotec). First, CD45⁺ cells were depleted by incubating 1 × 10⁷ total cells with 20 μL of CD45 MicroBeads and 80 μL of pre-chilled MACS buffer for 15 min on ice. The cell suspension was then passed through an LS column placed in a SuperMACS II separator, and the flow-through containing CD45⁻ cells was collected. These CD45⁻ cells were subsequently incubated with CD43 MicroBeads under the same conditions and processed through a second LS column to remove CD43⁺ cells. The resulting CD45⁻CD43⁻ cell population was collected and used for downstream scRNA-seq. The total cell count and viability of the sorted CD45⁻CD43⁻ population were reassessed using AOPI staining, performed by mixing 10 μL of chilled AOPI stain with 10 μL of the cell suspension and measured under the “Cell line, viability propidium iodide” setting, using a dilution factor of 2. Samples yielding more than 3 × 10⁵ cells with viability greater than 80% were fixed by adding 100 μL of pre-warmed Enhancer and 275 μL of 50% glycerol to 1 mL of chilled Quenching Buffer. The mixture was gently pipetted until homogeneous and stored at −80 °C for long-term preservation, for up to 6 months.
Single-cell RNA sequencing library preparation and data processing library preparation and sequencing
scRNA-seq libraries were prepared using the 10× Genomics Chromium Single Cell 3′ Library & Gel Bead Kit v3 (Cat. No. 1000075). Approximately 5,000–10,000 sorted CD45⁻ cells per sample were loaded into the Chromium Controller to generate gel bead-in-emulsion (GEM) droplets, enabling cell lysis and reverse transcription of mRNA within nanoliter compartments. Cell viability (>85%) was confirmed using trypan blue staining prior to GEM generation. Following cDNA amplification, sequencing libraries were constructed according to the manufacturer’s protocol and quantified using Qubit and Bioanalyzer. Libraries were sequenced on the Illumina NovaSeq 6000 platform using paired-end 150 bp reads (PE150), with a target depth of at least 100,000 reads per cell to ensure robust transcriptome coverage.
scRNA-seq data processing, annotation, and pathway analysis
scRNA-seq data were demultiplexed, aligned, and count matrices were generated using Cell Ranger v6.1.2 [84] aligning to GRCh38 with intronic regions included. Cell Ranger matrices were directly imported into Seurat v4. Cells [85] with 300–10,000 unique genes expressed, less than 10,000 unique UMIs, and less than 10% mitochondrial reads were retained in the final dataset. Individual samples were processed by log-normalization; the top 2,000 features were identified; the data was scaled; a principal component analysis (PCA) was computed, and the data was clustered via Louvain clustering using the top 30 principal components (PCs). We used DoubletFinder [86] to remove doublets on a per-sample basis using the top 30 PCs, an expected doublet rate equal to the number of cells in the sample or 125,000 cells (a default scaling factor), and other parameters remained as default. Doublet cells were removed, and samples were combined without computational batch correction. The full dataset was processed as above and visualized using UMAP with the top 30 PCs. Differentially expressed genes in each cluster were identified using the FindAllMarkers function in Seurat v4 with the Wilcoxon rank-sum test downsampled to a maximum of 1,000 cells per cluster, and these results were recorded in S1 Fig. Differentiation states of MSCs lineage were inferred using CytoTRACE (v0.3.3) [87] with default parameters using the log-normalized counts. To reconstruct the continuous differentiation trajectory from MSCs to mature osteoblasts (Mature_OB), we applied Monocle 2 (v2.4.0) [88]. The integrated Seurat object was converted into a CellDataSet object using the SeuratWrappers package, maintaining gene expression and metadata integrity. Genes used for ordering were selected based on differential expression across clusters, and dimensionality reduction was performed using DDRTree algorithm included in Monocle 2.
Generation of global and conditional Pcif1 knockout mice
For global knockout, the Pcif1−/− mice were generated by targeting at the exon 5, 6, and 7 of Pcif1 gene [16]. For conditional knockout, Pcif1fl/+ mice were generated by inserting dual loxP sites targeting exon 5, 6, and 7 [16]. MSC-specific Pcif1 knockout mice were obtained by mating Pcif1fl/+ mice with Prrx1-Cre mice (The Jackson Laboratory). Myeloid cell-specific Pcif1 knockout mice were achieved by crossing Pcif1fl/+ mice with LysM-Cre mice (The Jackson Laboratory). Prrx1-Cre;Pcif1fl/fl mice aged 3 weeks received weekly intraperitoneal injection of the WNT agonist rhRSPO2 (4 mg/kg) for five weeks and then sacrificed for subsequent examinations. All mice used in the study had a C57BL/6J genetic background, and were housed in SPF facilities following a 12-hour light-dark cycle (Animal Care Center of State Key Laboratory of Oral Diseases). All animal experiments were performed in compliance with institutional animal care and ethical guidelines (Research Ethics Committee of West China Hospital of Stomatology).
