Targetome-defined miR-181c signaling from extracellular vesicles governs periodontal MSC fate via RNF150-MAP3K5

Abstract

Background

In regenerative settings, robust differentiation of human mesenchymal stromal cells (hMSCs) requires precise coupling between post-transcriptional regulation and kinase-driven pathways to achieve optimal therapeutic efficacy. miRNA-mediated regulation is widely implicated in stem cell fate control, yet its mechanistic contribution to lineage commitment remains incompletely defined.

Methods

Multi-conditioned periodontal ligament–derived hMSCs were developed as the regenerative model for alveolar bone. Extracellular vesicle (EV)-encapsulated miRNAs were delineated through microarrays. AGO2 RNA-immunoprecipitation sequencing with transcriptome profiling was integrated to establish the miRISC-associated targetome. Protein-protein interactions and signaling hierarchy were examined by Co-IP, WB, and IFC under knockdown/overexpression with pharmacological interventions. The translational efficacy of EV and miRNA was validated using a mouse ligature-induced periodontitis model, followed by µCT and histological analyses.

Results

EVs produced by hMSCs with higher intrinsic osteogenic capacity exhibited enhanced osteo-inductivity, traced to the enrichment of miR-181c. miR-181c was consistently upregulated during osteogenesis, with gain- and loss-of-function producing concordant effects. Mechanistically, the E3 ubiquitin ligase RNF150 emerged as the main target, whose repression reduced MAP3K5 ubiquitin-mediated proteolysis and enabled p38 activation. Local administration of EVs and miR-181c both promoted alveolar bone regeneration process in vivo.

Conclusion

Collectively, EV–miR-181c–RNF150–MAP3K5–p38 axis was proposed linking miRNA-mediated repression to kinase activation and lineage commitment. This work provides the framework for how EV-delivered miRNAs gate hMSCs’ fate decision, with significance for acellular therapeutic modalities in periodontal and skeletal tissues.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12964-026-03153-y.

Highlights

Multi-condition screening establishes miR-181c as a functional determinant of hMSC-EVs.

miR-181c–associated targetome is defined in hMSCs via RIP-seq.

The RNF150/MAP3K5/p38 axis provides a mechanistic link between post-transcriptional regulation and osteogenic differentiation.

Engineered-EVs with exogenous loading support acellular regenerative therapy.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12964-026-03153-y.

Introduction

Despite extensive studies, microRNA (miRNA)–mediated regulation remains largely correlative, as the predicted targets often fail to converge on defined signaling pathways that can causally explain cell fate decisions [1, 2]. In particular, experimental strategies that link miRNA repression to downstream kinase activation through validated molecular intermediates remain limited [3, 4]. This gap is especially critical in regenerative biology, where precise coupling between post-transcriptional regulation and signaling cascades is required to drive robust and directional stem cell lineage commitment [5, 6].

Periodontitis is a progressive destructive disease of tooth-supporting tissues [7, 8] and a major unmet challenge in tissue regeneration, affecting more than 1.1 billion individuals worldwide and compromising both oral and systemic health [9]. Given the limited capacity of conventional surgical approaches to reconstruct the damaged periodontium [10], stem cell–based therapies have attracted increasing attention [11, 12]. We previously demonstrated that autologous transplantation of human mesenchymal stromal cell (hMSC) sheets promoted the formation of oriented periodontal ligament–like and cementum-mimicking structures in vivo [13–15]. However, accumulating evidence indicates that transplanted cells rapidly lose viability within periodontitis lesions [16, 17], suggesting that the therapeutic benefits are mediated less by durable engraftment than by the early secretome, which initiates regenerative responses [18].

As the therapeutic effects of MSCs are predominantly attributed to paracrine signaling, extracellular vesicles (EVs), particularly small EVs (sEVs) with diameters below 200 nm, have emerged as attractive acellular effectors that retain key bioactivities by modulating host MSC functions while reducing constraints associated with cell transplantation [19]. Accumulating evidence supports a central role of EV-associated miRNAs in periodontal regeneration [20, 21]. However, the substantial variability in EV composition and potency—shaped by donor heterogeneity, cellular state, and culture conditions—remains a major barrier to clinical translation [22, 23].

Therefore, it is essential to identify EV cargos that confer reproducible efficacy and elucidate the downstream molecular pathways causally linking EV delivery to specific differentiation outcomes. In the present study, we characterized the mechanisms by which EV-delivered miRNA regulate osteogenesis in a periodontal ligament–derived mesenchymal stromal cell model. By integrating multi-omics screening with experimental validation of miRNA-target interactions, this study establishes the molecular framework in which EV-delivered miR-181c promotes osteogenic lineage commitment through post-transcriptional repression of RNF150 and subsequent modulation of MAP3K5/MAPK signaling. Together with in vivo validation in animal models of periodontitis, these findings identify the EV/miR-181c/RNF150/MAP3K5 axis as a mechanistic basis for the standardized development of EV-based regenerative therapies for periodontal tissue repair.

Methods and materials

Cell lines

This study was conducted in accordance with the Declaration of Helsinki, with ethical approval from the Dental Research Ethics Committee of the Institute of Science Tokyo (D2020-077). Informed consent was obtained from all subjects. Teeth used in this study were free of periodontal disease and periapical pathology and were extracted for orthodontic purposes or due to impaction. Periodontal ligament (PDL) tissue was collected from seven healthy donors. As previously described, hMSCs were isolated from PDL tissues and respectively characterized for intrinsic osteogenic and proliferative capacities to establish primary cell bank [24–26]. Cells were cultured in Minimum Essential Medium alpha (α-MEM) (Thermo Fisher Scientific, 12571063) supplemented with 10% fetal bovine serum (FBS; Thermo Fisher Scientific, A5670701, Lot 253482), 100 U/mL penicillin, and 100 µg/mL streptomycin (Thermo Fisher Scientific, 15140122) ໿at 37 °C in a humidified 5% CO₂ atmosphere. Cells were subcultured with 0.25% trypsin-EDTA (Thermo Fisher Scientific, 25200056) and seeded in T-225 flasks (Corning, 431082) at a density of 1.3 × 10² cells/cm². Multi-flasks and Hyper-flasks (Falcon, 353143 & 10024) were subsequently used for large-scale cell expansion. The culture medium was changed every 3–4 days. Cells from the passage 7 to 10 were used in this study.

Extracellular vesicles collection, isolation, and characterization

Once confluency reached over 95%, cells were gently washed three times with Dulbecco’s phosphate buffered saline (DPBS; Fujiwako, 045-29795), then the culture medium was replaced with serum-and antibiotic-free α-MEM basal medium. Conditioned medium was collected after culturing for 24 h under 5% CO₂ at different settings (passage number: P7 and P10, temperatures: 37 °C and 25 °C).

Adapted from our previous protocol [27, 28], conditioned medium was centrifuged at 300 × g and then 3,000 × g for 10 min each at 4 °C. Further ultrafiltration was achieved with 100 kDa (Amicon Ultra‑14, Merck Millipore) and 10 kDa membrane (Amicon Ultra-4, Merck Millipore). EVs were separated by size-exclusion chromatography (qEV, Gen2 original/35, Izon). Fractions 8 to 10 were collected as the “EVs fraction”, while 11 to 14 were separated as “non-EVs fractions”. Nanoparticle-Tracking Analysis (NTA) platforms included Videodrop (Myriade, France), ZetaView (Particle Metrix, Germany), and NanoFCM Flow-NanoAnalyzer (NanoFCM, China). EVs’ diameter and concentration were measured with the high yielding purity (Fig. S1D), stability, and integrity (Fig. S1E–G). All samples were stored at -80 °C at 5 × 10⁹ particles/mL until further use.

