Exosomal miR-484 from lung adenocarcinoma promotes osteoclastogenesis and osteolytic bone metastasis by targeting PECAM1

Abstract

Bone metastasis is a frequent and devastating complication in lung adenocarcinoma (LUAD), where excessive osteoclast activation drives osteolytic destruction and skeletal-related events. How tumor-derived exosomes reprogram osteoclast precursors to establish a bone-metastatic niche remains incompletely understood. Here, we identify an exosomal miR-484–platelet endothelial cell adhesion molecule-1 (PECAM1) axis that links LUAD to pathological osteoclastogenesis and osteolytic bone metastasis. Using a highly bone-tropic Lewis lung carcinoma (BM-LLC) model, we show that BM-LLC-derived exosomes are preferentially internalized by osteoclast precursors and are markedly enriched in miR-484. Exosomal transfer of miR-484 increases intracellular miR-484 in these precursor cells and directly suppresses PECAM1, a negative regulator of osteoclastogenesis. PECAM1 repression activates NFATc1 and c-Fos, up-regulates osteoclast markers (TRAP, CTSK, RANK), and promotes osteoclast differentiation and bone matrix degradation in vitro. Gain- and loss-of-function studies demonstrate that PECAM1 overexpression or miR-484 inhibition attenuates BM-LLC exosome–induced osteoclastogenesis, whereas PECAM1 knockdown phenocopies miR-484 and further amplifies osteoclast activation. In vivo, systemic administration of BM-LLC exosomes exacerbates osteolytic lesions and bone tumor burden without altering lung colonization, while therapeutic delivery of antagomiR-484 partly restores PECAM1 expression, suppresses osteoclastogenic signaling, and alleviates bone loss. Circulating exosomal miR-484 levels correlate with the severity of osteolysis, supporting its potential as a liquid biopsy biomarker. Collectively, our data define exosomal miR-484–mediated PECAM1 suppression as a critical mechanism by which LUAD cells remodel the bone microenvironment, and highlight the miR-484–PECAM1axis as a tractable target for preventing or treating osteolytic bone metastasis.

Graphical Abstract

Introduction

Lung adenocarcinoma (LUAD), the most prevalent subtype of non-small cell lung cancer (NSCLC), accounts for approximately 40% of all lung cancer cases [1]. Although targeted therapies and immunotherapy have advanced significantly, tumor metastasis remains LUAD patients’ leading cause of mortality [2]. Bone metastasis occurs in 30 ~ 40% of LUAD cases, predominantly as osteolytic lesions. These lesions cause severe bone pain, pathological fractures, and hypercalcemia, drastically diminishing patients’ quality of life and survival [3]. Current clinical management relies on bisphosphonates (e.g., zoledronic acid) or RANKL inhibitors (e.g., denosumab), but these agents only delay bone destruction without effectively inhibiting tumor colonization [4]. Therefore, elucidating the molecular mechanisms of LUAD bone metastasis and identifying early diagnostic biomarkers are of paramount clinical significance.

The tumor microenvironment (TME) plays a critical role in cancer metastasis, with increasing evidence suggesting that tumor-secreted exosomes “pre-program” distant organ microenvironments to form a pre-metastatic niche (PMN) [5]. Exosomes (30–150 nm extracellular vesicles) carry bioactive molecules, including miRNAs, that modulate recipient cell behavior [6]. In bone metastasis, tumor-derived exosomes promote osteoclast (OC) differentiation while suppressing osteoblast activity, driving osteolytic destruction [7]. For instance, breast cancer exosomal miR-21 enhances OC activation by targeting PDCD4 [8], while prostate cancer exosomes deliver miR-141-3p to facilitate bone metastasis via the DLC1/MAPK pathway [9]. However, the mechanisms by which LUAD exosomes regulate OC differentiation remain incompletely understood.

Emerging studies highlight the role of platelet endothelial cell adhesion molecule-1 (PECAM1/CD31) in osteoclastogenesis. PECAM1, expressed in OC precursors, regulates OC differentiation through mechanisms of Src signaling [10–12]. Notably, PECAM1 deficiency in bone marrow monocytes leads to increased OC formation and bone loss [12]. In addition, PECAM1 interacts with SHP-1 to modulate monocyte-driven osteoclastogenesis [11]. These findings suggest that PECAM1 may serve as a key checkpoint in bone metastasis. Although PECAM1 has been identified as a prognostic biomarker in lung adenocarcinoma [13] and its expression signature is associated with early-stage LUAD [14], its functional link to LUAD-derived exosomes and osteolytic metastasis remains unexplored. Targeting pathogenic miRNAs with antagomiRs has emerged as a promising therapeutic strategy [15], warranting further exploration of its potential in bone metastasis-targeted therapy.

In this study, we established a bone-metastatic Lewis lung carcinoma (BM-LLC) model and employed multiple approaches to investigate the exosome-mediated mechanisms of LUAD (lung adenocarcinoma) bone metastasis. We identified miR-484 as a key exosomal miRNA that promotes OC differentiation by targeting PECAM1, leading to enhanced bone resorption. Through functional assays [Tartrate-resistant acid phosphatase (TRAP) staining, luciferase reporter assays, and Micro-computed tomography (micro-CT) analysis], we systematically demonstrated the exosomal miR-484/PECAM1 axis’s role as a critical osteolytic metastasis driver. Our findings offer new LUAD bone metastasis insights and highlight potential diagnostic/therapeutic targets.