Micro-CT and histomorphometric analysis
Femurs isolated from 3- or 6-week-old mice were fixed before scanned by a μCT 80 cabinet micro-CT scanner (Scanco Medical, Switzerland). The femurs were imaged with a spatial resolution of 8 μm (70 kV, 200 mA, 0.5 mm aluminum filter, 300 ms integration time). The bone parameters were measured using the SCANCO Evaluation software (version 1.1.19.0, Scanco Medical). Visualization of trabecular and cortical bone was conducted using SCANCO Visualizer software (version 1.1.18.0, Scanco Medical).
For histomorphometric evaluation, decalcified samples were prepared as five-μm-thick sections for H&E staining (Solarbio, China) and TRAP staining (Solarbio, China). For Von Kossa staining and calcein double-labeling experiment, undecalcified femurs were prepared as ten-μm-thick sections using a Leica CM3050 S cryostat (Leica Biosystems, USA). All parameters of trabecular bone of distal femurs were calculated by OsteoMeasure software (OsteoMetrics, USA).
Cell culture and mineralization assay
Murine MSCs were extracted from femurs of 6-week-old mice, and cultured in α-MEM medium (Gibco, USA) supplemented with 10% FBS, 100 units/mL penicillin and 100 μg/mL streptomycin (all from Gibco, USA). For osteogenic induction, the osteogenic medium was supplemented with 50 μg/mL ascorbic acid, 5 mM β-glycerophosphate, and 100 nM dexamethasone (all from Sigma), in addition to the original culture medium.
MSCs treated with osteogenic medium for 7 days were fixed and stained with ALP staining kit (Beyotime, China). Quantitative ALP activity was measured according to the manufacturer’s protocols (Nanjing Jiancheng, China). For ARS staining, MSCs receiving osteogenic induction for 14 days were stained with 1% Alizarin red S (Solarbio, China). Quantification of mineralization was assessed through the OD value at 450 nm by a spectrophotometer (Thermo Fisher Scientific) after destained with 10% cetylpyridinium chloride.
Bone marrow cells were isolated and treated with ACK lysis buffer, and then cultured overnight at 37°C. Nonadherent BMDMs were collected and cultured in DMEM complete medium (Gibco, USA) supplemented with 50 ng/mL M-CSF and 50 ng/mL RANKL (R&D Systems, USA) for 5 days.
Quantitative RT-PCR and western blot
RNA was extracted from MSCs after a 7-day osteogenic induction by Trizol reagent (Invitrogen, USA) and treated with a cDNA Synthesis kit following the instructions (Yeasen, China). The expression of each gene was measured using a 2−ΔΔCt method and normalized to Gapdh (Yeasen, China). Proteins were obtained by a total protein extraction kit (Cell signaling, USA), and then subjected to western blot analyses. The primers and antibodies used in the manuscript were listed in S4 and S5 Tables.
Bulk RNA sequencing
MSCs from Prrx1-Cre;Pcif1fl/fl mice and Pcif1fl/fl littermates were cultured in osteogenic medium for 7 days. Total RNAs were extracted by Trizol reagent (Invitrogen, USA) and purified by using poly-T oligo-attached magnetic beads. Sequencing libraries were generated and subjected to Illumina PE150. Clean data was processed and mapped to the Mus musculus reference genome (mm10) by HISAT2. Differentially expressed genes were selected by DESeq2 R package with |fold change| ≥ 2.0 and padj ≤ 0.05. GO and KEGG enrichment analyses were performed using clusterProfiler R package.
m6Am-Exo-Seq analysis
m6Am-Exo-Seq datasets were obtained from the published database GSE151229, and mapped to the Mus musculus reference genome (mm10). m6Am peaks of Wnt11, Fzd4 and Fgfr2 were visualized using IGV software.
m6AmPred prediction
Predictions of m6Am modification sites were performed using the m6AmPred website (http://180.208.58.19/m6ampred/index.html) [42]. The sequences of mouse Wnt11, Fzd4, and Fgfr2 transcripts were retrieved from the National Center for Biotechnology Information (NCBI) database. The calculated likelihood ratio is capped at 99, and the higher values indicate the greater possibility. The candidate m6Am sites within 500 nt downstream of the TSS were selected for further analyses.
