We determined the characteristics of the lncRNA transcript by comparing hBMSCs with IL-1β and control groups. To detect differentially expressed transcripts between samples and groups, the edgeR package was employed (www.bioconductor.org/packages/release/bioc/html/edgeR). HISAT2 [21] adoption of global and local search methods can be more effective than RNA sequencing data of spliced reads and is now the most accurate software with the highest ratios. Based on the results of differential analysis, using false discovery rate (FDR) < 0.05 and |log2FC| > 1, a total of 181 lncRNAs were significantly differentially expressed in hBMSCs induced by IL-1β osteogenesis; of these, 112 lncRNAs were up-regulated and 69 were down-regulated (Fig. 2A and B). Gene Ontology (GO) differential gene enrichment analysis showed that IL-1β osteogenic induction primarily focused on the biological process of positively regulating cell differentiation and GO term 0045597 (Fig. 2C and D).

Through flow cytometry, hBMSCs were confirmed to express low levels of CD19, CD34, and CD45 while showing high expression of CD44, CD90, and CD105 (Fig. 3A). Next, we cultured hBMSCs in osteogenic medium to induce osteogenic differentiation and evaluated their potential using Alizarin red S (ARS) (Fig. 3B). For hBMSCs in terms of multi-directional differentiation ability, we observed that the accumulation of lipids increased, lipid droplets gradually increased, and the fat differentiation potential was promoted (Fig. 3B). Alcian blue staining showed that the hBMSCs had the potential to differentiate into cartilage cells (Fig. 3B). We used the RNAplex [22] base dior to identify short interactions between 2 long-chain RNA software packages in order to predict and complement the mRNA antisense lncRNA combination, which contains the ViennaRNA package [23]. Three software packages—miReap, miRanda, and TargetScan—were used to predict miRNA–lncRNA interactions, and miRNA sequences and family data were retrieved from the TargetScan website (www.targetscan.org). The lncRNA–miRNA–mRNA network was built by compiling all coexpressions and competing triplets, and was visualized using Cytoscape V3.6.0 (http://www.cytoscape.org/). Five lncRNAs satisfied the requirements of the above 3 simultaneously. The purpose of this study was to identify the expression of targets by inhibiting and expressing these targets to observe their effects on osteogenesis. Therefore, a novel lncRNA, LncMSTRG25, and its coexpression with mRNA PAX8 and miRNA miR-939-5p attracted our attention (Fig. 3C). We first verified the up-regulation of LncMSTRG25 and PAX8 and the sustained down-regulation of miR-939-5p during osteogenesis using quantitative polymerase chain reaction (qPCR) (Fig. 3D to F). LncMSTRG25 cellular localization was predicted using LncLocator, iLoc-LncRNA, and RNALocate, and the results showed that LncMSTRG25 tends to be located in the cytoplasm. In addition, after the mRNA level of LncMSTRG25 was knocked down using small interfering RNA (siRNA), LncMSTRG25 was labeled using specific probes, and RNA fluorescence in situ hybridization (FISH) results showed that LncMSTRG25 was localized in the cytoplasm (Fig. 3G). Taken together, the findings demonstrated the successful identification of hBMSCs and showed expression changes in LncMSTRG25, PAX8, and miR-939-5p during the differentiation of hBMSCs into osteoblasts.

To further verify the role of LncMSTRG25, we designed 3 specific siRNAs targeting LncMSTRG25, and through qPCR, we chose 2 siRNAs with knockdown efficiency. We transfected siLncMSTRG25-1, siLncMSTRG25-2, and NC (Negative Control) into hBMSCs using lip3000, and PCR assay showed that LncMSTRG25 expression showed a marked decrease relative to the NC control. At the same time, the expression levels of PAX8 and osteogenic markers, including ALP, BMP2, RUNX2, COLL1, OPN, and OCN, were detected, and the results showed that LncMSTRG25 knockdown effectively inhibited osteogenic-related genes at the mRNA and protein levels (Fig. 4A to D). Next, we continued the osteogenic differentiation of hBMSCs for 14 d, and in vitro osteogenic potential was assessed using ARS and ALP (alkaline phosphatase) staining. Interestingly, the ARS and ALP intensities were significantly lower in LncMSTRG25 knockdown hBMSCs than they were in the NC control, reflecting reduced osteogenic differentiation potential (Fig. 4E). We induced osteogenic differentiation of hBMSCs for 3 d and analyzed the expression of PAX8, BMP2, and RUNX2 using immunofluorescence. A significant decrease in PAX8, BMP2, and RUNX2 fluorescence intensities was observed in siLncMSTRG25-treated hBMSCs relative to the NC control group (Fig. 4F). These data suggest that LncMSTRG25 is involved in hBMSC differentiation during development and plays a positive adjustment role.

