Exploration of Crucial miRNA Signatures and Molecular Mechanisms in the System of Muscle–Exosome–Bone: Evidence from Transcriptome Data

Abstract

Muscles and bones are adjacent in spatial position and closely related in function. Exosomes can achieve communication between donor and receptor cells by carrying molecules, such as miRNAs. Therefore, the purpose of this study is to use exosome-miRNA as a bridge, explore the correlation between clinical manifestations, and use bioinformatics and machine learning to cluster and screen exosome-miRNA sequencing data. In vitro and in vivo experiments were conducted to validate the screened molecules. Three parts were explored to identify miRNAs that play a key regulatory role in the muscle–exosome–bone system. Ultimately, it was found that miR-92a-1-5p may play a crucial role in this system; that is, atrophic muscle cells can inhibit osteogenic differentiation by releasing exosomes carrying miR-92a-1-5p into osteoblasts and targeting Col1a1.

Introduction

Postmenopausal osteoporosis generally occurs within 5–10 years after menopause. Osteoporosis is the main risk factor for osteoporotic fractures. The lifetime risk of osteoporotic fracture in women (40%) is higher than that of breast cancer, endometrial cancer, and ovarian cancer combined. Osteoporotic fractures and related complications may lead to disability or death in patients, posing a serious burden on both families and society.

The decrease in estrogen levels in postmenopausal women not only leads to a decrease in bone density and the redistribution of subcutaneous fat to visceral areas but also results in a decrease in muscle mass. Although bone density, muscle mass, and aging are relatedmeaning that bone density and muscle mass decrease with agethis decreasing trend is not linearly correlated. Studies have shown that the changes described above are likely to occur at an accelerated rate before and after menopause in women. Women’s muscle mass often gradually decreases after the third decade and accelerates to decline after the fifth decade. − Due to the close synergistic effect between muscles and skeletal muscles in biological functions, rapid loss of muscle tissue may lead to further acceleration of the bone loss rate.

In recent years, extracellular vesicles with a diameter between 40 and 160 nm, known as exosomes, have been involved in regulating various physiological and pathological reactions, such as inflammation, neurodegenerative diseases, tumor cell invasion and metastasis, vascular permeability generation, immune response, and drug resistance, based on their special regulatory mechanisms, which are released by donor cells and enter recipient cells, playing a critical role in intercellular communication. − Exosomes are widely present in more than 30 types of body fluids, including serum, plasma, urine, cerebrospinal fluid, and bile. Almost all types of cells can produce exosomes. Depending on their source, exosomes have different contents, such as proteins, RNA (miRNA, lincRNA, circRNA, etc.), DNA, amino acids, and metabolites. Among these, miRNA has advantages over other molecules, such as high abundance, wide research field, and depth.

Skeletal muscles are widely distributed in the human body, accounting for about 40% of body weight. Each skeletal muscle has a special shape and structure and is rich in blood vessels, lymph nodes, and neural structures. In addition, skeletal muscles also contain various cells, such as stem cells, fibroblasts, and immune cells, thus playing an important regulatory role in pathological and physiological mechanisms. , Skeletal muscles play important synergistic and auxiliary functions in various systems of the human body such as respiration and metabolism. Therefore, exosomes secreted by skeletal muscle cells can play important regulatory roles among multiple organs and tissues in the human body.

Based on the close connection between muscles and bones and the intercellular communication function of exosomes, we propose a hypothesis that muscle tissue may secrete exosomes with regulatory miRNAs into bones, thereby affecting the occurrence and development of the latter. This study verifies this hypothesis through the following steps: (i) Exploring the correlation between bone density and clinical information such as the relative cross-sectional area (RCSA) of paraspinal muscles through gray correlation analysis (GRA); (ii) Analyzing exosome miRNA sequencing data through bioinformatics and machine learning to screen miRNAs that are correlated with both muscle and bone clinical phenotypes; (iii) verifying whether miRNAs play a regulatory role in the muscle–exosome–bone system through in vitro and in vivo experiments.

Experimental Section

Sample Preparation and Demographic Data

This study recruited a total of 10 postmenopausal patients. First, we recorded these patients’ age, height, and weight; second, we detected the level of estradiol in the serum using the chemiluminescence method (Beckman, Switzerland); and finally, we performed BMD testing and MRI examination on the patients lumbar vertebrae (GE Medical Systems Lunar Prodigy, USA). To control for confounding bias, patients with the following diseases were excluded from the study: tumors; known chronic, systemic, metabolic, or endocrine diseases; inflammation; or acute trauma. To minimize deviations caused by hemolysis, the blood was collected and stored without violent vibration. The candidates were also requested to fast before blood collection during a standardized time window of the day (between 8:00 and 11:00 a.m.).