Materials and methods

Animal models

Female C57BL/6 mice (6–8 weeks old) were purchased from Shanghai Lingchang Biotechnology Co., Ltd. (Shanghai, China; license number SCXK 2023-0006). After one week of acclimation, mice were randomly assigned to all experimental groups using a computer-generated random number sequence to ensure unbiased distribution. All mice within a single cage received the same treatment. A highly bone-metastatic LLC (BM-LLC) model via tail artery injection of LLC cells (2 × 10⁵ cells/100 µL PBS) [16]. To generate the BM-LLC subline, tumor cells isolated from bone metastatic lesions were expanded and subjected to three rounds of in vivo selection (3 groups per round, n = 6 mice/group) [17]. For the exosome functional study depicted in Fig. 6, a total of n = 18 mice were randomly divided into three groups (n = 6 per group): NC-LLC-Exos, BM-LLC-Exos, and BM-LLC-Exos + antagomiR-484 (RiboBio #miR30003127-4-5). These mice were pretreated via tail vein injection with the respective exosomes and/or antagomiR-484. Specifically, antagomiR-484 was administered at a dose of 5 mg/kg by tail vein injection, using an exosome-loaded formulation (electroporated into 2.5 × 10¹⁰ particles of BM-LLC-Exos, equivalent to ~ 10 µg exosomal protein, based on the average PPR). The treatment schedule consisted of four doses: pretreatment on days − 3 and − 1 relative to tumor inoculation (day 0), followed by post-inoculation doses on days 3 and 6. Mice were monitored daily for clinical signs of toxicity, such as lethargy, piloerection, or abnormal posture. No overt signs of toxicity, including significant body weight loss, were observed in any group. Outcome assessment, including quantification of bone metastasis area, micro-CT analysis, and histological evaluation, was performed by investigators blinded to the group allocation. The personnel administering treatments were not involved in the data analysis to prevent bias. For intracaudal artery or intravenous injection, mice were anesthetized with 2% isoflurane delivered in oxygen via a precision vaporizer. At the experimental endpoint, the mice were euthanized by CO₂ inhalation in a induction chamber, followed by cervical dislocation to ensure death. All procedures were approved by the Institutional Animal Care and Use Committee (GK-2025-XM-0179). The protocol required the immediate euthanasia of any mouse exhibiting signs of severe distress, significant weight loss (> 20% of initial body weight), paralysis, or a tumor burden exceeding 1000 mm³. For the bone metastasis model, the primary endpoint was defined as the development of osteolytic lesions accompanied by morbidity, such as severe limping or reduced mobility. No animal was allowed to exceed these predefined humane endpoints. No mice were excluded from the analyses, as all subjects completed the study without reaching the predefined humane endpoints prior to the planned experimental endpoint.

Fig. 6.

In vivo targeting of miR-484–PECAM1 axis attenuates bone metastasis. A Schematic of the experimental timeline showing exosome pretreatment, tumor cell inoculation, and antagomir-484 treatment schedule. B, C Representative images (B) and lung metastasis nodule counts (C) (n = 6 mice per group; scale bar = 0.5 cm). D, E Representative images (D) and bone metastasis area quantification (E) (n = 6 mice per group). F Body weight changes of mice from day − 3 to day 21 (n = 6 mice per group). G, H Representative micro-CT images of tibiae. I, J Cortical thickness (I) and quantitative analysis of bone volume/total volume (BV/TV; J) (n = 6 mice per group). K Representative Western blot analysis of PECAM1, NFATc1, c-Fos, and OC markers (TRAP, CTSK, RANK) in bone marrow cells. L–Q Quantification of the Western blot signals shown in (K) for PECAM1 (L), TRAP (M), CTSK (N), RANK (O), NFATc1 (P), and c-Fos (Q) (n = 6 independent experiments). R Relative expression level of exosomal miR-484 in mouse serum. All data represent mean ± SD, ns P > 0.05, *P < 0.05, **P < 0.01 by one-way ANOVA with Tukey’s test

Exosome isolation and characterization

Exosomes were isolated from cell culture supernatants of cells cultured with exosome-depleted Fetal bovine serum (FBS) (prepared by ultracentrifugation at 110,000 × g for 18 h) by differential centrifugation: 300 × g (10 min), 2,000 × g (20 min), and 10,000 × g (30 min) to remove debris, followed by ultracentrifugation at 110,000 × g (70 min, 4 °C; Beckman Coulter Optima XE-90). Pellets were resuspended in PBS and characterized by: Morphology: Transmission electron microscopy (TEM; Hitachi HT7800) with negative staining; Size distribution: Nanoparticle tracking analysis (NTA; NanoSight NS300, Malvern); Marker detection: Western blot for exosomal markers: TSG101 (ab125011, Abcam), CD63 (ab134045, Abcam), CD81 (ab79559, Abcam) and negative control calnexin (ab22595, Abcam). The particle concentration and protein concentration (measured by BCA assay) were determined for each preparation. The particle-to-protein ratio (PPR) was consistent across batches, averaging approximately 2.5 × 10⁹ particles/µg of exosomal protein. Particle concentration was used as the primary metric for functional in vitro and in vivo dosing. To confirm the intravesicular localization of miR-484, an RNase protection assay was performed. Aliquots of exosomes were treated with RNase A (1 µg/mL, 37 °C, 30 min) either in the presence or absence of 1% Triton X-100. RNA was subsequently extracted and miR-484 levels were quantified by Quantitative polymerase chain reaction (qPCR). Degradation of miR-484 only upon Triton X-100 disruption confirmed its protection within the vesicular membrane. For functional studies, antagomiR-484 was loaded into exosomes (5 µg antagomiR per 2.5 × 10¹⁰ particles in 100 µL electroporation buffer) using electroporation (Bio-Rad Gene Pulser Xcell). For in vivo injections, a dose equivalent to 2.5 × 10¹⁰ particles per mouse was administered.

The identity and purity of the isolated vesicles were confirmed as exosomes according to MISEV guidelines, based on their isolation via differentialultracentrifugation, cup-shaped morphology observed by TEM, a size distribution profile (peak of 100–150 nm) as determined by NTA, positive enrichment of canonical exosomal protein markers (TSG101, CD63, CD81), absence of the negative cellular protein marker calnexin, and the consistent use of exosome-depleted FBS to ensure that the isolated vesicles were of cellular origin and not artifacts from the culture media.