MeRIP-qPCR
The enrichment of m6Am-modified RNA fragments was performed using the EpiQuik CUT&RUN m6A RNA Enrichment (MeRIP) Kit (EpigenTek, USA) according to the manufacturer’s instructions. Total RNA of MSCs was extracted and subjected to enrichment procedures. RNA fragments containing m6Am were captured, purified and reverse transcribed followed by RT-qPCR using site-specific primers. The relative m6Am enrichment levels were calculated as the percentage of the m6Am signal in the IP group compared to that in the IgG control.
RNA decay assay
MSCs from Prrx1-Cre;Pcif1fl/fl mice and Pcif1fl/fl littermates were seeded at the concentration of 300 thousand per well and subjected to osteogenic induction for 7 days. Then, the cells were treated with actinomycin D (10 μg/mL) for 0, 2, 4, and 6 hours. Total RNA was harvested at pre-determined time points for qRT-PCR analyses. Gene expression was normalized to t = 0.
Dual-luciferase reporter assay
293T cells were seeded in 24-well plates and co-transfected with 500 ng of the reporter plasmid and PCIF1 siRNA, along with a PCIF1 overexpression plasmid or a catalytically inactive PCIF1 mutant plasmid. The PCIF1 overexpression plasmid and PCIF1 mutant plasmid (aa160–163, LFPD to AFPA) were constructed as reported [13,15]. After 48 h, Firefly and Renilla luciferase activities were measured using the Dual-Luciferase Reporter Assay System (Promega, USA). Firefly luciferase activity was normalized to Renilla.
Statistical analysis
All data were expressed as mean ± SD. Comparison between two independent groups was performed using an unpaired two-tailed Student t test. Comparisons among multiple groups (i.e., more than two groups) were conducted using one-way ANOVA, followed by Tukey’s post hoc test for pairwise comparisons. A p-value below 0.05 was regarded as statistically significant.
Supporting information
S1 Fig. Heatmap showing normalized gene expression, scaled by row (gene), for key cell lineage marker genes across the corresponding cell clusters.
https://doi.org/10.1371/journal.pbio.3003739.s001
(TIF)
S2 Fig. Global deletion of Pcif1 decreases bone mass of female mice.
(A) Representative Micro-CT images of femurs from 6-week-old female Pcif1−/− knockout mice and Pcif1+/+ littermates. Scale bar, 200 μm. (B) Quantitative analyses of distal trabecular bone and midshaft cortical thickness of femurs (n = 8). (C) Representative images of Von Kossa staining of undecalcified femoral sections. Scale bar, 500 μm. (D) Representative TRAP images of trabecular bone from the femoral metaphysis. Scale bar, 50 μm. (E) Quantitation of osteoblast and osteoclast numbers of trabecular bone (n = 8). (F) Representative images demonstrating calcein double-labeling patterns in trabecular bone from the femoral metaphysis. Scale bar, 20 μm. (G) Quantitation of mineralization apposition rate (MAR) and bone formation rate (BFR/BS) of trabecular bone (n = 8). The data underlying panels B, E, and G can be found in S1 Data (Sheet S2 Fig).
https://doi.org/10.1371/journal.pbio.3003739.s002
(TIF)
S3 Fig. LysM-driven Pcif1 knockout does not cause changes of bone mass of female mice.
(A) Representative Micro-CT images of femurs from 6-week-old female LysM-Cre;Pcif1fl/fl mice and Pcif1fl/fl littermates. Scale bar, 200 μm. (B) Quantitation of bone parameters (n = 8). The data underlying panel B can be found in S1 Data (Sheet S3 Fig).
https://doi.org/10.1371/journal.pbio.3003739.s003
(TIF)
S4 Fig. Bulk RNA-seq and m6Am-exo-seq.
(A) Volcano plot showing differentially expressed genes. (B) qRT-PCR analysis of Wnt11, Fzd4, and Fgfr2 in MSCs cultured in vitro without osteogenic induction (n = 3). (C) Diagram of integrative analysis. The data underlying panel B can be found in S1 Data (Sheet S4 Fig).
https://doi.org/10.1371/journal.pbio.3003739.s004
(TIF)
S1 Table. GWAS summary statistics data source.
https://doi.org/10.1371/journal.pbio.3003739.s005
(PDF)
S3 Table. Sample characteristics stratified by sex (with missing category).
https://doi.org/10.1371/journal.pbio.3003739.s007
(PDF)
S1 File. Bone-related SNPs for m6Am regulators.
https://doi.org/10.1371/journal.pbio.3003739.s010
(XLSX)
S2 File. GTEx eQTL outputs of m6Am regulators.
https://doi.org/10.1371/journal.pbio.3003739.s011
(XLSX)
S1 Data. Excel file with all individual numerical values corresponding to the data presented in the main and supporting figures.
https://doi.org/10.1371/journal.pbio.3003739.s017
(XLSX)
Acknowledgments
We acknowledge Xiao Zhang, Qiang Guo, Yushun Wang, Ning Ji, Yuwen Luo, and Xuelin Huang (State Key Laboratory of Oral Diseases) for technique support. We also acknowledge the Louisiana Optical Network Infrastructure (LONI) and Tulane’s Cypress High Performance Computing System for providing the computational resources used in this study.