Lv-LncMSTRG25 and Lv-NC were transfected into hBMSCs to study the osteogenic regulatory effect of LncMSTRG25 overexpression on hBMSCs. The MOI (multiplicity of infection) selection for hBMSCs is shown (Fig. 5A). The positive proportion of green fluorescent cells reached >90% in the MOI = 5 group. Accordingly, an MOI of 5 was chosen for the following experiments. According to the results of the Lv-LncMSTRG25 transfection group, LncMSTRG25 expression was approximately 130 times (Fig. 5B) and significantly inhibited compared with that of LncMSTRG25 miR-939-5p (Fig. 5C). Lv-LncMSTRG25 significantly increased the expression levels of mRNA and proteins for PAX8 and osteogenic markers (ALP, BMP2, RUNX2, COLL1, OPN, and OCN) in hBMSCs when compared with those of Lv-NC. (Fig. 5D to F). The results of ALP and ARS staining showed that Lv-LncMSTRG25 significantly promoted ALP activity and calcium deposition in hBMSCs undergoing osteogenic differentiation for 14 d (Fig. 5G). In addition, the results of cellular immunofluorescence and phalloidinium staining showed that the fluorescence intensity of PAX8, BMP2, and RUNX2 was significantly increased in Lv-LncMSTRG25-transfected hBMSCs compared with that in Lv-NC-transfected hBMSCs (Fig. 5H). These results imply that Lv-LncMSTRG25 substantially enhanced osteogenic differentiation in hBMSCs.

To verify miR-939-5p in hBMSCs during osteogenetic differentiation, we compared miR-939-5p agomir or antagomir transfection of hBMSCs with that of the NC agomir and antagomir groups, which were approximately 10 and 30 times higher, respectively (Fig. 6A). The expression of mRNA and proteins was analyzed using PCR and Western blotting, and the results demonstrated that the antagomir significantly increased the expression of PAX8 and key osteogenic markers at both the mRNA and protein levels in hBMSCs, whereas the agomir had the opposite effect (Fig. 6B to D). At the same time, ALP and ARS staining showed that 14 d of osteogenesis-induced inhibition of miR-939-5p promoted osteogenesis, whereas miR-939-5p inhibited osteogenesis (Fig. 6E). Cellular immunofluorescence and phalloidin staining yielded similar results. Fluorescence intensities of PAX8, BMP2, and RUNX2 were significantly elevated in hBMSCs transfected with the miR-939-5p antagomir, in contrast to the NC group; conversely, hBMSCs transfected with the miR-939-5p agomir exhibited the opposite pattern (Fig. 6F). Taken together, our data support that miR-939-5p has a notable impact on the osteogenic differentiation of hBMSCs.