Cell culture and functional assays

NC-LLC, BM-LLC, and intermediate-selection LLCs were maintained in DMEM supplemented with 10% FBS (Gibco, #11965092) for exosome production and functional studies, while RAW264.7 cells (a murine osteoclast precursor cell line; Zhejiang Ruyao Biotechnology Co., #RY-M0109) were cultured under the same conditions and differentiated into OCs using RANKL (50 ng/mL; R&D Systems, #462-TEC-010) and Macrophage colony-stimulating factor (M-CSF; 25 ng/mL; R&D Systems, #416-ML-050) as positive controls (OC-induced). Metastatic potential was characterized through collagen I-coated [18] Transwell invasion assays (Corning, #354483; 8-µm pores). Cells (5 × 104/well) in serum-free DMEM were allowed to invade toward 10% FBS for 24 h. After fixation (4% PFA) and crystal violet staining (0.1%), quantification was performed by analyzing 5 random fields/insert using ImageJ. For osteoclastogenesis assays, RAW264.7 cells were treated with approximately 6.25 × 10⁹, 1.25 × 10¹⁰, or 2.5 × 10¹⁰ particles/mL (equivalent to ~ 2.5, ~ 5, and ~ 10 µg protein/mL, respectively, based on the average PPR) of NC-LLC-Exos or BM-LLC-Exos for 5 days (medium refreshed every 48 h). TRAP-positive multinucleated (≥ 3 nuclei) cells quantified microscopically (Olympus IX73; 5 fields/group; ImageJ) and analyzed for size distribution (ImageJ). Nuclear and cellular size differences between treatment groups were assessed via H&E staining (fixed cells) and phase-contrast microscopy (live cells) to evaluate differentiation status.

Dual-luciferase reporter assay

Wild-type or mutant PECAM1 3’UTR sequences were cloned into the pmirGLO vector (Promega, #E1330) and co-transfected with miR-484 mimic (RiboBio #miR10003127-1-5) into HEK293T cells (purchased from Zhejiang Ruyao Biotechnology Co., Ltd., Ningbo, China; #RY-H0076). Luciferase activity was measured after 48 h (Dual-Luciferase Reporter Assay System, Promega, #E1910).

Quantitative PCR (qPCR) analysis

Total RNA was extracted using Trizol reagent (Invitrogen, #15596026) and reverse transcribed into cDNA with the PrimeScript RT Reagent Kit (Takara, #RR037A) under two conditions: 37 °C for 15 min and 85 °C for 5 s. Quantitative PCR was performed using TB Green Premix Ex Taq II (Takara, #RR820A) on a QuantStudio 6 system (Applied Biosystems). Cycling parameters included initial denaturation at 95 °C for 30 s, followed by 40 cycles of 95 °C for 5 s and 60 °C for 30 s. Melting curve analysis was conducted from 65 °C to 95 °C to verify amplification specificity. The following primers were used: ITGAM (forward: 5′-AAG CAG CTG AAT GGG AGG AC-3′, reverse: 5′- TAG ATG CGA TGG TGT CGA GC-3′), PECAM1 (forward: 5′-CAA GGC CAA ACA GAA ACC CG-3′, reverse: 5′- TCG ACC TTC CGG ATC TCA CT-3′), ITGA2 (forward: 5′-TTT CCT ACC GTT TGC CCC TC-3′, reverse: 5′-CCT AGA GAC CAG GCC AAA GC-3′), MCAM (forward: 5′-CGG GTG TGC CAG GAG AG-3′, reverse: 5′-CCA CAC TTG AGA AGG GCT GT-3′), β-actin (internal control; forward: 5′-CAT CCG TAA AGA CCT CTA TGC CAA C-3′, reverse: 5′-ATG GAG CCA CCG ATC CAC A-3′), miR-484 (forward: 5′-TCA GGC TCA GTC CCC TC-3′, reverse: 5′-GTG CAG GGT CCG AGG T-3′; Qiagen, #339306) and miR-16-5p (normalizer; 5′-TAG CAG CAC GTA AAT ATT GGC-3′, reverse: 5′-GTG CAG GGT CCG AGG T-3′; Qiagen, #339306). All reactions were performed in triplicate, and relative gene expression was calculated using the 2^−ΔΔCt method.

Western blotting

Proteins were extracted with RIPA buffer (Beyotime, #P0013B), separated by SDS-PAGE, and transferred to PVDF membranes. After blocking (5% skim milk), membranes were incubated with primary antibodies overnight at 4 °C: PECAM1 (1:1000, Abcam, #ab28364), Nuclear factor of activated T-cells, cytoplasmic 1 (NFATc1; 1:800, Abcam, #ab2796), TRAP (1:1000, Abcam, #ab52750), Cathepsin K (CTSK; 1:1000, Abcam, #ab19027), Receptor activator of nuclear factor kappa-B (RANK; 1:800, Abcam, #ab200369), c-Fos (1:1000, CST, #2250), TSG101, CD63, CD81 and calnexin. HRP-conjugated secondary antibodies (1:5000, CST, #7074/#7076) were applied for 1 h at RT, followed by ECL detection (Bio-Rad, #1705060). Band densitometry was performed using ImageJ software (Fiji distribution). For quantification, rectangular regions of interest (ROIs) of identical size were defined around each target band. The local background intensity, measured from an adjacent area within the same lane, was subtracted from each band. The signal intensity of each target protein was then normalized to the corresponding β-actin band from the same sample to control for loading variations.

Sequencing

Total RNA was sequenced on an Illumina NovaSeq 6000 (150 bp paired-end). Raw reads were quality-checked (FastQC), aligned to the mouse genome (GRCm38) using HISAT2, and analyzed for differentially expressed genes (DEGs) with DESeq2, using thresholds of |log2FC| > 1 and a Benjamini-Hochberg adjusted P-value (FDR) < 0.05.

Small RNA libraries were prepared using the NEBNext Multiplex Small RNA Library Prep Kit (NEB, #E7300) and sequenced on an Illumina HiSeq 2500. Novel miRNAs were identified by miRDeep2, and targets were annotated with TargetScan and miRDB. Differentially expressed miRNAs were identified using DESeq2 with thresholds of |log2FC| > 1 and FDR < 0.05.