References
- 1. Li Q, Jiang S, Lei K, Han H, Chen Y, Lin W, et al. Metabolic rewiring during bone development underlies tRNA m7G-associated primordial dwarfism. J Clin Invest. 2024;134(20):e177220. pmid:39255038
- 2. Sheng R, Meng W, Zhang Z, Yin Q, Jiang S, Li Q, et al. METTL3 regulates cartilage development and homeostasis by affecting Lats1 mRNA stability in an m6A-YTHDF2-dependent manner. Cell Rep. 2024;43(8):114535. pmid:39088322
- 3. Yang C, Dong Z, Ling Z, Chen Y. The crucial mechanism and therapeutic implication of RNA methylation in bone pathophysiology. Ageing Res Rev. 2022;79:101641. pmid:35569786
- 4. You Y, Liu J, Zhang L, Li X, Sun Z, Dai Z, et al. WTAP-mediated m6A modification modulates bone marrow mesenchymal stem cells differentiation potential and osteoporosis. Cell Death Dis. 2023;14(1):33. pmid:36650131
- 5. Zhou X, Yang X, Huang S, Lin G, Lei K, Wang Q, et al. Inhibition of METTL3 alleviates NLRP3 inflammasome activation via increasing ubiquitination of NEK7. Adv Sci (Weinh). 2024;11(26):e2308786. pmid:38696610
- 6. Akichika S, Hirano S, Shichino Y, Suzuki T, Nishimasu H, Ishitani R, et al. Cap-specific terminal N6-methylation of RNA by an RNA polymerase II-associated methyltransferase. Science. 2019;363(6423):eaav0080. pmid:30467178
- 7. Liu J, Li K, Cai J, Zhang M, Zhang X, Xiong X, et al. Landscape and regulation of m6A and m6Am methylome across human and mouse tissues. Mol Cell. 2020;77(2):426-440.e6.
- 8. Boulias K, Toczydłowska-Socha D, Hawley BR, Liberman N, Takashima K, Zaccara S, et al. Identification of the m6Am methyltransferase PCIF1 reveals the location and functions of m6Am in the transcriptome. Mol Cell. 2019;75(3):631-643.e8. pmid:31279658
- 9. Sendinc E, Valle-Garcia D, Dhall A, Chen H, Henriques T, Navarrete-Perea J, et al. PCIF1 catalyzes m6Am mRNA methylation to regulate gene expression. Mol Cell. 2019;75(3):620-630.e9. pmid:31279659
- 10. Sun H, Li K, Zhang X, Liu J, Zhang M, Meng H, et al. m6Am-seq reveals the dynamic m6Am methylation in the human transcriptome. Nat Commun. 2021;12(1):4778. pmid:34362929
- 11. Wei J, Liu F, Lu Z, Fei Q, Ai Y, He PC, et al. Differential m6A, m6Am, and m1A demethylation mediated by FTO in the cell nucleus and cytoplasm. Molecular Cell. 2018;71(6):973-985.e5.
- 12. An H, Hong Y, Goh YT, Koh CWQ, Kanwal S, Zhang Y, et al. m6Am sequesters PCF11 to suppress premature termination and drive neuroblastoma differentiation. Mol Cell. 2024;84(21):4142-4157.e14. pmid:39481383
- 13. Zhuo W, Sun M, Wang K, Zhang L, Li K, Yi D, et al. m6Am methyltransferase PCIF1 is essential for aggressiveness of gastric cancer cells by inhibiting TM9SF1 mRNA translation. Cell Discov. 2022;8(1):48. pmid:35597784
- 14. Wang L, Wu L, Zhu Z, Zhang Q, Li W, Gonzalez GM, et al. Role of PCIF1-mediated 5’-cap N6-methyladeonsine mRNA methylation in colorectal cancer and anti-PD-1 immunotherapy. EMBO J. 2023;42(2):e111673. pmid:36514940
- 15. Li K, Chen J, Zhang C, Cheng M, Chen S, Song W, et al. The CTBP2-PCIF1 complex regulates m6Am modification of mRNA in head and neck squamous cell carcinoma. J Clin Invest. 2023;133(20):e170173. pmid:37643007
- 16. Song W, Liu L, Liang H, Cheng H, He W, Yin Q, et al. m6Am methyltransferase PCIF1 regulates periodontal inflammation. J Dent Res. 2024;103(11):1130–40.