Microarray profiling, RNA-sequencing, and bioinformatics analysis

Multi-conditioned EVs derived from two hMSC lines exhibiting the highest and lowest osteogenic capacities (named as Osteo_High/Low) were profiled in microarray. EV samples were subjected to miRNA purification using the miRNeasy Micro Kit (Qiagen, 217084). Qualified samples were analyzed by microarray, in which hybridization was performed on the 3D-Gene^® Human miRNA Oligo chip platform, followed by raw data extraction, background correction, and normalization. For RNA-seq and RIP-seq (sample preparation explained in later sections), data were obtained from the hMSC line with the highest proliferative/osteogenic capacity. RNA was purified using the RNeasy Mini Kit (Qiagen, 74104). Libraries were constructed via poly(A) enrichment of mRNA and sequenced on the NovaSeqX Plus system (Illumina, San Diego, US) with 150 bp paired-end reads. Low-quality and adaptor sequences were trimmed using fastp (v0.22.0). Processed paired-end reads were aligned to the human reference genome (GRCh38.p14) with HISAT2 (v2.1.0). Transcript assembly and abundance estimation were performed using StringTie (v2.1.1). Differential expression analysis was carried out with the edgeR (v4.0.16) and limma_3.58.1 (v1.32.0) in R, applying thresholds of adjusted P < 0.01, FDR < 0.1, and fold change ≥ 1.5. The P-values were corrected for multiple testing using the Benjamini-Hochberg method and converted to q values. Gene set enrichment analysis (GSEA, v4.3.3) was performed from the Molecular Signatures Database (MSigDB) under default settings. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted using the DAVID tool (https://davidbioinformatics.nih.gov/) (for KEGG adjusted P < 0.05, FDR < 0.1; for GO: adjusted P < 0.01, FDR < 0.05). Heatmaps, Venn diagrams, and enrichment plots of differentially expressed genes (DEGs) were generated via the OmicStudio (https://www.omicstudio.cn/).

Western blot

To detect protein expression on EVs, cells, antibody-conjugated beads, or lysates, proteins were extracted using RIPA buffer (Cell Signaling Technology, #9806) supplemented with Protease/Phosphatase Inhibitor (Cell Signaling Technology, #5872). Protein concentrations were measured by Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, USA). Electrophoresis was performed on NuPAGE Bis-Tris Mini Protein Gels (4–12%, Invitrogen, NP0322BOX) and transferred to PVDF (Invitrogen, IB24002). After blocking (Cytiva Lifescience, RPN418; Blocking One-P, Nacalai, 05999-84), the membranes were incubated overnight at 4 °C with primary antibodies, followed by TBS buffer washing (Bio-Rad, #1706435), and then incubated for 1 h with the host-matched secondary antibodies. Bands were visualized using a chemiluminescence kit (ECL, Cytiva Lifescience, RPN2209) and imaged with a FUSION SOLO Chemiluminescence System (Vilber, France). Band intensities were quantified by ImageJ (v1.8.0, NIH, Bethesda, MD, USA) and normalized by α-Tubulin. Primary antibodies were used at the manufacturers’ recommended dilutions, including: α-Tubulin (Cell Signaling Technology, #3873), CD9 (Fujifilm, 014-27763), CD63 (Fujifilm, 012-27063), Flotillin-1 (Cell Signaling Technology, #18634), Flotillin-2 (C42A3, Cell Signaling Technology, #3436), Annexin V (Cell Signaling Technology, #8555S), GP96/HSP90B1 (R&D, MAB7606), Calnexin (Cell Signaling Technology, #2679S), Anti-DDDDK-tag mAb (MBL, M185-3 L), Anti-MYC-tag pAb (MBL, 562), ASK1 (F-9) (Santa Cruz, sc-5294), p38 MAPK (Cell Signaling Technology, #9212), phospho-p38 MAPK (Thr180/Tyr182, Cell Signaling Technology, #9211), and Ubiquitin (Abcam, ab7780). Secondary antibodies were goat anti-rabbit IgG HRP-conjugated (R&D, HAF008) and goat anti-mouse IgG HRP-conjugated (R&D, HAF007). All antibodies were diluted in Can Get Signal Solution™ (Toyobo, NKB-201).

Osteogenic differentiation, alkaline-phosphatase (ALP) activity, and alizarin-red staining

For osteogenesis-related experiments, after cell adherence, complete medium was replaced with osteogenic induction medium (OIM) supplemented with 100 µM ascorbic acid (Fujifilm, 019-12063), 10 mM β-glycerophosphate (Sigma-Aldrich, G9422), and 10 nM dexamethasone (Fuji Pharma, 804Y3088). For experiments involving miRNA or siRNA transfection, antibiotics were omitted in OIM for the entire culture period. After seven days of osteogenic induction, alkaline phosphatase (ALP) activity was quantified (Wako Pure Chemical Industries, 297-93501) according to the manufacturer’s protocol, with absorbance measured at the optical density (O.D.) of 405 nm by microplate reader (SpectraMax ABS Plus, Molecular Devices). To assess osteogenic mineralization, cells were stained with 1% Alizarin Red S solution (Fujifilm Wako, 011-01192) at day 14 and 21 following standard protocols. Calcium deposition was observed by scanning images and microscope analysis (NIKON, Eclipse Ti2).

Real-time polymerase chain reaction (RT-qPCR)

Total RNA was extracted by the RNeasy mini kit (Qiagen, Dusseldorf, Germany) followed by the cDNA synthesis using the Superscript VILO Kit (Invitrogen). Total miRNA was purified by miRNeasy Micro Kit (Qiagen, 217084) and reverse transcribed by TaqMan™ Advanced miRNA cDNA Synthesis Kit (Invitrogen, A28007). For quantitative analysis of miRNA or mRNA expression levels, Real-Time PCR (QuantStudio 3, Applied Biosystems) was performed. The specific Taqman™ Gene Expression assay probes (Invitrogen, 4453320) utilized in this study were as listed in the resources table. The expression levels were determined as fold changes through the 2^−ΔΔCt method.

miRNA/siRNA transfection and lentivirus infection

To achieve miRNA overexpression and knockdown, transient transfection was performed using Lipofectamine RNAiMAX (Invitrogen, 13778030). Commercially available miRNA mimics and inhibitors were purchased from Invitrogen, including: hsa-miR-181c-5p (MC10181, MH10181), hsa-miR-378a-3p (MC11360), hsa-miR-584-5p (MC11470), and hsa-miR-339-5p (MC12347). mirVana™ miRNA Mimic Negative Control #1 (4464058) and miRNA Inhibitor Negative Control #1 (4464076) were used as negative controls. For gene knockdown, Sigma-Aldrich predesigned siRNAs were used as listed in the resources table. MISSION^® siRNA Universal Negative Control #2 (SIC002) served as a negative control. To exclude potential off-target effects of siRNAs, all siRNA sequences were analyzed using the NCBI BLASTN suite (https://blast.ncbi.nlm.nih.gov/) and confirmed with specificity. Plasmid vectors and lentivirus were synthesized and packaged by VectorBuilder (Guangzhou, China), driven by the human EF1A promoter and encoded either RNF150 mRNA tagged with EGFP and MYC (pLV\[Exp]-EGFP\T2A\Puro-EF1A> hRNF150\[NM\_020724.2]/MYC) or ASK1 mRNA tagged with FLAG (pLV\[Exp]-Puro-EF1A> hMAP3K5\[NM\_005923.4]/FLAG). Empty control vector (pLV\[Exp]-EGFP/Puro-EF1A> mCherry, VB010000-9492agg) served as the negative control (vector designs are provided in Fig. S6). For transduction, the virus was titrated to 1 × 10⁸ TU/mL at a 1:500 MOI dilution. After 48 h post-infection, the cells were harvested for protein extraction or further downstream analysis.