Generation of PECAM1-modified cell lines

PECAM1 Overexpression (OE) Construct. The full-length coding sequence (CDS) of mouse PECAM1 was amplified by PCR and cloned into the pCDH-CMV-MCS-EF1-Puro lentiviral vector (System Biosciences, #CD510B-1). The insert was verified by Sanger sequencing. For lentiviral production, HEK293T cells were co-transfected with the recombinant plasmid and packaging plasmids (psPAX2 [Addgene, #12260] and pMD2.G [Addgene, #12259]). Viral supernatant was collected 48 h post-transfection, filtered through a 0.45 μm membrane, and used to infect RAW264.7 cells. Stable cell lines were selected with 2 µg/mL puromycin (InvivoGen, #ant-pr-1) for 7 days, and overexpression was confirmed by qPCR and Western blot.

PECAM1 Knockdown Construct. A short hairpin RNA (shRNA) targeting PECAM1 (5′-CCG GGA GCT GAA TGT TGT GTC TCG GCT CGA GCC GAG ACA CAA CAT TCA GCT CTT TTT G-3′) was cloned into the pLKO.1-puro vector (Addgene, #8453), with a scrambled shRNA as negative control. Lentiviral packaging and infection of RAW264.7 cells were performed as described above. Knockdown efficiency was validated by qPCR (primers as in Sect. 4) and Western blot (anti-PECAM1, Abcam #ab28364).

miRNA functional and rescue assays

RAW264.7 cells were transfected with either a miR-484 mimic or a miR-484 inhibitor (RiboBio #miR20003127-1-5) using Lipofectamine 3000 (Invitrogen #L3000015), with scrambled oligonucleotides as controls. Cells were treated with 2.5 × 10¹⁰ particles/mL NC-LLC-Exos or BM-LLC-Exos for 72 h post-transfection. To systematically dissect the miR-484–PECAM1 regulatory axis, eight experimental groups were established: Control (PBS-treated); NC-LLC-Exos (basal exosome control); BM-LLC-Exos (bone-metastatic exosomes); OE-PECAM1 + BM-LLC-Exos (PECAM1 overexpression rescue); NC-LLC-Exos + miR-484 mimic (miR-484 gain-of-function); NC-LLC-Exos + miR-484 mimic + miR-484 inhibitor (miR-484 inhibition rescue); sh-PECAM1 + BM-LLC-Exos (PECAM1 knockdown); sh-PECAM1 + BM-LLC-Exos + miR-484 inhibitor (dual modulation).

Bioinformatics analysis

Putative target genes of miR-484 were predicted using TargetScan (v7.2), miRDB (2020), and miRWalk (v3.0). Only genes identified in ≥ 2 databases were retained for further analysis. DEGs from RNA-seq were analyzed with clusterProfiler (v3.18) for pathway enrichment, including: Disease, Reactome, KEGG and GO enrichment. Gene sets related to OC differentiation and cell adhesion were prioritized.

Experimental design, statistical analysis, and rigor

Definition of Replicates: For all in vitro experiments, an “independent replicate” (denoted as n in the figures) refers to a biological replicate. This is defined as an experiment performed on a separate day using freshly prepared reagents (e.g., exosome batches, cell culture media) and cells from distinct passages. Each independent experiment contained technical replicates (e.g., triplicate wells in qPCR or multiple imaging fields), but the reported ‘n’ value represents the number of independent biological replicates. For in vivo studies, ‘n’ refers to the number of individual animals per group.

Statistical Analysis: Data are expressed as mean ± SD for normally distributed continuous variables. The specific statistical tests used are detailed in the figure legends. Briefly, for comparisons between two groups, an unpaired two-tailed Student’s t-test (or Welch’s t-test if variances were unequal) was performed after verifying normality and homogeneity of variances. For comparisons involving three or more independent groups, one-way ANOVA was conducted, followed by Tukey’spost hoc test for multiple comparisons. All statistical analyses were performed using GraphPad Prism 8.0.

Sample Size Justification and Power Analysis: The sample size for in vivo experiments (n = 6 mice per group) was determined based on previous experience with similar bone metastasis models [19], which indicated that this group size provides sufficient statistical power to reliably detect significant inter-group differences. A post-hoc power analysis conducted for the primary endpoints (bone tumor area and BV/TV) further confirmed that the chosen sample size was adequate to identify the observed biological effects with high confidence.

Results

Establishment of a highly bone-metastatic LLC model and characterization of tumor-derived exosomes

To investigate the role of tumor-derived exosomes in promoting OC activation, bone matrix degradation, and subsequent enhancement of bone metastasis, we established a reliable LUAD bone metastasis model. Using an arterial caudal injection approach (reported to outperform intracardiac injection [16]) combined with three rounds of in vivo selection and in vitro culture/injection, we isolated a highly bone-metastatic LLC subline (BM-LLC), as illustrated in Fig. 1A. The selection process significantly increased the incidence of bone metastasis (NC-LLC: 22.23%; 1st selection: 44.43%; 2nd selection: 77.80%; BM-LLC: 94.43%; P < 0.05 for 2nd/3rd vs. NC-LLC; Fig. 1B). It also shortened metastasis latency from 27.33 to 15.47 days (P < 0.01; Fig. 1C). Transwell assays (collagen I-coated) showed that invasiveness increased with each selection round (Fig. 1D, P < 0.05), with BM-LLC cells (3rd selection) exhibiting approximately 1.6-fold higher invasiveness than ordinary LLC cells (P < 0.01), indicating acquired bone tropism. Isolated exosomes from BM-LLC (BM-LLC-Exos) exhibited typical cup-shaped morphology (Fig. 1F) and a size distribution peaking at 100–150 nm (Fig. 1G), similar to control exosomes (NC-LLC-Exos). Western blot confirmed enrichment of exosomal markers (TSG101/CD63/CD81) and absence of calnexin (Fig. 1H). When co-cultured with RAW264.7 cells (OC precursors), the control group (exosome-free medium) showed no green fluorescence signal, while exosome-treated groups exhibited clear internalization of labeled exosomes (Fig. 1I). Notably, PKH67-labeled BM-LLC-Exos were internalized by RAW264.7 cells approximately 2.4-fold more efficiently than NC-LLC-Exos (P < 0.01), with predominant cytoplasmic localization (Fig. 1J), suggesting enhanced uptake by OC precursors. These results indicate that the established BM-LLC subline not only exhibits enhanced bone metastatic capacity and invasiveness but also secretes exosomes with significantly increased uptake efficiency by OC precursors, suggesting a potential role of tumor-derived exosomes in facilitating PMN formation and promoting OC activation.