- 17. Xiang B, Zhang M, Li K, Zhang Z, Liu Y, Gao M, et al. The epitranscriptional factor PCIF1 orchestrates CD8+ T cell ferroptosis and activation to control antitumor immunity. Nat Immunol. 2025;26(2):252–64. pmid:39762445
- 18. Wu Y, Xie L, Wang M, Xiong Q, Guo Y, Liang Y, et al. Mettl3-mediated m6A RNA methylation regulates the fate of bone marrow mesenchymal stem cells and osteoporosis. Nat Commun. 2018;9(1):4772. pmid:30429466
- 19. Zhang Q, Riddle RC, Yang Q, Rosen CR, Guttridge DC, Dirckx N, et al. The RNA demethylase FTO is required for maintenance of bone mass and functions to protect osteoblasts from genotoxic damage. Proc Natl Acad Sci U S A. 2019;116(36):17980–9. pmid:31434789
- 20. Yahara Y, Barrientos T, Tang YJ, Puviindran V, Nadesan P, Zhang H, et al. Erythromyeloid progenitors give rise to a population of osteoclasts that contribute to bone homeostasis and repair. Nat Cell Biol. 2020;22(1):49–59. pmid:31907410
- 21. Zheng C-X, Sui B-D, Qiu X-Y, Hu C-H, Jin Y. Mitochondrial regulation of stem cells in bone homeostasis. Trends Mol Med. 2020;26(1):89–104. pmid:31126872
- 22. Wang Y, Li Y, Cai Y, Yang X, Li H, Wang Q, et al. Dimethyl citraconate alleviates periodontitis via activating the NRF2 cascade. J Dent Res. 2025;104(8):910–9. pmid:40289519
- 23. Tang Y, Wu X, Lei W, Pang L, Wan C, Shi Z, et al. TGF-beta1-induced migration of bone mesenchymal stem cells couples bone resorption with formation. Nat Med. 2009;15(7):757–65. pmid:19584867
- 24. Guan M, Yao W, Liu R, Lam KS, Nolta J, Jia J, et al. Directing mesenchymal stem cells to bone to augment bone formation and increase bone mass. Nat Med. 2012;18(3):456–62. pmid:22306732
- 25. Salhotra A, Shah HN, Levi B, Longaker MT. Mechanisms of bone development and repair. Nat Rev Mol Cell Biol. 2020;21(11):696–711. pmid:32901139
- 26. Zhu S, Chen W, Masson A, Li Y-P. Cell signaling and transcriptional regulation of osteoblast lineage commitment, differentiation, bone formation, and homeostasis. Cell Discov. 2024;10(1):71. pmid:38956429
- 27. Deng P, Yuan Q, Cheng Y, Li J, Liu Z, Liu Y, et al. Loss of KDM4B exacerbates bone-fat imbalance and mesenchymal stromal cell exhaustion in skeletal aging. Cell Stem Cell. 2021;28(6):1057-1073.e7. pmid:33571444
- 28. Wei Q, Nakahara F, Asada N, Zhang D, Gao X, Xu C, et al. Snai2 maintains bone marrow niche cells by repressing osteopontin expression. Dev Cell. 2020;53(5):503-513.e5. pmid:32413329
- 29. Tran B, Nguyen ND, Center JR, Eisman JA, Nguyen TV. Association between fat-mass-and-obesity-associated (FTO) gene and hip fracture susceptibility. Clin Endocrinol (Oxf). 2014;81(2):210–7. pmid:24106974
- 30. Bandyopadhyay S, Duffy MP, Ahn KJ, Sussman JH, Pang M, Smith D, et al. Mapping the cellular biogeography of human bone marrow niches using single-cell transcriptomics and proteomic imaging. Cell. 2024;187(12):3120-3140.e29. pmid:38714197
- 31. Wang L, You X, Lotinun S, Zhang L, Wu N, Zou W. Mechanical sensing protein PIEZO1 regulates bone homeostasis via osteoblast-osteoclast crosstalk. Nat Commun. 2020;11(1):282. pmid:31941964
- 32. McDonald MM, Khoo WH, Ng PY, Xiao Y, Zamerli J, Thatcher P, et al. Osteoclasts recycle via osteomorphs during RANKL-stimulated bone resorption. Cell. 2021;184(5):1330-1347.e13. pmid:33636130
- 33. Cao S, Li Y, Shen L, Shao B, Yu L, Li J, et al. Functionalized virus nanoparticles alleviates osteoporosis via targeting the function of RANK-specific motifs. ACS Appl Mater Interfaces. 2023;15(27):32272–80. pmid:37381119