RNA-immunoprecipitation and miRNA algorithmic prediction

RNA immunoprecipitation (RIP) coupled with high-throughput sequencing is an advanced tool for systematically identifying the binding of miRNAs and target genes within the miRNA-induced Silencing Complex (miRISC) [29]. According to the protocol of the RiboCluster Profile RIP-Assay Kit (MBL, cat# RN1005), the RNA-binding protein immunoprecipitation (RIP) was performed using anti-AGO2 antibody (MBL, RN003M). Samples were collected from cells overexpressing either Negative Control mimic or miR-181c mimic. In detail, 1.5 × 10⁷ cells were transfected with 100 nM mimic for 24 h, lysed in 0.5 mL of miLysis buffer containing PMSF (Cell Signal Technology, #8553) and Ribonuclease Inhibitor (Invitrogen, 10777019) with nuclear fraction removed. Following the instruction of Pierce™ Protein G Plus Agarose (Thermo Fisher Scientific; Cat#20397), pre-cleared lysates were incubated with beads conjugated to either AGO2 antibody or IgG Isotype control (MBL, M076-3), with gentle rotation at 4 °C for 4 h. After washing with miWash buffer, the immunoprecipitated mRNA was extracted using RNeasy Mini Kit (Qiagen, 74104) with integrity assessed with Agilent 2100 Bioanalyzer (Agilent, USA). To further narrow down the scope of target genes and elucidate the miRNA-mediated post-transcriptional regulatory network, algorithm prediction platforms were used to evaluate seed-region complementary affinity and site accessibility, including TargetScan [30] and IntaRNA [31].

EV labeling, uptake, and exogenous loading

ExoSparkler Exosome Labeling Kit (Dojindo, 343–09661) was used to label the membrane structure of EVs and administered to cells at varying concentrations for different time durations. Labeled EVs indicated a clear time- and dose-dependent internalization pattern (Fig. 1E, Fig. S1C) and a preferential homing uptake (Fig. S1H). Exo-Fect™ Exosome Transfection Kit (System Biosciences, EXFT10A-1) was utilized to exogenously load miRNAs into EVs [32]. According to the manufacturer, 1 × 10 ⁶ particles/mL EVs were mixed with fluorescence-tagged miRNAs or miR-181c mimics in the reagent and incubated for 10 min at 37 °C. The miRNA-engineered EVs were then isolated and could be further dyed with a membrane labeling kit or directly added to the medium in order to observe the uptake of intracellular fluorescent signal.

Fig. 1.

Multi-conditioned hMSC-EVs exhibit disparities in osteo-inductivity. A Schematic workflow showing the study design, EV collection, isolation, characterization, and storage conditions. B Flow cytometry analysis of MSC surface markers, including CD73, CD90, CD105, CD19, CD34, CD11b, CD45, and HLA-DR. C Nanoparticle tracking analysis of EV size and distribution. Results of SEC fraction 8–10 were shown. D Western blot examining surface markers of EV, including CD9, CD63, FLOT-1, and Annexin V, as well as contamination markers GP96 and Calnexin. E Confocal images showing uptake efficiency of fluorescent-labeled EVs at different concentrations (5 × 10⁷, 5 × 10⁶ particles/mL). F ARS staining (left) and RT-qPCR analysis of osteogenic marker Integrin Binding Sialoprotein (IBSP) (right) of Osteo_High and Osteo_Low cells, cultured under different passages (P7, P10) and temperatures (37 °C, 25 °C) for day 21(n = 3, one-way ANOVA with Dunnett’s post-hoc). G-H ARS staining (G) and RT-qPCR analysis of osteogenic markers (H) IBSP, Osteopontin (SPP1), Osteocalcin (BGLAP), and Osterix (SP7) in Osteo_High hMSCs treated with EVs obtained under multi-conditions for 14 days (n = 3, one-way ANOVA with Dunnett’s post-hoc). Data are presented as the mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, P > 0.05. Exact P-values and statistical methods are provided in the source data. Scale bars in F, G = 10 μm; H, J = 100 μm

Protein-protein interaction, co-immunoprecipitation, and ubiquitination assay

Protein-protein interaction (PPI) models between RNF150 and the canonical upstream kinases, including MAPK11/13/14, MAP2K3/6, and MAP3K5, were constructed using AlphaFold3 [33] and PRODIGY [34]. Among these kinases, MAP3K5 exhibited the most stable interaction with the predicted binding free energy (∆G) of − 12.6 kcal·mol^{− 1} (Fig. S6B). To experimentally validate it, we established an overexpression system by MAP3K5-FLAG and RNF150-MYC plasmids (Fig. S6C). HEK293T cells (ATCC, CRL-3216) were transfected with the plasmids mentioned above, using FuGene HD (Promega, Madison, WI) at a concentration of 0.05 µg/cm². After 48 h, the cells were lysed by IP buffer (Thermo Fisher Scientific, 87788). Cell lysates were incubated with Anti-DDDDK-tag mAb-Magnetic Beads (MBL, M185-11) or Anti-MYC-tag mAb-Magnetic Beads (MBL, M047-11) at 4 °C for 3 h with slow rotation. To investigate RNF150-mediated ubiquitination, endogenous ubiquitination assays were conducted as follows: Forty-eight hours after siRNA transfection and overexpression, cells were treated with or without MG132 (Fujiwako, 139-18451; final concentration: 10 µM) and hydrogen peroxide solution (3%wt H₂O₂, Sigma-Aldrich, 323381, final concentration: 1 mM) for 9 h to modulate proteolysis of ubiquitinated proteins. Cell lysates were immunoprecipitated and collected for immunoblotting. Oxidative stress was stimulated by H₂O₂ (final concentration: 700 µM) for 1 h, and the reactive oxygen species (ROS) intensity was measured by the Highly Sensitive DCFH-DA Kit (Dojindo, R252).

Immunofluorescence staining

Cells were seeded in confocal dishes. After fixation with 4% paraformaldehyde, blocking was performed using Blocking One Histo (Nacalai, 06349-64). Cells were incubated in Immunoreaction Enhancer (TOYOBO, NKB-501) overnight at 4 °C with corresponding primary antibodies (refer to Western blot section for antibody information). Cells were washed with 0.05% TBST buffer followed by secondary antibodies incubation at a 1:500 dilution at room temperature for 60 min. Cells were mounted with ProLong™ Gold Antifade Mountant (Thermo Fisher Scientific, P36931). Images were captured using a Nikon microscope (AX/AXR with NSPARC, Japan). For intensity quantification, 10 μm z-stack images were acquired and processed into maximum intensity projections using ImageJ (NIH, Bethesda, MD, USA).

Dual-luciferase reporter assay

To verify whether NDRG4 and RNF150 are target genes of miR-181c-5p, each 3′-end-untranslated region (UTR) containing potential binding sites was cloned from total RNA reverse-transcribed cDNA of hMSCs using corresponding primers (Supplemental Table 1). The NDRG4 (full-length) and two RNF150 mRNA fragments (Poorly-conserved site 4396’-4403’ and conserved site 7068’-7079’) were inserted independently into the pmirGLO Dual-Luciferase miRNA Target Expression Vector (Promega, E1330). HEK293T cells were firstly inoculated with NDRG4-pmirGLO, RNF150-pmirGLO, or empty vector for 24 h, and subsequently transfected with 50 nM miR-181c-5p mimic or NC mimic. 48 h post-transfection, cells were lysed for luciferase activity measurement using the Nano-Glo^® Dual-Luciferase^® Reporter Assay System (Promega, N1610) by Varioskan LUX Multimode Microplate Reader (Invitrogen) and normalized to Renilla luciferase activity.

Inhibitors and activators administration

CultureSure DMSO (Fujiwako, 031-24051) was used to dissolve the p38 phosphorylation antagonist SB203580 (KOM, AG-CR1-0030) and the agonist Anisomycin (Fujiwako, 017-16861). SB203580 at a final concentration of 10 µM or Anisomycin at 10 nM was added to the osteogenic induction medium for routine culture. To maintain the drug concentration, the medium was replaced every three days. Equivalent DMSO or D-PBS was applied as a control.