Bioinformatics predictions suggest that LncMSTRG25 may be a target of miR-939-5p. Figure 7A shows the predicted bioinformatics binding sites between the 2 mRNAs. The dual-luciferase reporter assay further demonstrated that cotransfection with miR-939-5p mimics significantly decreased the relative luciferase activity of LncMSTRG25-WT when compared with the NC group. However, in the presence of miR-939-5p mimics, the luciferase activity of LncMSTRG25-MUT carrying the miR-939-5p binding site mutation remained largely unchanged (Fig. 7B). To examine the role of LncMSTRG25 in the osteogenic differentiation process controlled by miR-939-5p, we first used the coding LncMSTRG25 lentivirus transfection of hBMSCs LncMSTRG25 to stabilize its expression. Cotransfection with Lv-LncMSTRG25 and miR-939-5p mimics partially restored miR-939-5p levels in hBMSCs (Fig. 7C). Next, we assessed the effects of Lv-LncMSTRG25 and miR-939-5p agomir on the osteogenic differentiation ability of hBMSCs. PCR and Western blotting validation showed that in the process of osteogenic differentiation, compared with hBMSCs transfected with Lv-LncMSTRG25, hBMSCs cotransfected with Lv-LncMSTRG25 and miR-939-5p mimics saw reduced expression of the target gene PAX8 and osteogenesis-related genes, including ALP, BMP2, RUNX2, COLL1, OPN, and OCN (Fig. 7D to F). Using FISH, we further verified that LncMSTRG25 and miR-939-5p in hBMSCs were located in the cytoplasm (Fig. 7G). At the same time, compared with the transfection of Lv-LncMSTRG25 hBMSCs, hBMSCs cotransfected with Lv-LncMSTRG25 and miR-939-5p mimics showed reduced ALP and ARS staining intensities (Fig. 7H). Immunofluorescence results for PAX8, BMP2, and RUNX2 in hBMSCs showed a similar trend (Fig. 7I). Our results show that miR-939-5p partially blocks the osteopromotive effect of LncMSTRG25. These findings imply that LncMSTRG25 regulates osteogenesis by interacting with miR-939-5p.

To further investigate the role of PAX8 in osteogenic differentiation of hBMSCs, we designed 3 siRNAs specifically targeting PAX8. We used lip3000 to transfect siPAX8 into hBMSCs, and verified the results by PCR. Two siRNAs with the highest knockdown efficiencies were selected, and PAX8 expression was significantly reduced. We then examined osteogenic markers, including ALP, BMP2, RUNX2, COLL1, OPN, and OCN, at the mRNA and protein levels, and the results showed that PAX8 knockdown effectively inhibited osteogenesis-related gene expression (Fig. 8A to C). We induced osteogenic differentiation of hBMSCs for 3 d and analyzed the expression of PAX8, BMP2, and RUNX2 using immunofluorescence. Compared with the NC control group, the siPAX8, PAX8, BMP2, and RUNX2 groups decreased (Fig. 8D). Next, we proceeded to osteogenic induction of hBMSCs (as described above) for 14 d and evaluated their osteogenic differentiation potential in vitro using ARS and ALP staining. Similarly, compared with the NC group, PAX8 knockdown hBMSCs showed lower ARS and ALP intensities, indicating that the ability of hBMSCs to differentiate osteogenically was impaired after PAX8 knockdown (Fig. 8E). These data support the notion that PAX8 is an important regulator of osteogenic differentiation in hBMSCs.

To predict and verify the potential of miR-939-5p downstream targets, we analyzed the RNAInter database and identified miR-939-5p and PAX8 between the 2 binding sites (Fig. 9A). PAX8-WT1 and PAX8-WT2 were labeled as the binding sites and verified using dual-luciferase reporter plasmids. Relative to the NC group, transfection with miR-939-5p mimics and PAX8-WT significantly decreased luminous intensity. Next, after cotransfection with miR-939-5p mimics, we constructed PAX8-MUT carrying miR-939-5p binding site mutations, labeled as PAX8-MUT1 and PAX8-MUT2. Luciferase activity remained unchanged following cotransfection with miR-939-5p mimics (Fig. 9B). The data imply that miR-939-5p has the potential to interact with both of the predicted binding sites, PAX8 sites 1 and 2. Next, we performed RNA immunoprecipitation (RIP) experiments to explore whether LncMSTRG25 is associated with PAX8. PCR by immunoprecipitation revealed a significant enrichment (25-fold up-regulation) of LncMSTRG25 in the PAX8 group (anti-PAX8) compared with that in the control immunoglobulin G (anti-IgG) group (Fig. 9C). At the same time, in immunoprecipitates containing PAX8, the levels of miR-939-5p were much higher (by 29 times) than those in IgG immune precipitation (Fig. 9D). PCR products further analyzed by agarose electrophoresis showed the anti-PAX8 group enrichment of LncMSTRG25 (Fig. 9E), suggesting that LncMSTRG25 of miR-939-5p has a sponge effect and there is interaction among all 3. Next, we used RNAplex [22,23] to predict whether PAX8 is a potential target of LncMSTRG25. A biotin-labeled LncMSTRG25-specific probe was designed for further validation, and an RNA pull-down assay was performed on hBMSCs. Enrichment of silver staining and Western blotting showed PAX8 protein bands (Fig. 9F to G). To further clarify LncMSTRG25 and PAX8 on osteogenic differentiation, we used PCR and Western blotting to verify PAX8 and osteogenic markers. Lv-LncMSTRG25 and siPAX8 were cotransfected; compared with hBMSCs transfected with Lv-LncMSTRG25, hBMSCs cotransfected with Lv-LncMSTRG25 and siPAX8 showed that the up-regulation of PAX8, ALP, BMP2, RUNX2, COLL1, OPN, and OCN caused by LncMSTRG25 overexpression was inhibited to varying degrees (Fig. 9H to J). The intensity of ALP and ARS staining was significantly weaker than that in the Lv-LncMSTRG25 group (Fig. 9K), indicating that the promotion of osteogenesis by LncMSTRG25 could be partially blocked by siPAX8. Finally, PAX8, BMP2, and RUNX2 immunofluorescence results showed similar trends (Fig. 9L). Collectively, our results indicate that the osteogenic differentiation-promoting activity of LncMSTRG25 is partially regulated by PAX8 knockdown.