- 34. Akoumianakis I, Polkinghorne M, Antoniades C. Non-canonical WNT signalling in cardiovascular disease: mechanisms and therapeutic implications. Nat Rev Cardiol. 2022;19(12):783–97. pmid:35697779
- 35. Chae W-J, Bothwell ALM. Canonical and non-canonical Wnt signaling in immune cells. Trends Immunol. 2018;39(10):830–47. pmid:30213499
- 36. Holzem M, Boutros M, Holstein TW. The origin and evolution of Wnt signalling. Nat Rev Genet. 2024;25(7):500–12. pmid:38374446
- 37. Ko S-B, Mihara E, Park Y, Roh K, Kang C, Takagi J, et al. Functional role of the Frizzled linker domain in the Wnt signaling pathway. Commun Biol. 2022;5(1):421. pmid:35513706
- 38. Kushwaha P, Kim S, Foxa GE, Michalski MN, Williams BO, Tomlinson RE, et al. Frizzled-4 is required for normal bone acquisition despite compensation by Frizzled-8. J Cell Physiol. 2020;235(10):6673–83. pmid:31985040
- 39. Xie Y, Su N, Yang J, Tan Q, Huang S, Jin M, et al. FGF/FGFR signaling in health and disease. Signal Transduct Target Ther. 2020;5(1):181. pmid:32879300
- 40. Kluiver TA, Lu Y, Schubert SA, Kraaier LJ, Ringnalda F, Lijnzaad P, et al. Divergent WNT signaling and drug sensitivity profiles within hepatoblastoma tumors and organoids. Nat Commun. 2024;15(1):8576. pmid:39567475
- 41. Zhu Q, Fu Y, Cui C-P, Ding Y, Deng Z, Ning C, et al. OTUB1 promotes osteoblastic bone formation through stabilizing FGFR2. Signal Transduct Target Ther. 2023;8(1):142. pmid:37024477
- 42. Jiang J, Song B, Chen K, Lu Z, Rong R, Zhong Y. m6AmPred: Identifying RNA N6, 2′-O-dimethyladenosine (m6Am) sites based on sequence-derived information. Methods. 2022;203:328–34.
- 43. Liu Y, Fan M, Yang J, Mihaljević L, Chen KH, Ye Y, et al. KAT6A deficiency impairs cognitive functions through suppressing RSPO2/Wnt signaling in hippocampal CA3. Sci Adv. 2024;10(20):eadm9326. pmid:38758792
- 44. Wang L, Hu F, Cui Q, Qiao H, Li L, Geng T, et al. Structural insights into the LGR4-RSPO2-ZNRF3 complexes regulating WNT/β-catenin signaling. Nat Commun. 2025;16(1):362. pmid:39753551
- 45. Sun R, He L, Lee H, Glinka A, Andresen C, Hübschmann D, et al. RSPO2 inhibits BMP signaling to promote self-renewal in acute myeloid leukemia. Cell Rep. 2021;36(7):109559. pmid:34407399
- 46. Yue Z, Niu X, Yuan Z, Qin Q, Jiang W, He L, et al. RSPO2 and RANKL signal through LGR4 to regulate osteoclastic premetastatic niche formation and bone metastasis. J Clin Invest. 2022;132(2):e144579. pmid:34847079
- 47. Zhang Q, Kang Y, Wang S, Gonzalez GM, Li W, Hui H, et al. HIV reprograms host m6Am RNA methylome by viral Vpr protein-mediated degradation of PCIF1. Nat Commun. 2021;12(1):5543. pmid:34545078
- 48. Tartell MA, Boulias K, Brunsting EL, Bloyet L-M, Greer EL, Whelan SPJ. Methylation of viral mRNA cap structures by PCIF1 attenuates the antiviral activity of interferon-β. Proc Natl Acad Sci U S A. 2021;118(29):e2025769118. pmid:34266951
- 49. Wang L, Wang S, Wu L, Li W, Bray W, Clark AE, et al. PCIF1-mediated deposition of 5’-cap N6,2’-O-dimethyladenosine in ACE2 and TMPRSS2 mRNA regulates susceptibility to SARS-CoV-2 infection. Proc Natl Acad Sci U S A. 2023;120(5):e2210361120. pmid:36689652
- 50. Sun H, Zhang M, Li K, Bai D, Yi C. Cap-specific, terminal N6-methylation by a mammalian m6Am methyltransferase. Cell Res. 2018;29(1):80–2.