Apoptosis assays

hMSCs were seeded overnight, then transfected with miRNA/siRNA or transduced with lentivirus. Osteogenic induction medium was replaced after two days and maintained for additional 7 days. Live cell images were continuously monitored using an automatic system (CM30, Evident, Japan). Cell viability was evaluated using MTT assay (G4000, Promega, USA) according to the manufacturer’s instructions. Mitochondrial membrane potential was stained using JC-1 MitoMP Detection Kit (Dojindo, Japan) and evaluated by the ratio of red aggregate intensity to green JC-1 monomer fluorescence. DNA fragmentation was detected using TUNEL in situ Apoptosis Detection Kit (Takara Bio, Japan), followed by nuclear counterstaining with Hoechst 33,342 (Invitrogen, USA). The proportions of TUNEL-positive area were quantified in relative to nuclei area. Plasma-membrane integrity was assessed by 7-aminoactinomycin D (7-AAD; BD Pharmingen, USA) and labeled-cell counts were calculated by ImageJ (v1.8.0, NIH, Bethesda, MD, USA).

Animal experiments and allocations

Animal ethical approval was issued by the Animal Care and Use Committees of the Institute of Science Tokyo (A2025-148C4). C57BL/6J WT mice (8-weeks-old) were purchased from CLEA (Tokyo, Japan). Mice were acclimated for 1 week before induction under specific-pathogen-free condition on a 12 h light cycle with free access to water and food. Mice were randomly allocated into experimental or control groups (PBS, EV, NC-AgomiR, miR-181c), with 4 individuals in each group. PBS group served as control for EV group (derived from Osteo_H_P7_37°C, concentration: 2 × 10⁹ Part/mL). miR-Up™ AgomiR (GenePharma, China) of Negative Control and hsa-miR-181c-5p (B06015, B04008) were synthesized and diluted in normal saline (concentration: 50 nM). Anesthesia (medetomidine-midazolam-butorphanol) were injected intraperitoneally according to body weight. 6 − 0 silk suture (Nitcho Industry, Japan) was ligated around the cervical portion of the right maxillary second molar, while the contralateral non-ligated maxillary molar served as an internal control [35]. Suture maintenance was checked every three day. Suture was maintained for 14 days to accumulate plaque and removed 24 h before drug administration. Local injection was achieved with 32-gauge insulin syringes into four different sites (10 µl/site), surrounding the distal, mesial, buccal, and palatal gingiva of the second molar under a stereomicroscope. Routine drug administration (in total 4 times, every two days) and euthanized sacrifice were respectively conducted according to the timeline (Fig. 7A).

Fig. 7.

EVs and miR-181c alleviate periodontitis-induced bone loss in vivo. A Schematic illustration of the experimental workflow. The suture was ligated around the maxillary second molar. After suture removal, PBS, EV, NC-AgomiR, or miR-181c was locally injected every 2 days for a total of four administrations. Samples were later harvested for µCT and histological analyses. B Conserved sequence in hsa/mmu-miR-181c-5p and synthesized AgomiR-181c-5p. C Representative µCT images of maxillary molars: 3D reconstructions and 2D sagittal sections are shown. EV and miR-181c treatments preserved alveolar bone morphology and improved crest height and continuity. D Quantitative analysis of bone loss (CEJ-ABC distance), BV/TV, relative bone volume, and trabecular number (n = 4, one-way ANOVA with Dunnett’s post-hoc test). E Representative H&E staining of second molar. H&E images show that the heavy neutrophil infiltration and alveolar resorption were partially restored by EV and miR-181c. *P < 0.05, **P < 0.01, ***P < 0.001; ns, P > 0.05. Scale bars in C = 1000 μm, E = 300 μm

Micro-computed tomography (µCT) and alveolar bone quantification

To observe and evaluate alveolar bone loss, the maxilla was fixed with 4% paraformaldehyde and analyzed with µCT (SMX-100CT, Shimadzu, Japan). Three-dimensional reconstruction was performed by RadiAnt Dicom Viewer (Medixant, Poland). Distances from the cementoenamel junction to the alveolar bone crest (CEJ-ABC) were measured at six sites of the second molars (mesio-buccal, buccal, disto-buccal, mesio-palatal, palatal, and disto-palatal) in multi-plane reconstruction view of buccal-lingual cross section, set aligned to long axis of the molar [36]. The average bone loss was calculated by subtracting the CEJ–ABC distances between the ligated and the non-ligated sides. Bone indexes were calculated using TRI/3D-BON-FCS (RATOC, JP). The analyzed region of interest consisted of 50 slices from the root bifurcation to the root apex of the maxillary second molar. Bone volume/total volume (BV/TV), relative bone volume (BV-ligated/BV-non-ligated), Bone Mineral Density (BMD), and trabecular number, separation, thickness (Tb.N, Tb.Sp, Tb.Th) were respectively analyzed. Images from different specimens were evaluated in a random and blinded sequence.

Histological analysis

After µCT analyses, samples were decalcified in EDT-X (Falma, Japan) for 2 weeks, embedded in OCT (45833, Sakura, Japan) and cryo-sectioned (15 μm). Sections with pulp chamber and the two roots of the second molar were selected for further analysis. Hematoxylin–eosin (H&E) staining were performed for evaluation of the periodontal tissue destruction.

Statistical analysis

Data were presented as Mean ± SD of at least three independent experiments. For microscopic images, at least five representative zones from three independent wells are analyzed. To address donor heterogeneity and reproducibility, functional phenotypes were validated in hMSC lines derived from four independent donors. Graphs were created and analyzed by Graphpad Prism (GraphPad Software, v9.5.1, Boston, Massachusetts, USA). An unpaired two-tailed Student’s t-test was performed for two groups. For multiple groups’ comparison, Brown-Forsythe and Welch ANOVA tests were performed, followed by Dunnett T3 multiple comparison tests. For comparisons among grouped data, two-way ANOVA with a Tukey post-hoc test was performed. The statistical significance was determined by P-values less than 0.05 and were labeled as follows: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, and non-significant (ns) when P > 0.05. The sample values were plotted as individual dots in each panel and corresponding statistical analysis can be referred to the source data.

Results

Multi-conditioned hMSC-EVs exhibit disparities in osteo-inductivity

To establish EVs as the functionally interpretable unit for downstream mechanistic analysis, we firstly characterized hMSC-derived EVs and assessed how cellular states influence their osteogenic efficacy (Fig. 1A). In accordance with ISCGT2025 and MISEV2023 guidelines [37, 38] (Fig. S1A), hMSC-derived EVs displayed a peak diameter of 70–150 nm and expressed canonical EV surface markers (Fig. 1B–D), with high purity confirmed by particle-to-protein ratios (Fig. S1B). Time- and dose-dependent internalization of labeled EVs was observed in recipient hMSCs, supporting efficient cellular uptake (Fig. 1E, Fig. S1C).