We isolated hBMSCs from the bone marrow of patients with limb-long bone fractures; all procedures were performed according to the ethical review board approval of the Union Hospital Affiliated to Huazhong University of Science and Technology (UHCT-IEC-SOP-016-03-01). hBMSCs from 3 donors (average age, 28 years) were cultured in a Dulbecco's modified Eagle's medium/F12 (Gibco, USA), 15% fetal bovine serum (ExCell Bio, China), and 1% penicillin–streptomycin (Beyotime, China) growth medium. hBMSCs at a density of 1 × 10⁵ cells/cm² were seeded in medium-sized cell culture dishes, and the culture conditions were 37 °C and 5% CO₂. The medium was refreshed every 3 d. Separation was based on a previously reported protocol for hBMSCs [49]. The differentiation of hBMSCs was induced using IL-1β (5 ng/ml) (MedChemExpress, USA). Flow cytometry was used to sort and identify specific markers of hBMSCs, CD19, CD34, CD45, CD44, CD90, and CD105 (all from BD, USA).

GenePharma (Suzhou, China) was responsible for the design and synthesis of the miR-939-5p antagomir, miR-939-5p mimics, siRNA, siLncMSTRG25,siPAX8, miR-939-5p agomir, and a negative control. The sequences are shown in Table S1. We used Lipofectamine 3000 plasmid transfection reagent (Invitrogen, Carlsbad, CA, USA) and the corresponding negative controls. In slow-virus infection, LncMSTRG25 induced a slow virus and was provided by Genechem (Shanghai, China). A density of 5 × 10⁴ cells/ml was used to seed hBMSCs in 6-well plates, and once the cells reached 70% confluence, they were infected with lentivirus at an MOI of 5. Lv-NC is an empty lentiviral system lacking insertion sequences.

Following total RNA extraction, ribosomal RNA was removed to enhance the retention of all coding RNA and lncRNA. RNA was randomly fragmented into short lengths, and the resulting fragments served as a template for first-strand cDNA synthesis using random hexamers. The second-strand cDNA was synthesized by adding buffer, deoxynucleotide triphosphates (with dUTP replacing dTTP), ribonuclease (RNase) H, and DNA polymerase I. The purified product was obtained using a QiaQuick PCR kit, eluted with EB buffer, and then processed with end repair, base A addition, and sequencing adapter ligation. Uracil-N-glycosylase was employed to degrade the second strand. Fragment size selection for PCR amplification was carried out through agarose gel electrophoresis. Sequencing was performed on Illumina HiSeq 4000, and GeneDenovo conducted the data analysis (Biotechnology Co. Ltd., Guangzhou, China).

Third-generation hBMSCs were grown in an osteogenic differentiation medium for 7 d (Cyagen, USA). At 3-d intervals, the medium was refreshed, and after induction, the cells were fixed with 4% paraformaldehyde. To evaluate the differentiation ability, ALP staining kits from Beyotime (China) were employed. According to the manual configuration of the ALP staining fluid, the cells were incubated with light for 30 to 60 min. The cells were observed and photographed under a light microscope.