- 51. Pandey RR, Delfino E, Homolka D, Roithova A, Chen K-M, Li L, et al. The mammalian Cap-specific m6Am RNA methyltransferase PCIF1 regulates transcript levels in mouse tissues. Cell Rep. 2020;32(7):108038. pmid:32814042
- 52. Sheng R, Wang Y, Wu Y, Wang J, Zhang S, Li Q, et al. METTL3-mediated m6 A mRNA methylation modulates tooth root formation by affecting NFIC translation. J Bone Miner Res. 2021;36(2):412–23. pmid:32936965
- 53. Xu R, Sheng R, Lin W, Jiang S, Zhang D, Liu L, et al. METTL3 modulates Ctsk+ lineage supporting cranial osteogenesis via hedgehog. J Dent Res. 2024;103(7):734–44. pmid:38752256
- 54. Geula S, Moshitch-Moshkovitz S, Dominissini D, Mansour AA, Kol N, Salmon-Divon M, et al. Stem cells. m6A mRNA methylation facilitates resolution of naïve pluripotency toward differentiation. Science. 2015;347(6225):1002–6. pmid:25569111
- 55. Udagawa N, Koide M, Nakamura M, Nakamichi Y, Yamashita T, Uehara S, et al. Osteoclast differentiation by RANKL and OPG signaling pathways. J Bone Miner Metab. 2021;39(1):19–26. pmid:33079279
- 56. Thomas GP, Baker SU, Eisman JA, Gardiner EM. Changing RANKL/OPG mRNA expression in differentiating murine primary osteoblasts. J Endocrinol. 2001;170(2):451–60. pmid:11479141
- 57. Matsushita Y, Nagata M, Kozloff KM, Welch JD, Mizuhashi K, Tokavanich N, et al. A Wnt-mediated transformation of the bone marrow stromal cell identity orchestrates skeletal regeneration. Nat Commun. 2020;11(1):332. pmid:31949165
- 58. Joeng KS, Lee Y-C, Lim J, Chen Y, Jiang M-M, Munivez E, et al. Osteocyte-specific WNT1 regulates osteoblast function during bone homeostasis. J Clin Invest. 2017;127(7):2678–88. pmid:28628032
- 59. Okuchi Y, Reeves J, Ng SS, Doro DH, Junyent S, Liu KJ, et al. Wnt-modified materials mediate asymmetric stem cell division to direct human osteogenic tissue formation for bone repair. Nat Mater. 2021;20(1):108–18. pmid:32958876
- 60. Luther J, Yorgan TA, Rolvien T, Ulsamer L, Koehne T, Liao N, et al. Wnt1 is an Lrp5-independent bone-anabolic Wnt ligand. Sci Transl Med. 2018;10(466):eaau7137. pmid:30404864
- 61. Hu L, Chen W, Qian A, Li Y-P. Wnt/β-catenin signaling components and mechanisms in bone formation, homeostasis, and disease. Bone Res. 2024;12(1):39. pmid:38987555
- 62. Rauner M, Baschant U, Roetto A, Pellegrino RM, Rother S, Salbach-Hirsch J, et al. Transferrin receptor 2 controls bone mass and pathological bone formation via BMP and Wnt signaling. Nat Metab. 2019;1(1):111–24. pmid:30886999
- 63. Pandur P, Läsche M, Eisenberg LM, Kühl M. Wnt-11 activation of a non-canonical Wnt signalling pathway is required for cardiogenesis. Nature. 2002;418(6898):636–41. pmid:12167861
- 64. Jiang W, Guan B, Sun H, Mi Y, Cai S, Wan R, et al. WNT11 Promotes immune evasion and resistance to Anti-PD-1 therapy in liver metastasis. Nat Commun. 2025;16(1):1429. pmid:39920102
- 65. Friedman MS, Oyserman SM, Hankenson KD. Wnt11 promotes osteoblast maturation and mineralization through R-spondin 2. J Biol Chem. 2009;284(21):14117–25. pmid:19213727
- 66. Caetano da Silva C, Edouard T, Fradin M, Aubert-Mucca M, Ricquebourg M, Raman R, et al. WNT11, a new gene associated with early onset osteoporosis, is required for osteoblastogenesis. Hum Mol Genet. 2022;31(10):1622–34. pmid:34875064
- 67. Gu Q, Tian H, Zhang K, Chen D, Chen D, Wang X, et al. Wnt5a/FZD4 mediates the mechanical stretch-induced osteogenic differentiation of bone mesenchymal stem cells. Cell Physiol Biochem. 2018;48(1):215–26. pmid:30007964
- 68. Xu W, Luo F, Wang Q, Tan Q, Huang J, Zhou S, et al. Inducible activation of FGFR2 in adult mice promotes bone formation after bone marrow ablation. J Bone Miner Res. 2017;32(11):2194–206. pmid:28650109