Two hMSC lines exhibiting the most pronounced osteogenic potential difference were designated as Osteo_High and Osteo_Low (Fig. S1I), screened from seven independent donors [25]. To further assess the effects of culture temperature (37 °C vs. 25 °C) and passage numbers (P7 vs. P10) on osteogenesis, ALP assay was performed (Fig. S1J), as well as ARS staining and RT-qPCR of osteogenic markers IBSP, SP7, BGLAP, SPP1 (Fig. 1F and Fig. S1K). In general, Osteo_High cells at younger passages cultured at optimal temperature elicited the strongest efficiency, followed by 25 °C culture, elder passage, and lastly the Osteo_Low cells. In the following step, multi-conditioned hMSC-derived EVs were evaluated for osteo-inductivity respectively. EVs derived from Osteo_Low cells failed to induce obvious hMSC osteogenesis, whereas EVs secreted by Osteo_High markedly enhanced differentiation, with similar tendency of declination by increasing passages and reduced temperature (Fig. 1G–H). Collectively, these observations indicate that EV-mediated osteogenic efficacy is tightly coupled to the intrinsic differentiation capacity and cultural status of parental cells, highlighting the presence of functionally determinant EV cargos.

miR-181c as the determinant of hMSC lineage decision

To identify the molecular determinants responsible for osteogenic effect, microarray analysis was performed on the aforementioned hMSC–derived EVs (Supplemental Table 2). The heatmap and Venn diagram revealed marked differences in miRNA profiles (Fig. 2A–B). Upregulated miRNAs from the Osteo_High_P7_37°C group were subjected to prediction algorithm [30, 39] and intersecting candidates were analyzed by functional annotation. Gene ontology (GO) indicated enrichments in signal transduction, cell differentiation, and phosphorylation (Fig. 2C). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway suggested activation of cascades related to stem cell pluripotency and metabolism, including MAPK, cAMP, etc. (Fig. 2D). Gene set enrichment analysis (GSEA) further suggested in positive regulation of osteoblast differentiation, oxidative stress, and cell proliferation (Fig. 2E and Fig. S2A–C).

Fig. 2.

Profiling and screening of EV-encapsulated cargo. A Heatmap of miRNA expression profiles in EVs isolated from different donors (Osteo_High/Low cell lines) under various culture conditions (Passage 7, 10 and Temperature 37 °C, 25 °C). B Venn diagram of miRNA expression across the samples, with an enlarged box showing the miRNAs stably detected in Osteo_High_P7_37 °C group (n = 2). C GO functional annotation of the upregulated miRNAs. D KEGG analysis of upregulated miRNAs revealed enrichment in signaling pathways regulating stem cell pluripotency. E Gene set enrichment analysis (GSEA) indicated involvements in osteoblast differentiation and cell proliferation. F RT-qPCR analysis of intracellular miRNA expression levels in different cultured samples, normalized to snU6 as the internal control (n = 3, two-way ANOVA with Bonferroni’s post-hoc test). G ALP activity under transfection of candidate miRNA mimics at 50 nM (n = 4, one-way ANOVA with Dunnett’s post-hoc test). H Scanning and microscopic images of ARS staining after transfection with candidate miRNAs (left), with quantification of mineral nodule area (right) (n = 5, one-way ANOVA with Dunnett’s post-hoc test). Data are presented as the mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, P > 0.05. Exact P-values and statistical methods are provided in the source data. Scale bars in J = 100 μm

In search of the miRNAs with optimal osteo-inductive capacity in EVs, we focused on those consistently detected in the Osteo_High_ P7_37°C group, comprising 26 miRNAs (Fig. 2B). Among them, five candidates: miR-181c-5p, miR-378a-3p, miR-339-5p, miR-584-5p, and miR-1323 were supported (Fig. S2F), with miR-181c-5p exhibiting the highest reads (Fig. S2D). Cellular validation confirmed that miR-181c-5p was the most abundantly expressed component (Fig. 2F). For functional validation, synthetic mimics were transfected (Fig. S2E), among which miR-181c-5p exerted the strongest efficiency (Fig. 2G).

A gradual increase in endogenous miR-181c expression was observed from throughout osteogenic induction (Fig. 3A). Gain- and loss-of-functions were performed to bidirectionally confirmed its positive role (Fig. 3B–C and Fig. S3C). Inhibition of miR-181c markedly promoted lipid accumulation, as evidenced by increased lipid droplet formation (Fig. 3B and Fig. S3D). To comprehensively characterize its stage-specificity, multiple markers were assessed by from days 7–28 (Fig. 3D and Fig. S3F): miR-181c enhanced ALPL expression during the early differentiation phase, upregulated SP7 and SPP1 at intermediate, and increased the IBSP, RUNX2, and COL1A1 for maturation. In contrast, treatment with miR-181c inhibitor partially suppressed osteo-markers (Fig. S3E). Together, these findings establish miR-181c as a functional determinant of hMSC lineage commitment rather than a passive component enriched in osteogenic EVs.

Fig. 3.

miR-181c as the determinant of hMSC lineage. A RT-qPCR analysis of endogenous miR-181c expression levels in hMSCs during the osteogenic induction process (day 0-day 28). snU6 was used as an internal control. B–C ARS staining (B, day 21) and Oil Red O staining (C, day 35) of hMSCs with quantification, after transfection with miR-181c mimic or inhibitor. D RT-qPCR analysis of osteogenic marker genes (ALPL, SPP1, SP7, IBSP, BGLAP, RUNX2, COL1A1) at different time points (day 7-day 28), after transfection with miR-181c mimic (n = 3, two-way ANOVA with Bonferroni’s post-hoc). E Illustration of EV bio-engineering. miRNAs were loaded into hMSC EVs. F NTA analysis of the original and engineered EVs demonstrating consistency in size distribution. G Confocal images of cytoplasmic co-localization between EVs (labeled with Exosparkler-Green) and exogenously-loaded miRNA (tagged with Texas-Red). H ALP assay showing exogenous miR-181c loading partially restored the osteogenic ability of Osteo_Low-derived EVs (n = 4, one-way ANOVA with Dunnett’s post-hoc test). Data are presented as the mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, P > 0.05. Exact P-values and statistical methods are provided in the source data. Scale bars in H = 10 μm, J = 100 μm

To determine whether similar potency could be achieved under a natural-mimicking delivery, and to support the therapeutic potential of EV-mediated intercellular transfer, miR-181c was exogenously loaded into EVs using a bioengineering approach (Fig. 3E). NTA analysis confirmed that the physicochemical properties of the engineered EVs remained unchanged (Fig. 3F). Confocal revealed the successful encapsulation and efficient delivery to recipients (Fig. 3G). Furthermore, when miR-181c mimics were loaded into the EVs derived from Osteo_Low cells—which by themselves have low-efficiency osteogenic differentiation—both ARS staining and RT-qPCR demonstrated a pronounced capacity restoration (Fig. S3G-H), even comparable to those Osteo_High-derived EVs (Fig. 3H). Collectively, miR-181c is reinforced as the principal functional cargo which could strengthen osteo-deficient hMSC EVs.

miRISC-mediated targetome reveals RNF150 mRNA repression

To move beyond phenotypic association and define how miR-181c exerts osteogenic regulation, we next sought to identify its direct targets within the miRNA Induced Silencing Complex (miRISC) by AGO2 RNA-immunoprecipitation (RIP) (Fig. 4A). Immunoblotting and Coomassie Blue confirmed successful enrichment of AGO2 (Fig. 4B). Immunoprecipitated mRNAs were extracted and subjected to RIP-seq (Fig. S4A). In parallel, bulk RNA-seq was conducted 48 h after transfection to capture post-transcriptional alterations induced by miRNA-mediated degradation (Supplemental Table 3). Volcano plots displayed that RIP-seq identified 418 differentially expressed genes (DEGs), including 224 upregulated and 194 downregulated targets, whereas RNA-seq detected 575 significantly downregulated genes (Fig. 4C). KEGG of RIP-seq showed enrichment of the spliceosome pathway, revealing the potential relation to post-transcriptional RNA processing (Fig. S4B).