Fig. 1.

Establishment of bone-metastatic LLC model and characterization of tumor-derived exosomes. A Schematic of the in vivo selection protocol for generating the BM-LLC subline. B Bone metastasis incidence across selection rounds (n = 6 mice per group). C Metastasis latency from tumor inoculation to detectable bone lesions (n = 6 mice per group). D-E Transwell invasion assay through collagen I-coated membranes showing increased invasiveness with selection rounds. (n = 3 independent experiments; scale bar = 100 μm) F Representative TEM image of isolated exosomes exhibiting typical cup-shaped morphology (scale bar = 1 μm). G Size distribution of exosomes measured by nanoparticle tracking analysis. H Western blot analysis of exosomal markers (TSG101, CD63, CD81) and negative control calnexin. I Confocal microscopy images of PKH67-labeled exosome uptake in RAW264.7 cells; control (exosome-free medium) shows no fluorescence signal (n = 3 independent experiments; 3 fields analyzed per experiment; scale bar = 25 μm). J Quantitative analysis of exosome uptake efficiency by RAW264.7 cells (n = 3 independent experiments). All data represent mean ± SD. ns P > 0.05, *P < 0.05, **P < 0.01 by one-way ANOVA with Tukey’s test or Student’s t-test

Functional validation of bone-metastatic exosomes in promoting OC differentiation

Multi-level assays demonstrated the pro-osteoclastogenic effects of BM-LLC-Exos. TRAP staining revealed that both NC-LLC-Exos and BM-LLC-Exos induced the formation of multinucleated OCs in a dose-dependent manner (approximately 6.25 × 10⁹ to 2.5 × 10¹⁰ particles/mL), with no significant differences among NC-LLC-Exos groups (P > 0.05), all of which reached levels comparable to those induced by M-CSF + RANKL. In contrast, BM-LLC-Exos at each concentration significantly enhanced OC formation compared to NC-LLC-Exos (P < 0.01) (Fig. 2A–C). At 2.5 × 10¹⁰ particles/mL, BM-LLC-Exos generated 71.00 TRAP⁺ cells/field versus 12.33 by NC-LLC-Exos (P < 0.01) (Fig. 2B), and induced a TRAP⁺ area of 2489 μm² compared to 864 μm² in the NC-LLC-Exos group (P < 0.01) (Fig. 2C). Even at the lowest concentration (6.25 × 10⁹ particles/mL), BM-LLC-Exos produced 4.9-fold more TRAP⁺ cells and 2.6-fold larger TRAP⁺ area than the control (P < 0.01). Mechanistically, western blot analysis confirmed that both exosome treatments up-regulated OC markers relative to the control (P < 0.01), including TRAP, CTSK, and RANK (Fig. 2D–G). Notably, BM-LLC-Exos further enhanced the expression of TRAP and RANK at all concentrations (P < 0.05), while a significant increase in CTSK was observed only at 2.5 × 10¹⁰ particles/mL (P < 0.05), compared to NC-LLC-Exos. Phase-contrast microscopy and H&E staining further validated that BM-LLC-Exos promoted larger cell size and increased multinucleation (≥ 3 nuclei) compared to the smaller and fewer nuclei induced by NC-LLC-Exos (Fig. 2H–I). Collectively, these results confirm the potent role of BM-LLC-Exos in activating OC differentiation.

Fig. 2.

Bone-metastatic exosomes promote OC differentiation in vitro. A Representative images of TRAP-stained OCs treated with different concentrations of exosomes (scale bar = 100 μm). B Quantification of TRAP⁺ multinucleated (≥ 3 nuclei) cells per field (n = 3 independent experiments; 5 fields analyzed per experiment). C Measurement of total TRAP⁺ area (n = 3 independent experiments; 5 fields analyzed per experiment). D Western blot analysis of OC markers (TRAP, CTSK, RANK) in exosome-treated RAW264.7 cells. E–G Quantification of the Western blot signals shown in (D) for TRAP (E), CTSK (F), and RANK (G) (n = 3 independent experiments). H Phase-contrast microscopy images of OC morphology (scale bar = 25 μm). I H&E staining showing multinucleation (scale bar = 50 μm). All data represent mean ± SD, ns P > 0.05, *P < 0.05, **P < 0.01, #P < 0.05,##P < 0.01 by one-way ANOVA with Tukey’s test

Integrated multi-omics analysis identifies miR-484–PECAM1 as a key regulatory axis

Integrated transcriptomic and miRNA sequencing analyses revealed key regulators involved in OC differentiation. Volcano plot (Fig. 3A-B) and heatmap (Fig. 3C) analyses showed DEGs in RAW264.7 cells treated with BM-LLC-Exos (|log2FC| > 1, FDR < 0.05). Pathway enrichment analysis (Disease Ontology, Reactome, KEGG, and GO) identified 16 °C-related gene clusters, including “collagen biosynthesis” and “plasma membrane components” (Fig. 3D). Venn analysis further pinpointed ITGAM, PECAM1, and ITGA2 as shared candidates across four major databases (Fig. 3E). qPCR validation confirmed that PECAM1 was significantly downregulated (0.32-fold, P < 0.01) in BM-LLC-Exos-treated cells compared to NC-LLC-Exos, while the other two genes showed no significant reduction (Fig. 3F). Small RNA sequencing was performed to profile miRNA expression in different exosome groups. Volcano plot and density-CPM analysis revealed 198 differentially expressed miRNAs in BM-LLC-Exos (|log2FC| > 2, FDR < 0.05) (Fig. 3G–H). Intersection analysis with miRDB predicted miR-484 and miR-466k as the top candidate miRNAs targeting PECAM1 (Fig. 3I). The expression levels of these two candidate miRNAs were subsequently validated by qPCR. Given that miR-484 expression was significantly higher than that of miR-466k (P < 0.01; Fig. 3J), miR-484 was selected for further investigation.