- 69. Merrill AE, Sarukhanov A, Krejci P, Idoni B, Camacho N, Estrada KD, et al. Bent bone dysplasia-FGFR2 type, a distinct skeletal disorder, has deficient canonical FGF signaling. Am J Hum Genet. 2012;90(3):550–7. pmid:22387015
- 70. Zhang H, Wang Z, Liu Z, Li X, Sun W, Zhen W, et al. Nuclear FGF2 orchestrates phase separation-mediated rDNA chromatin architecture to control BMSCs cell fate. Bone Res. 2025;13(1):80. pmid:40993129
- 71. Szenker-Ravi E, Altunoglu U, Leushacke M, Bosso-Lefèvre C, Khatoo M, Thi Tran H, et al. RSPO2 inhibition of RNF43 and ZNRF3 governs limb development independently of LGR4/5/6. Nature. 2018;557(7706):564–9. pmid:29769720
- 72. Knight MN, Karuppaiah K, Lowe M, Mohanty S, Zondervan RL, Bell S, et al. R-spondin-2 is a Wnt agonist that regulates osteoblast activity and bone mass. Bone Res. 2018;6:24. pmid:30131881
- 73. Liu JZ, McRae AF, Nyholt DR, Medland SE, Wray NR, Brown KM, et al. A versatile gene-based test for genome-wide association studies. Am J Hum Genet. 2010;87(1):139–45. pmid:20598278
- 74. Sudmant PH, Rausch T, Gardner EJ, Handsaker RE, Abyzov A, Huddleston J, et al. An integrated map of structural variation in 2,504 human genomes. Nature. 2015;526(7571):75–81. pmid:26432246
- 75. Durinck S, Spellman PT, Birney E, Huber W. Mapping identifiers for the integration of genomic datasets with the R/Bioconductor package biomaRt. Nat Protoc. 2009;4(8):1184–91. pmid:19617889
- 76. Harrison PW, Amode MR, Austine-Orimoloye O, Azov AG, Barba M, Barnes I, et al. Ensembl 2024. Nucleic Acids Res. 2024;52(D1):D891–D9.
- 77. GTEx Consortium. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science. 2020;369(6509):1318–30. pmid:32913098
- 78. Warwick AN, Zuckerman B, Ung CY, Luben R, Olvera-Barrios A. gtexr: a convenient R interface to the Genotype-Tissue Expression (GTEx) portal API. J Open Source Softw. 2025;10(109):8249.
- 79. Gamazon ER, Wheeler HE, Shah KP, Mozaffari SV, Aquino-Michaels K, Carroll RJ, et al. A gene-based association method for mapping traits using reference transcriptome data. Nat Genet. 2015;47(9):1091–8. pmid:26258848
- 80. Barbeira AN, Dickinson SP, Bonazzola R, Zheng J, Wheeler HE, Torres JM, et al. Exploring the phenotypic consequences of tissue specific gene expression variation inferred from GWAS summary statistics. Nat Commun. 2018;9(1):1825. pmid:29739930
- 81. Gusev A, Ko A, Shi H, Bhatia G, Chung W, Penninx BWJH, et al. Integrative approaches for large-scale transcriptome-wide association studies. Nat Genet. 2016;48(3):245–52. pmid:26854917
- 82. Hu Y, Li M, Lu Q, Weng H, Wang J, Zekavat SM, et al. A statistical framework for cross-tissue transcriptome-wide association analysis. Nat Genet. 2019;51(3):568–76. pmid:30804563
- 83. Jia P, Dai Y, Hu R, Pei G, Manuel AM, Zhao Z. TSEA-DB: a trait-tissue association map for human complex traits and diseases. Nucleic Acids Res. 2020;48(D1):D1022–30. pmid:31680168
- 84. Zheng GXY, Terry JM, Belgrader P, Ryvkin P, Bent ZW, Wilson R, et al. Massively parallel digital transcriptional profiling of single cells. Nat Commun. 2017;8:14049. pmid:28091601
- 85. Butler A, Hoffman P, Smibert P, Papalexi E, Satija R. Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol. 2018;36(5):411–20. pmid:29608179
- 86. McGinnis CS, Murrow LM, Gartner ZJ. DoubletFinder: doublet detection in single-cell RNA sequencing data using artificial nearest neighbors. Cell Syst. 2019;8(4):329-337.e4. pmid:30954475
- 87. Gulati GS, Sikandar SS, Wesche DJ, Manjunath A, Bharadwaj A, Berger MJ, et al. Single-cell transcriptional diversity is a hallmark of developmental potential. Science. 2020;367(6476):405–11. pmid:31974247
- 88. Qiu X, Mao Q, Tang Y, Wang L, Chawla R, Pliner HA, et al. Reversed graph embedding resolves complex single-cell trajectories. Nat Methods. 2017;14(10):979–82. pmid:28825705