Fig. 4.

miR-181c exerts RNF150 mRNA degradation through miRISC. A Schematic illustration of RNA immunoprecipitation sequencing (RIP-seq) and bulk RNA-seq. B Western blot and Coomassie Blue staining confirming successful enrichment of the AGO2-miRISC complex. C Volcano plots showing the differentially expressed genes (DEGs) of RNA-seq and RIP-seq. D Venn diagram showing the screening strategy of miRNA targetome, intersected by RIP-seq DEGs, RNA-seq downregulated genes and overlapped with TargetScan predictions. E Table summarizing miRNA-mRNA interaction information with scoring indexes predicted by TargetScan and IntaRNA. F ARS staining following siRNA knockdown of candidate miRNA targets, including GNB4, DRAM1, RPSP6KA6, TRIM2, ATP11C, NDRG4, and RNF150. G–H ALP activity (G) (n = 4, unpaired two-tailed t tests) and RT-qPCR of BGLAP (H) increased after siRNF150 and siNDRG4 knockdown (n = 3, unpaired two-tailed t tests). I RT-qPCR validation of RNF150 and NDRG4 downregulated by miR-181c overexpression (n = 3, unpaired two-tailed t tests). J Putative miR-181c binding sites and sequence in RNF150 and NDRG4 mRNA 3’UTR region, adapted from Targetscan. K Dual-luciferase reporter assays for RNF150 and NDRG4 mRNA (n = 4, one-way ANOVA with Dunnett’s post-hoc test). Data are presented as the mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, P > 0.05. Exact P-values and statistical methods are provided in the source data. Scale bars in F = 100 μm, L = 10 μm

In order to refine a comprehensive miR-181c targetome, DEGs from RIP-seq, downregulated genes from RNA-seq, and genes with conserved binding site predictions were intersected, yielding 7 overlapping candidates: RNF150, GNB4, NDRG4, DRAM1, RPS6KA6, ATP11C, and TRIM2 (Fig. 4D), following the binding potential evaluation with TargetScan and IntaRNA analyses [30, 31]. All of the candidate transcripts’ conserved sites in 3’UTR regions showed RNA hybrid free energy ranging from − 20.1 to − 27.4 kcal/mol, indicating the thermodynamically feasible miRNA–mRNA duplexes and structural accessibility (Fig. 4E). To define the dominant functional target in osteogenic relevant context, we further performed siRNA-mediated knockdown (Fig. 4F). siGNB4, siDRAM1, and siRPS6KA6 resulted in significant suppression in mineral deposition, which were in contradiction to miR-181c phenotype. Although siTRIM2 and siATP11C slightly increased ARS staining, their impacts on regulating ALP activity were relatively weak and did not fully reproduce the miR-181c phenotype, and thus RNF150 and NDRG4 were prioritized for subsequent mechanistic validation (Fig. S4C, Fig. 4G–H). Both candidates’ mRNA expression levels were markedly reduced following miR-181c transfection, suggesting their potential involvements (Fig. 4I). Hence, two putative binding sites within the RNF150 mRNA 3′UTR (conserved and poorly-conserved) as well as the full-length NDRG4 mRNA 3′UTR were cloned for dual-luciferase reporter assays (Fig. 4J), which demonstrated that co-transfection with miR-181c significantly reduced luciferase activity only for the construct of RNF150 (Fig. 4K), indicating it as the most credible target of miR-181c. We then performed Western blotting to examine intracellular RNF150 expression following miR-181c interventions (Fig. S4I). miR-181c mimic significantly reduced RNF150 protein abundance to approximately 40%. The inhibitor produced only slight upward trend in expression, which is plausible as the inhibition of a single endogenous miRNA may not cause substantial target variation when its basal expression does not reach functional threshold [40, 41]. In conclusion, among the candidate targets, RNF150 satisfied both phenotypic criteria and direct molecular-verifiable binding, positioning it as a core mediator of miR-181c–driven osteogenic regulation.

To directly test whether RNF150 acts as a functional downstream effector of miR-181c, we conducted reciprocal rescue experiments by combining miR-181c gain/loss-of-function with RNF150 overexpression or knockdown (Fig. S4G-H). miR-181c mimic promoted osteogenic differentiation, but co-overexpression of RNF150 by lentivirus partially reversed matrix mineralization and osteogenic markers. Conversely, miR-181c inhibitor weakened differentiation, whereas siRNF150 alleviated its inhibitory effect. Furthermore, we examined whether miR-181c-mediated silencing depends on an intact miRISC-associated repression machinery. Given that the processing bodies are cytoplasmic granules composed of translationally repressed mRNAs and decay-associated ribonucleoproteins, and are scaffolded in part by the bridging protein TNRC6A [42, 43], miR-181c was transfected after TNRC6A knockdown. Immunofluorescence revealed a marked reduction in the co-localization between AGO2 and miRNA signals (Fig. S4E). Consistently, the loss of TNRC6A attenuated miR-181c-mediated RNF150 silencing and partially restored its expression levels (Fig. 4F). Overall, these results support a model in which miR-181c exerts its post-transcriptional silencing effect through direct pairing with RNF150 mRNA in an AGO2-TNRC6A-dependent miRISC complex.

Osteogenesis governed by MAP3K5 ubiquitination and p38 activation

Having established RNF150 as a direct target of miR-181c, we next asked whether this regulatory event is coupled to kinase-driven osteogenic signaling. To elucidate the pathway through which miR-181c-mediated RNF150 silencing promotes osteogenic differentiation, KEGG analysis was performed based on the upregulated genes from the miR-181c-overexpressed RNA-seq. Significant enrichments were found in RAP1, APELIN, and MAPK signaling, along with the pathways associated with cell cycle and stem cell pluripotency (Fig. 5A). Consistent with the GO analysis, which highlighted the phosphatase/protein kinase activities (Fig. S5A–D), kinase enrichment analysis also identified multiple members of the MAPK family, particularly with high scores for MAPK14 (p38α) (Fig. 5B).

Fig. 5.

EV/miR-181c/RNF150 activates the p38-MAPK pathway. A KEGG analysis of the upregulated genes from miR-181c overexpression RNA-seq, highlighting MAPK pathway. B Kinase enrichment analysis of miR-181c overexpression RNA-seq, indicating activation of MAPK14 (p38α), MAPK1, and MAPK8. C Western blot of total p38 and phosphorylated-p38 (p-p38) at different time frames (10 min, 1 h, 4 h) following EV treatment, miR-181c mimic transfection, and siRNF150 knockdown. α-Tubulin as an internal control. D Immunofluorescence staining images of p38 (green) and p-p38 (red). E–G Effects of p38-specific inhibitor SB203580 on ARS staining in hMSCs treated with EVs (E), miR-181c mimic (F), and siRNF150 (G). H Effects of p38-specific inhibitor SB203580 on ALP activity in hMSCs treated with EVs, miR-181c mimic, and siRNF150 (n = 4, one-way ANOVA with Dunnett’s post-hoc test). Data are presented as the mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, P > 0.05. Exact P-values and statistical methods are provided in the source data. Scale bars in G, J, M = 100 μm, D = 50 μm

We thus hypothesized that EVs with miR-181c cargo promoted osteogenic differentiation of hMSCs via the p38-mediated MAPK pathway. Phosphorylation levels were examined at the Thr180/Tyr182 (common residues shared by p38 isoforms). Notably, each treatment—EVs, transfection of miR-181c mimic or siRNF150—triggered a marked increase in phospho-p38 (p-p38) within 10–60 min, followed by a subsequent decline over 4 h (Fig. 5C and Fig. S5E). Immunofluorescence staining similarly confirmed a visible enhancement in p-p38 intensity (Fig. 5D and Fig. S5F). To determine whether p38 served as an essential mediator of the EV/miR-181c/RNF150 axis, selective inhibitor SB203580 was administered. Strikingly, SB203580 abolished the ARS staining (Fig. 5E, F, and G; Fig. S5G) and ALP activity (Fig. 5H) induced by the aforementioned stimulations. RT-qPCR also revealed a reduction in the osteogenic markers following inhibitor treatment (Fig. S5H, I, and J). Protein–protein interaction modeling nominated MAP3K5 as the most stable RNF150 interactor among canonical upstream kinases in the p38 MAPK pathway (Fig. S6B). This predicted interaction was supported by structural interface features (Fig. 6A) and was subsequently validated experimentally. In HEK293T cells co-expressing MAP3K5–FLAG and RNF150–MYC, reciprocal co-immunoprecipitation confirmed direct binding between RNF150 and MAP3K5 (Fig. 6B; Fig. S6C).