Fig. 3.

Integrated multi-omics analysis identifies miR-484–PECAM1 as a key regulatory axis. A, B Volcano plot of DEGs in RAW264.7 cells treated with BM-LLC-Exos (|log2FC| > 1, FDR < 0.05). C Heatmap of DEG clusters. D Pathway enrichment analysis of OC-related gene categories (Disease/Reactome/KEGG/GO). E Venn diagram of shared target genes from multiple databases. F qPCR validation of PECAM1 downregulation (n = 3 independent experiments, each with technical triplicates). G, H Volcano plot of differentially expressed miRNAs (NC-LLC-Exos vs. BM-LLC-Exos; |log2FC| > 1, FDR < 0.05) with density distribution of CPM values. I Venn diagram comparing miRNAs upregulated in BM-LLC-Exos (|log2FC|>2, FDR < 0.05) versus miRNAs predicted to target PECAM1. J Expression comparison of miR-484 and miR-466k in BM-LLC-Exos (n = 3 independent experiments). All data represent mean ± SD, ns P > 0.05, *P < 0.05, **P < 0.01 by Student’s t-test

miR-484 directly targets PECAM1 to regulate OC differentiation

To validate the functional interaction between miR-484 and PECAM1, dual-luciferase reporter assays were performed. miRDB analysis predicted two conserved binding sites for mmu-miR-484 within the PECAM1 3′UTR: Site 1 (8-mer, positions 104–111) and Site 2 (7-mer, positions 217–223) (Fig. 4A). Transfection with miR-484 mimic significantly reduced the luciferase activity of the wild-type PECAM1 3′UTR reporter by 58.6% (P < 0.01 vs. NC mimic), while mutation of both binding sites abolished this inhibitory effect (Fig. 4B), confirming direct targeting. Functional assays further demonstrated that miR-484 consistently downregulated PECAM1 expression in RAW264.7 cells at both the mRNA (0.30-fold, P < 0.01; Fig. 4C) and protein levels (0.48-fold, P < 0.01; Fig. 4D–E). In contrast, MCAM, a related family member included as a negative control, remained unaffected (mRNA: 0.94-fold; protein: 0.92-fold), supporting the specificity of miR-484–PECAM1 targeting. To investigate the functional role of PECAM1 in osteoclastogenesis, we established PECAM1-overexpressing (OE-PECAM1) and PECAM1-knockdown (sh-PECAM1) RAW264.7 cell lines (Fig. 4F). qPCR (Fig. 4G–H) and Western blot (Fig. 4I–K) confirmed efficient modulation of PECAM1 expression (OE: 3.83-fold mRNA, 2.36-fold protein; sh: 0.31-fold mRNA, 0.34-fold protein; all P < 0.01), providing essential tools for subsequent functional studies. An RNase protection assay confirmed the intravesicular localization of miR-484, as it was only degraded by RNase upon disruption of the exosomal membrane with Triton X-100 (P < 0.01; Fig. 4M), supporting that its functional delivery is exosome-dependent.

Fig. 4.

miR-484 directly targets PECAM1 to regulate OC differentiation. A miR-484 binding sites in PECAM1 3’UTR predicted by miRDB and TargetScan with conservation analysis across species. B Dual-luciferase reporter assay validating miR-484-PECAM1 interaction using wild-type (WT) and mutant (Mut) constructs (n = 3 independent experiments, each with technical triplicates). C, D qPCR analysis of PECAM1 (C) and MCAM (D) mRNA levels after miR-484 mimic transfection (n = 3 independent experiments, each with technical triplicates). E Representative Western blot images of PECAM1 and MCAM protein expression. F, G Quantification of the Western blot signals shown in (E) for PECAM1 (F) and MCAM (G) (n = 3 independent experiments). H, I Validation of PECAM1-overexpressing (OE) and knockdown (sh) RAW264.7 cells by qPCR (n = 3 independent experiments). J Representative Western blot images for PECAM1 modulation. K, L Quantification of the Western blot signals shown in (J) for PECAM1 (K) and MCAM (L) (n = 3 independent experiments). MCAM served as negative control for target specificity. M RNase protection assay analyzed by qPCR verifies the intravesicular localization of miR-484 (n = 3 independent experiments). All data represent mean ± SD, ns P > 0.05, *P < 0.05, **P < 0.01 by Student’s t-test

Rescue experiments confirm the functional role of the miR-484–PECAM1 axis in OC differentiation