For 14 d, the third-generation hBMSCs were cultured in osteogenic differentiation induction medium provided by Cyagen (USA); medium changes were performed every 3 d. After induction, the hBMSCs were fixed with 4% paraformaldehyde. ARS dyeing liquid (Solarbio, China) mineralization nodule staining was used to evaluate osteogenic differentiation ability. The samples were subjected to manual dyeing for 30 min, visualized under an optical microscope, and documented with photographs (Nikon, Tokyo, Japan).

Third-generation hBMSCs were planted on 12-well culture plates, mounted on a chip, and then allowed to develop in the osteogenetic differentiation medium for 3 d. The expression levels of PAX8, BMP2, and RUNX2 were assessed through immunofluorescence staining. Briefly, cells were fixed at the end of induction with 4% formaldehyde, followed by membrane disruption with 0.5% Triton. Cells were subsequently washed 3 times with PBS. Blocking of nonspecific cell surface binding was performed using goat serum. PAX8 was diluted 1:50 with an antibody diluent (Proteintech, China, 10336-1-AP). RUNX2 (Affinity China, AF5186) and BMP2 (Affinity China, AF5163) were diluted by 1:100, added to the 12-well culture plates, and then incubated at 4 °C overnight. Three washes with TBST (tris-buffered saline–Tween 20) were performed on the cells, and they were then maintained at room temperature with Alexa Fluor 594 mark of goat anti-rabbit secondary antibody (Proteintech, China, RGAR004) before being incubated at 37 °C in the dark for 2 h. Three washes with phosphate-buffered saline (PBS) were performed on the cells at room temperature. To each well, 150 μl of prepared fluorescein isothiocyanate-labeled phalloidin working solution (Solarbio, China) was added. Cells on coverslips were covered, and incubation was carried out in the dark at room temperature for 30 min. Finally, nuclei were stained with 50 μl of blue skies 4′,6-diamidino-2-phenylindole (DAPI) (Beyotime, China) and incubated in the dark at room temperature for 6 min, followed by 3 PBS washes. The slides were removed and placed on slides, and half a drop of an anti-fluorescence quenching agent was added. The slides were sealed with a coverslip and photographed using a fluorescence camera in a dark room (Nikon, Japan).

RNA from hBMSCs was isolated using Trizol reagent (Takara, China), and the One-Step qPCR kit was used for mRNA cDNA synthesis (11142ES60, YeaSen, China) and used cDNA Synthesis Kit (11148ES50, YeaSen, China) for microRNA; total RNA was reverse-transcribed into cDNA. For qPCR, the PCR kit (11201ES08, YeaSen, China) and a fluorescence PCR system were utilized (Bio-Read, USA, CFX Connect). To normalize mRNA and lncRNA levels, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the internal reference, while U6 small nuclear RNA (snRNA) was used for miRNA quantification. Finally, we calculated the relative expression of the target RNA by the 2^−ΔΔCt method. The design and synthesis of all primers were conducted by Sangon Biological (Shanghai, China). The sequences of the primers are shown in Table S2.

After 7 d of osteogenic differentiation of hBMSCs, cell lysates were prepared using RIPA (radioimmunoprecipitation assay) and protease inhibitors (Beyotime, China) for protein extraction. Protein concentration was quantified using a BCA assay kit from Beyotime (China). This was followed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis on polyvinylidene difluoride membranes (Millipore, USA). Samples were blocked using 5% milk powder, followed by primary antibody incubation The primary antibodies used were against PAX8 (Proteintech China,10336-1-AP), OPN (Proteintech China, 22952-1-AP), COLL1 (Proteintech China, 28459-1-AP), OCN (Abclonal China, A20800), Runx2 (Affinity China, AF5186), BMP2 (Affinity China, AF5163), ALP (Affinity China, DF6225), and GAPDH (Proteintech China, 60004-1-Ig). Subsequently, samples were incubated at 4 °C overnight with the secondary antibody (Proteintech China, RGAR001). A chemiluminescence instrument was employed for exposure, and the ImageJ software was used to analyze the images (Bio-Rad, USA,1708370).