Fig. 6.

RNF150 regulates the ubiquitination of MAP3K5. A 3D structural model of RNF150 (brown)-MAP3K5 (blue) with interface details. B Co-IP and Western blot analyses validating the direct interaction between MAP3K5-FLAG and RNF150-MYC. C Western blot analysis of MAP3K5 ubiquitination following H₂O₂ stimulation and siRNF150 treatment. D DCFH-DA staining images of ROS intensity, demonstrating the attenuation of H₂O₂-induced oxidative stress by EVs. E–F ARS staining in hMSCs overexpressing MAP3K5 (E) and RNF150 (F), with or without the p38 inhibitor SB203580. (n = 3, one-way ANOVA with Dunnett’s post-hoc test). G–H RT-qPCR analysis in hMSCs overexpressing MAP3K5 (G) and RNF150 (H), with or without the p38 agonist anisomycin (n = 3, one-way ANOVA with Dunnett’s post-hoc test). Data are presented as the mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, P > 0.05. Exact P-values and statistical methods are provided in the source data. Scale bars in G, I = 100 μm

Pathway enrichment analysis further implicated ubiquitin-mediated proteolysis among miR-181c targets (Fig. S6D), consistent with the known E3 ubiquitin ligase activity of RNF150. Given that MAP3K5 undergoes ubiquitination in response to oxidative stress, we examined whether RNF150 regulates MAP3K5 stability. Under proteasome inhibition, oxidative stress induced robust ubiquitination of MAP3K5, which was markedly attenuated by RNF150 knockdown (Fig. 6C). Notably, hMSC-derived EVs partially mitigated H₂O₂-induced oxidative stress (Fig. 6D), suggesting that EV-delivered miR-181c may influence MAP3K5 signaling in stress-responsive contexts. Since the MAP3K5 was initially characterized as apoptosis signal-regulating kinase 1 (ASK1) [44, 45], we sought to examine whether MAP3K5-mediated osteogenesis was accompanied with cell death or apoptosis burden. Transcriptomic analysis, live imaging, and viability assay revealed no significant cell death phenomena. JC-1, TUNEL, and 7-AAD staining also did not observe extensive apoptotic hMSCs (Fig. S6E–F, Supplemental Table 4).

Finally, to assess the functional roles of MAP3K5 and RNF150 in the p38 signaling cascade, rescue experiments were conducted. Lentiviral overexpression of MAP3K5 significantly promoted osteogenesis, whereas this effect was abolished by the p38 inhibitor (Fig. 6E&G). Conversely, RNF150 overexpression impaired osteogenic differentiation, which was reversed by co-treatment with the p38 agonist anisomycin (Fig. 6F&H). To further verify whether the established osteogenic effects were reproducible across different individual backgrounds, multi-donor validations of ARS staining were performed, showing coherent and robust trends (Fig. S7A). Taken together, these findings establish a causal signaling axis in which EV-delivered miR-181c suppresses RNF150, attenuates MAP3K5 ubiquitin–proteasome degradation, and thereby promotes p38 activation to drive osteogenic differentiation.

EVs and miR-181c alleviate ligature-induced periodontitis bone loss in vivo

For translational purposes, we continued to validate EVs and miR-181c as potential therapeutic candidates for periodontal regeneration in vivo using a ligature-induced periodontitis mouse model (Fig. 7A). The mature miR-181c-5p sequences is identical between human and mouse, suggesting functional conservation across species and the feasibility of animal experiments (Fig. 7B). µCT reconstruction images showed that 2 weeks of ligature placement successfully induced pronounced alveolar bone resorption, likely through sustained plaque accumulation and local inflammation (Fig. 7C). After model establishment, mice received repeated local injections of PBS, EV, NC-AgomiR, or miR-181c (Fig. 7A). The PBS- and NC-AgomiR treatment continued to show collapse of the alveolar crest, irregular erosion surfaces, disrupted bone continuity, and blurred bone margins (Fig. 7C). In contrast, EV and miR-181c treatments attenuated the increase in CEJ-ABC distances, restoring the contour with smoother boundaries. In particular, furcation exposure was suppressed, and sagittal sections further revealed partial recovery of interdental bone height (Fig. 7C). Quantitative analysis demonstrated the significant increase in BV/TV and relative bone volume (Fig. 7D), although the trend of trabecular number, separation, and thickness did not reach statistical significance (Fig. S7B).

Histological staining similarly supported the radiological findings. H&E sections showed that the Ligation, PBS, and NC-AgomiR groups displayed marked tissue disorganization, including inflammatory infiltration in connective tissue and alveolar crest resorption, accompanied by sparse and fragmented trabecular structures. After EV or miR-181c treatment, the alveolar and furcation involvements were partially restored, indicating a shift toward a less destructive bone-remodeling state (Fig. 7E). Collectively, consistent with the in vitro findings described above, local administration of EVs alleviated plaque-induced alveolar bone destruction and improved the structural integrity. The comparable protective potency achieved after miR-181c delivery further indicated it as a key functional cargo mediating the bone-regenerative effects of EVs.

Discussion

Among oral diseases, periodontal alveolar bone defect remains one of the most challenging aspects for contemporary regenerative approaches. Maximizing the osteo-inductive potential of hMSCs is therefore of particular interest, as accumulating evidence indicates that hMSC-derived EVs exert sustained regenerative effects within the periodontitis microenvironment [46]. However, donor heterogeneity, batch-to-batch variation, and differences in culture conditions substantially influence EV composition and potency [47], underscoring the need to optimize both parental cell selection and culture parameters for bone regeneration. In our preliminary work, hMSC lines established from independent donors exhibited marked differences in intrinsic proliferative and osteogenic capacity [25]. Consistent with this, EVs secreted by Osteo_High hMSCs at younger passage under 37 °C displayed the strongest osteo-inductive efficacy. Previous studies have reported that EV secretion dynamics and proteomic contents vary with time and temperature [48, 49]. While various approaches, including genetic modification, mechanical stimulation, or chemical supplementation, have been explored to enhance EV bioactivity, our findings demonstrate that temperature control combined with intrinsic pre-screening is sufficient to selectively enrich EVs with the pro-regenerative cargos. These results highlight the importance of defining EV preparations in which specific molecular cargos, rather than global bioactivity, account for reproducible osteo-inductive outcomes.

Because a single miRNA can theoretically regulate hundreds of mRNA targets, miR-181c has been implicated in diverse biological contexts, including inflammation [50], tumorigenesis [51], mitochondrial transfer [52], and metabolism [53], with numerous targets validated in vitro. In the maxillofacial context, miR-181c has been shown to play an essential role in skeletal development and bone homeostasis, as mice lacking multiple miR-181c alleles exhibit reduced body size [54]. miR-181c also regulates human jawbone-derived osteoblast differentiation via the Notch2/Runx2 axis [55], suppresses osteoclastogenesis through SFRP1/Wnt/β-catenin signalling [56], and mitigates perimenopausal osteoporosis via the WTAP/YTHDC1/SFRP1 pathway [57]. Despite these functional insights, direct evidence for condition- and tissue-specific target engagement by miR-181c has remained limited. Among six mature members of the miR-181 family, only a single AGO2–chimeric eCLIP analysis in miR-181a/b-1 knockout murine thymus has systematically interrogated target engagement [58], and comparable analyses in hMSCs have been lacking. To address this gap, we combined AGO2-centered RNA immunoprecipitation sequencing with transcriptome profiling to capture miRISC-bound mRNAs and their corresponding degradation signatures. This integrative strategy defined a miRISC-associated targetome of miR-181c in human primary MSCs and led to the algorithmic analysis and experimental validation of seven direct targets. Within this set, TRIM2 has been linked to miR-181c in ischemic-hypoperfusion brain injury [59], whereas ATP11C has only been proposed computationally [60]. RPS6KA6, RNF150, GNB4, NDRG4, and DRAM1 represent a set of previously-unrecognized miR-181c targets. Notably, among the potential candidates, RNF150 was prioritized for mechanistic investigation, as it consistently matched with miR-181c-mediated downstream regulation in both molecular verification and phenotypic readouts. Such finding underscores the limitations of in silico prediction and emphasize the need for context-resolved, experimentally-validated approaches to define authentic miRNA regulatory networks.