To definitively establish the functional role of the miR-484–PECAM1 axis, we performed a series of rescue experiments. TRAP staining showed that BM-LLC-Exos treatment significantly increased OC formation (75.67 vs. 12.67 cells/field for NC-LLC-Exos, P < 0.01; Fig. 5A–B). This effect was fully reversed by PECAM1 overexpression (52.33cells/field, P < 0.01). Conversely, in NC-LLC-Exos-treated cells, miR-484 mimic increased OC numbers to levels comparable to the BM-LLC-Exos group (71.00 cells/field, 3.2-fold increase, P < 0.01; Fig. 5D–E), while miR-484 inhibitor attenuated this enhancement (39.33 cells/field, P < 0.05). In BM-LLC-Exos-treated cells, PECAM1 knockdown further increased OC numbers (94.67 cells/field, 1.4-fold, P < 0.01; Fig. 5G–H), which was suppressed by miR-484 inhibitor (67.67 cells/field, P < 0.05). OC size followed similar trends: BM-LLC-Exos increased cell area (2260 μm² vs. 787.3 μm² for control, P < 0.01; Fig. 5C), an effect mitigated by PECAM1 overexpression (1692 μm², P < 0.05). miR-484 mimic similarly enlarged OCs (1710 μm², 2.2-fold; Fig. 5F), while PECAM1 knockdown further increased cell size (3493 μm²; Fig. 5I), which was reversible by miR-484 inhibition (2141 μm², P < 0.01). Western blot analysis (Fig. 5J) revealed that PECAM1 overexpression significantly reduced BM-LLC-Exos-induced expression of OC markers (TRAP: 28.1%; CTSK: 20.7%; RANK: 31.4%; Fig. 5K–M) and transcription factors (NFATc1: 23.8%; c-Fos: 28.4%; all P < 0.01; Fig. 5N–O). miR-484 mimic upregulated the expression of these markers and NFATc1/c-Fos (ranging from 18.6% to 30.9%, P < 0.01) in NC-LLC-Exos-treated cells, while miR-484 inhibitor suppressed this activation (50–100% inhibition, P < 0.05 for all except RANK). Conversely, PECAM1 knockdown further enhanced marker expression (14.9–22.3%, P < 0.01) in BM-LLC-Exos-treated cells, which was also reversed by miR-484 inhibitor (over 100% suppression, P < 0.01). Collectively, these data demonstrate that BM-LLC-Exos deliver miR-484 to inhibit PECAM1, thereby activating NFATc1/c-Fos signaling and promoting osteoclastogenesis.

Fig. 5.

Rescue experiments validate the functional role of the miR-484-PECAM1 axis in OC differentiation. A–I Representative TRAP staining (left; A, D, G) and quantitative analysis (right) showing OC number (B, E, H) and size (C, F, I) under indicated treatments (n = 3 independent experiments; 5 fields analyzed per experiment; scale bar = 100 μm). PECAM1 overexpression partially reversed BM-LLC-Exos-induced osteoclastogenesis, while miR-484 mimic in NC-LLC-Exos groups recapitulated BM-LLC-Exos effects. J Representative Western blot analysis of OC markers (TRAP, CTSK, RANK) and signaling molecules (NFATc1, c-Fos). K–O Quantification of the Western blot signals shown in (J) for TRAP (K), CTSK (L), RANK (M), NFATc1 (N), and c-Fos (O) (n = 3 independent experiments). PECAM1 overexpression or miR-484 inhibition attenuated BM-LLC-Exos-induced activation of osteoclastogenic pathways. All data represent mean ± SD, *P < 0.05, **P < 0.01 by one-way ANOVA with Tukey’s test

In vivo targeting of the miR-484–PECAM1 axis attenuates LUAD bone metastasis

To investigate the role of BM-LLC-Exos in promoting bone metastasis and evaluate the therapeutic potential of inhibiting the miR-484–PECAM1 axis, we employed antagomiR-484 to rescue LLC-induced bone metastasis in a model pretreated with BM-LLC-Exos. The experimental timeline is shown in Fig. 6A. Mice were pretreated with 2.5 × 10¹⁰ particles (equivalent to ~ 10 µg protein) of exosomes (NC-LLC-Exos, BM-LLC-Exos, or BM-LLC-Exos + antagomiR-484) on days − 3 and − 1, inoculated with LLC cells on day 0, and received additional exosome injections on days 3 and 6. Tissues and serum were collected at the endpoint (day 21) for analysis. As shown in Fig. 6B–C, no significant difference in lung nodule count was observed among the three groups (P > 0.05). However, BM-LLC-Exos pretreatment significantly promoted bone metastasis compared with NC-LLC-Exos, resulting in increased bone tumor area (54.40 ± 11.65 vs. 1.28 ± 2.55 mm², P < 0.01; Fig. 6D–E). Treatment with antagomiR-484 reduced bone tumor burden (28.27 ± 3.95 mm², P < 0.01). Body weight loss was notably aggravated in the BM-LLC-Exos group by day 20 (P < 0.05), and although antagomiR-484 treatment led to a partial recovery, the difference was not statistically significant (Fig. 6F). Micro-CT analysis (Fig. 6G–H) revealed severe osteolysis (BV/TV: 0.13 ± 0.03 vs. 0.31 ± 0.04, P < 0.05; Fig. 6J) and cortical thinning (thickness: 0.12 ± 0.01 vs. 0.17 ± 0.01 mm, P < 0.05; Fig. 6I) in the BM-LLC-Exos group compared to the NC-LLC-Exos group. AntagomiR-484 treatment ameliorated bone destruction, significantly improving BV/TV (0.23 ± 0.03, P < 0.05) and showing a trend toward restored cortical thickness (0.15 ± 0.02 mm, P > 0.05). Mechanistically, Western blot analysis (Fig. 6K) showed that BM-LLC-Exos downregulated PECAM1 expression (0.43 ± 0.06-fold; Fig. 6L) and upregulated NFATc1 (1.85 ± 0.19-fold; Fig. 6P), c-Fos (1.69 ± 0.20-fold; Fig. 6Q), and OC markers (TRAP: 1.56 ± 0.14, Fig. 6M; CTSK: 1.85 ± 0.18, Fig. 6N; RANK: 1.75 ± 0.15-fold, Fig. 6O) in bone marrow cells. These effects were reversed by antagomiR-484, with PECAM1 expression showing the most pronounced recovery (> 100% increase), while other markers were reduced by 30–65% (P < 0.05). Serum miR-484 levels correlated with osteolytic severity across groups (NC-LLC-Exos: 1.0 ± 0.14; BM-LLC-Exos: 1.56 ± 0.16; antagomiR-484: 1.19 ± 0.11; P < 0.05; Fig. 6R), underscoring the diagnostic potential of circulating miR-484. Collectively, these results demonstrate that targeting the miR-484–PECAM1 axis alleviates LUAD bone metastasis by suppressing exosome-mediated OC activation. Serum miR-484 levels reflect pathological bone loss, highlighting its dual utility as a therapeutic target and diagnostic candidate biomarker.

Discussion

Our study reveals a novel exosome-mediated mechanism in LUAD bone metastasis, In which tumor-derived miR-484 promotes osteolytic destruction through targeted suppression of PECAM1 in osteoclast precursors. These findings deepen the understanding of metastatic niche formation and identify promising diagnostic and therapeutic opportunities.