Gnecreate (Wuhan, China) was used to design and synthesize recombinant dual-luciferase reporter vectors pmirGLO-LncMSTRG25-WT, pmirGLO-LncMSTRG25-MT, pmirGLO-PAX8-WT1, pmirGLO-PAX8-MT1, pmirGLO PAX8-WT2, pmirGLO PAX8-MT2, and miR-939-5p mimic miR-939-5p NC. hBMSCs were seeded in 96-well plates and cultured to 75% confluence. Lipofectamine 3000 (Invitrogen) was used as the luciferase reporter carrier, miR-939-5p analog, or its negative control (miR-NC). After 48 h, cells were harvested and the firefly and sea kidney luciferase activities were quantified using a dual-luciferase reporter system (Promega, WI, USA). The sequences are shown in Table S3.

FISH kits provided by Gemma (Suzhou, China) were used to detect the expression of LncMSTRG25 and miR-939-5p in the cytoplasm of hBMSCs. In brief, cells were washed at the end of the intervention in PBS and fixed in 4% formaldehyde, and fragmentation was carried out with 0.5% Triton X for 5 min at 4 °C. Subsequently, cells were incubated overnight in the dark at 37 °C for 30 min with 200 μl of degeneration-specific probe mixture added to each hole. Fluorescent tags LncMSTRG25 and miR-939-5p-specific probe were designed and synthesized using GenePharma (Suzhou, China). The Cy3-labeled LncMSTRG25 probe sequence was 5′-CTGAACTATGGAGAGGTGCACAAACCCATGTGTAGACTGATAGCTGTGCACAGTACCATT-3′. The FAM-labeled miR-939-5p probe sequence was 5′-CACCCCCAGAGCCTCAGCTCCCCA-3′. Fluorescent signals were detected using confocal microscopy after staining the nuclei with DAPI (Carl Zeiss, Germany).

Following the manufacturer's instructions, RIP experiments were conducted using the Magna RIP kit (Millipore, USA). Cells (2 × 10⁷) were collected, to which we added cell lysis buffer, protease inhibitors, and RNase inhibitor cell lysis solution; the samples were then centrifuged at 4 °C and 10,000g. The supernatant was removed, and 50 μl of cell lysate was used as the input for qPCR. We took 500 μl of cell lysis solution and added anti-PAX8 (Cell Signaling Technology, USA, D2S2I) magnetic beads or IgG (Proteintech, China, 30000-0-AP) immune coprecipitation magnetic beads. The samples were marked as IP (immunoprecipitation) and IgG groups, respectively. Then, the samples were subjected to a 10 rpm rotating reaction at 4 °C overnight. After thorough washing, the immunoprecipitated RNA was extracted. qPCR was used to detect immune RNA expression during coprecipitation. The PCR products were visualized using ethyl bromide dye on 2% agarose gel, and 6 × DNA sample buffer (Beyotime, China) was added.

A Pierce Magnetic RNA-Protein Pull-Down Kit was utilized for performing RNA pull-down (Thermo Fisher Scientific). We used a biotin-labeled LncMSTRG25 probe sequence (5′-CAAACCCATGTGTAGACTGA-3′-biotin) and contrast probe (5′-TGCTTTGCACGGTAACGCCTGTTTT-3′-biotin). The LncMSTRG25 specificity of the biotin probe was designed and synthesized using GenePharma (Suzhou, China). To obtain cell lysates, the cell lysis buffer was combined with protease and RNase inhibitors and incubated on ice for 20 min. The sample was centrifuged at 12,000g and 4 °C to obtain the supernatant. After the pre-processing of streptavidin magnetic beads, we combined the tubes of magnetic beads with 50 pmol purpose probe or 50 pmol contrast probe, and then subjected them to a 10 rpm rotating reaction for 30 min at 4 °C. The fluid cell lysis supernatant was discarded, and then the sample was subjected to rotating incubation at 4 °C for 1 h. Samples were then washed to collect RNA magnetic bead compounds, and the final protein product was eluted. The obtained products were analyzed for PAX8 expression by Western blotting and silver staining.

Quantitative variables are expressed as the mean ± standard deviation, and comparisons between the 2 groups were performed using a 2-way t test. For comparisons among multiple groups, one-way analysis of variance (ANOVA) was used, with the Bonferroni test for post hoc pairwise comparisons.