Bioinformatic analyses and in vitro experiments revealed that hMSC-derived EVs and miR-181c converge on p38 activation, which is required for osteo-inductive effects. MAPKs are highly conserved serine/threonine kinases in eukaryotes that function through canonical three-tiered signaling cascades, among which p38 plays a critical role in osteoblastogenesis [61]. Conditional deletion of p38 results in impaired osteoblast differentiation, reduced mineral density, and decreased bone volume [62]. Mechanistically, nuclear p38 enhances osteogenic gene expression through activation of transcription factors such as Runx2 and Osx [63, 64], consistent with our observations in hMSCs and underscoring its relevance to alveolar bone remodeling [65]. Among all miR-181c targets identified in this study, RNF150 was the central mediator of miR-181c-induced osteogenesis. We therefore focused on the interaction between RNF150 and MAP3K5, a key upstream kinase in the p38 MAPK cascade. RNF150, an E3 ubiquitin ligase [66], produced two distinct protein isoforms upon overexpression, of which only the ~ 55 kDa form was co-immunoprecipitated with MAP3K5, suggesting isoform-specific ubiquitin ligase activity. This observation aligns with previous reports that MAP3K5 interacts with multiple RING-domain–containing RNF family ligases, and that the knockdown of these RNFs reduces MAP3K5 ubiquitin-dependent proteolysis [67]. In this context, our data identify RNF150 as a negative regulator of MAP3K5 stability in hMSCs.

Meanwhile, MAP3K5, also known as apoptosis signal-regulating kinase 1 (ASK1), has been primarily associated with stress-induced cell death, inducing apoptosis in osteoblastic cells under mechanical stress [68], ferroptosis in diabetic osteoblasts [69], and impaired chondrocyte survival and cartilage formation [70]. However, accumulating evidence also established its versatile biological outputs to be multi-faceted and context-dependent [45], not only restricted to cell death, but also involved in p38-mediated differentiation and proliferation. For example, Takeda et al. demonstrated that it can promote neuronal stem cell differentiation and survival [71, 72]. The MAP3K5-p38 cascade has been proposed as a positive regulator of keratinocyte, myogenic, and adipocyte differentiation, respectively [73–75]. Stress-induced MAP3K5 activation can initiate regenerative stem cell proliferation rather than apoptosis in Drosophila [76, 77]. In conjunction with the negative apoptotic indicators provided in our study, we consider that in the specific context of hMSCs, MAP3K5 accumulation could be primarily interpreted as a differentiation signal inducer promoting osteogenesis through transient activation of p-p38. Furthermore, given that MAP3K5 acts as a cellular stress sensor [78] and that periodontitis lesions are characterized by elevated oxidative stress [79], it will be of interest to determine whether EV-mediated miR-181c delivery additionally modulates oxidative stress responses in regenerative settings. As hMSC-delivered EVs partially attenuated H₂O₂-induced ROS accumulation in our system, further studies will be required to clarify the contribution of this pathway to stress adaptation.

Using a mouse model of ligature-induced periodontitis, we validated the therapeutic potential of hMSC-derived EVs and miR-181c in alleviating alveolar bone destruction. Local administration reduced bone resorption by approximately 25% (Fig. 7). Previous in vivo studies have reported a role for miR-181c functions in maintaining bone mass in hormone-related osteoporosis, mainly focused on investigating its upstream modulators [80, 81]. Beyond supporting the existing evidence of miR-181c in suppressing bone loss, our plaque-driven model further extends its capability to an inflammatory pathological microenvironment, where excessive release of pro-inflammatory factors drives the progression of periodontitis [82]. From the in vitro experiments, we clarified that the miR-181c/RNF150/MAP3K5 axis promotes hMSC osteogenic differentiation. Although crest height and relative volume were improved, parameters such as bone density or trabecular number did not show significant recovery. This may be attributable to the short observation window, during which the lesion still underwent early-stage remodeling [83]. In addition, as the therapeutic agents were locally administered as liquid formulations, repeated injections were required to maintain an effective concentration. Frequent punctures may cause additional injury and variability. Therefore, future studies integrating controlled-release scaffold systems such as hydrogels may prolong retention and facilitate the evaluation of long-term efficacy in clinically applicable settings [84].

Beyond defining the proof-of-concept mechanism, we further considered the translational potential of EV-based therapeutic strategy. Given the intrinsically low copy number of miRNAs within individual EVs [85], we applied post-packaging strategy to enhance miR-181c loading into hMSC-derived EVs without compromising vesicular integrity. While the effects of key mediators have been tested among multiple donors, to encompass the biological diversity of the broader patient population, further studies involving more diverse clinical backgrounds will be required for translational generalizability. Overall, our findings demonstrate the feasibility of EV–miR-181c axis in protecting against inflammatory alveolar destruction and facilitating bone regeneration.

Conclusion

In summary, we identify miR-181c as a pivotal functional cargo within hMSC-derived EVs that mechanistically links EV delivery to osteogenic differentiation. By integrating AGO2-centered RNA-immunoprecipitation sequencing with functional validation, we demonstrate that miR-181c represses RNF150 within the miRISC complex, thereby attenuating ubiquitin-mediated degradation of MAP3K5 and enabling activation of p38 MAPK signaling. This EV/miR-181c/RNF150/MAP3K5/p38 signaling axis defines a causal framework through which miRNAs can couple post-transcriptional repression to kinase-driven lineage commitment. Together, these findings highlight an EV-based, cargo-level approach for rewiring osteogenic signaling pathways, with direct evidence for periodontal bone regeneration and broader implications for acellular regenerative therapies targeting skeletal tissues.

Abbreviations

hMSCs

Human Mesenchymal Stromal Cells

EVs

Extracellular Vesicles

miRNA/miR

MicroRNA

miRISC

MiRNA-induced Silencing Complex

AGO2

Argonaute2

RIP

RNA Immunoprecipitation

RNF150

Ring Finger Protein 150

MAP3K5

Mitogen-Activated Protein Kinase Kinase Kinase 5 (Apoptosis Signal-Regulating Kinase 1)

Authors' contributions

JY.G and AH.L participated in study design, investigation, collected and analyzed data, methodology development, and drafted the manuscript; Y.Y contributed to pilot experiments and reviewed the manuscript; M.H and S.S contributed to study concept, technical design, and reviewed the manuscript. I.T was responsible for supervision, study concept, funding source, and final revision of the manuscript. All authors have read and approved the final manuscript.

Funding

This work was supported by JSPS KAKENHI Grants-in-Aid for Scientific Research B (22H03273 and 23K24531) and Cell Therapy Technology Inc, Tokyo, JP. The funders did not influence the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

The microarray and RIP/RNA sequencing data have been deposited in series GSE311799, GSE312257 as well as supplemental data files. Raw data and detailed protocols that support this paper are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All experiments involving humans were conducted according to the ethical policies and procedures approved by the Dental Research Ethics Committee of the Institute of Science Tokyo (D2020-077). Animal ethical approval was issued by the Animal Care and Use Committees of the Institute of Science Tokyo (A2025-148C4).

Consent for publication

All authors read and approved the manuscript.

Competing interest

The authors declare no competing interests.