We employed a bone-metastatic LLC model generated via tail artery injection and in vivo selection. This established a tumor subline (BM-LLC) that demonstrated significantly enhanced bone tropism compared to conventional methods [16]. The derived BM-LLC subline exhibited hallmark metastatic traits, including increased collagen I affinity. Notably, BM-LLC-derived exosomes showed preferential uptake by OC precursors, suggesting active organotropic targeting - a phenomenon increasingly recognized in metastatic progression but still poorly understood mechanistically [20].

Central to our findings is the identification of miR-484 as a direct regulator of PECAM1-mediated OC differentiation. Although PECAM1 is established in vascular biology [21], its bone metastasis role was unclear. Rescue experiments showed PECAM1 overexpression significantly mitigated BM-LLC-Exos-induced OC activation (though not fully abolished), suggesting other regulatory components may participate. The miR-484-PECAM1 interaction specificity was rigorously validated. This included 3’UTR mutagenesis and consistent suppression at mRNA/protein levels. MCAM served as a negative control in these assays.

Therapeutic potential was evidenced by antagomiR-484 treatment in vivo, which significantly attenuated osteolysis and preserved bone architecture in our murine model. While not directly compared to clinical bisphosphonates or denosumab in this study, the degree of bone protection achieved supports the therapeutic potential of targeting the exosomal miR-484–PECAM1 axis. Unlike current anti-resorptive agents that act on mature osteoclasts, antagomiR-484 intervenes earlier by disrupting osteoclast precursor programming, suggesting a complementary mechanism that may enhance existing therapeutic strategies. Notably, serum exosomal miR-484 levels in tumor-bearing mice correlated with osteolytic severity. The miR-484 sequence is highly conserved between mice and humans (miRDB analysis), supporting translational relevance. The natural bone-targeting capacity of BM-LLC-derived exosomes observed in our study may inform future delivery system development, though engineered nanoparticles could offer more controlled targeting approaches [22–24].

Several limitations should be considered. First, our murine model, while robust, cannot fully replicate the chronic progression of human LUAD bone metastasis. Second, the sample sizes in some experimental groups were relatively small, which was primarily due to ethical guidelines governing animal use and the need to minimize the total number of subjects while maintaining statistical rigor. Third, although we isolated vesicles according to MISEV guidelines and confirmed the presence of exosomal markers (TSG101, CD63, CD81) and the absence of calnexin, we cannot entirely rule out the potential co-isolation of other extracellular vesicles or non-vesicular contaminants. Furthermore, the cellular sources and the specific regulatory mechanisms governing the selective packaging of miR-484 into exosomes remain to be elucidated. Furthermore, while our RNase protection assay confirmed the intravesicular localization of miR-484 and functional rescue experiments established its necessity, we cannot formally rule out the potential contribution of exosomal surface proteins in facilitating target cell uptake. Future studies employing protease treatment of exosomes could help dissect the specific role of surface molecules in this process. Meanwhile, Future studies should explore clinical validation in patient samples, optimization of delivery systems (building upon the natural tropism of tumor exosomes or developing bone-targeted nanoparticles), and potential synergies with existing therapies like denosumab or radionuclides. Beyond this specific axis, our findings highlight the broader potential of targeting tumor-derived exosomal miRNAs as a novel therapeutic strategy, warranting the exploration of other candidate miRNAs involved in LUAD bone metastasis.

Conclusion

This work identifies the exosomal miR-484–PECAM1 axis as a key driver of osteolytic metastasis in LUAD, mechanistically linking tumor-derived exosomes to bone microenvironment reprogramming. Beyond revealing fundamental biology, our findings highlight miR-484 as both a candidate liquid biopsy marker and a potential therapeutic target. The demonstrated efficacy of antagomiR-484, combined with the intrinsic targetability of this pathway, suggests new avenues for overcoming current limitations in bone metastasis management. Future translation of these findings may contribute to improved outcomes for patients with this devastating complication.

Author contributions

Zhongkai Tong conceived the study, designed the experiments, and wrote the original draft. Mengqing Hu, Xiaoxiao Zhu, and Zhenyue Ye performed the experiments and collected the data. Cenli Wang, Xiaofei Liang, and Chunli Wu contributed to data analysis and interpretation. Zhenyan Li and Lin He assisted in methodology development and validation. Jian Li supervised the overall project, provided critical revisions, and secured funding. Zhaoxing Dong contributed to study design and served as co-corresponding author. Yong Zhou provided overall supervision, critical revisions, and served as corresponding author. All authors read and approved the final manuscript.

Funding

This work was supported by the Zhejiang Provincial Natural Science Foundation (No. LQ21H160014), the Medical Health Science and Technology Project of the Zhejiang Provincial Health Commission (No. 2023KY275), the Ningbo Municipal Key Clinical Specialty in Respiratory Medicine (No. 2025002), the Ningbo Clinical Research Center for Respiratory Diseases (No. 2022L004), the Ningbo Natural Science Foundation (No. 2023J318), and the Key Research Foundation of Hwa Mei Hospital, University of Chinese Academy of Sciences (No. 2022HMZD09).

Dataavailability

The datasets generated and/or analysed during the current study are available in the ArrayExpress (https://www.ebi.ac.uk/arrayexpress) repository under accession number E-MTAB-15762 for transcriptome sequencing and E-MTAB-15763 for miRNA sequencing.

Declarations

Ethics approval and consent to participate

The animal study protocol was approved by the Institutional Animal Care and Use Committee (IACUC) of Guoke Ningbo Life Science and Health Industry Research Institute (Approval No. GK-2025-XM-0179). All experiments were conducted in accordance with the relevant guidelines and regulations, specifically the GB/T35892-2018 “Guidelines for Ethical Review of Experimental Animal Welfare”. Not applicable. This study does not involve human participants.

Consent for publication

Not applicable. This study does not contain data from any individual person.

Competing interests

The authors declare no competing interests.