Establishment of the miRNA Sequencing Matrix

Total RNA was extracted using the mirVana miRNA Isolation Kit (Ambion) according to the manufacturer’s protocol. Quantitation of total RNA was carried out using the NanoDrop 2000 (Thermo Fisher Scientific Inc., USA). RNA integrity was assessed by an Agilent 2100 Bioanalyzer (Agilent Technologies, USA). 1 μg of RNA of each sample was used for small RNA library construction using TruSeq Small RNA Sample Prep Kits (Cat. No. RS-200-0012, Illumina, USA) following the manufacturer’s recommendations. Briefly, the total RNA was ligated to adapters at each end. The adapter-ligated RNA was then reverse transcribed into cDNA and amplified by PCR. The PCR products ranging from 140 to 160 bp were isolated and purified as small RNA libraries. Library quality was assessed on an Agilent 2100 Bioanalyzer system using DNA High Sensitivity Chips. The libraries were finally sequenced using the Illumina HiSeq X Ten platform, and 150 bp paired-end reads were generated. The raw sequencing data are deposited in the Gene Expression Omnibus under accession number GSE255659. Before conducting data analysis, we normalized the sequencing matrix using TPM.

Construction of the mRNA Sequencing Matrix

A total of 1 μg of RNA per sample was used as input material for the RNA sample preparations. Sequencing libraries were generated using the NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, USA) following the manufacturer’s recommendations, and index codes were added to attribute sequences to each sample. After the construction of the library, a qubit2.0 fluorometer was used for preliminary quantification and dilution of the library. After the library reached a concentration of 1.5 ng/μL, the insert size of the library was detected by the Agilent 2100 Bioanalyzer. After the insert size met expectations, qRT-PCR was used to accurately quantify the effective concentration of the library (the effective concentration of the library is higher than 2 nM) to ensure the quality of the library. The clustering of the index-coded samples was performed on a cBot Cluster Generation System using the TruSeq PE Cluster Kit v3-cBot-HS (Illumina) according to the manufacturer’s instructions. After cluster generation, the library preparations were sequenced on an Illumina NovaSeq platform, and 150 bp paired-end reads were generated. The raw sequencing data are deposited in the Gene Expression Omnibus under accession number GSE255763.

Gray Relational Analysis (GRA)

In this study, we used GRA to analyze the correlation between BMD and other clinical data. GRA is a multifactor statistical analysis method, that is based on the geometric shape of the data sequence to determine the degree of correlation. The closer the geometric shape is, the greater the degree of correlation, and vice versa. The specific steps for using GRA in this study are as follows:

(1) Positive normalization of indicators:

① Convert extremely small indicators to extremely large indicators:

$$\mathrm{Max}-x$$ 1

Among them, max is the maximum value among all the data in a certain indicator;

② Convert interval-type indicators to extremely large indicators:

$$M=\max \{a-\min \{x_i\},\max \{x_i\}-b\},{\mathrm{x̃}}_i=\{\begin{array}{l}1-\frac{a-x_i}{M},x_i<a;\\ 1,a\le x_i\le b;\\ 1-\frac{x_i-b}{M},x_i>b.\end{array}$$ 2

Among them, {x_i } is a set of interval-type indicator sequences, and the optimal interval is [a, b].

(2) The sequence forms involved in this analysis include mother sequences and subsequences. The mother sequence, also known as the reference sequence, refers to the sequence that needs to be studied or the data sequence that reflects the behavioral characteristics of the system, similar to the dependent variable Y, which is denoted as x ₀ here. A subsequence, also known as a comparative sequence, is a data sequence composed of factors that affect the behavioral characteristics of a system, similar to the independent variable X, denoted here as (x ₁, x ₂, ..., x_m ). In this study, we aim to investigate BMD as the mother sequence (x ₀), with other influencing factors as subsequences. x ₁ represents RCSA, x ₂ represents age, x ₃ represents the E2 level, and x ₄ represents BMI.

(3) Preprocessing variables:

① Perform dimensionless processing on the data to eliminate the influence of dimensionality caused by different units between the data;

② Reduce the data range to between [0, 1] or around 1 to simplify calculations.

In general, the preprocessing methods for variables in gray relational analysis are initialization and averaging, while the method used in this study is averaging. The specific method is to divide each element in the indicator by the average value of the indicator. Let the standardized matrix be Z, and the elements in Z are denoted as z_ij :

$$z_{ij}=\frac{x_{ij}}{\stackrel{\circledR }{x_{ij}}}$$ 3

(4) Calculate the correlation coefficient (gray coefficient) between each indicator in the subsequence and the parent sequence as

$$y(x_0(k){,x}_i(k))=\frac{a+\rho b}{|x_0(k)-x_i(k)|+\rho b}(i=1,2,...,m,k=1,2,...,n)$$ 4

where a and b are the minimum and maximum differences between the two poles, respectively, and ρ is the resolution coefficient (usually taken as 0.5).

$$a=\begin{array}{l}\min \\ i\end{array}\begin{array}{l}\min \\ k\end{array}|x_0(k)-x_i(k)|$$ 5

$$b=\begin{array}{l}\max \\ i\end{array}\begin{array}{l}\max \\ k\end{array}|x_0(k)-x_i(k)|$$ 6

(5) Calculate gray correlation degree:

$$y(x_0{,x}_i)=\frac{1}{n}\sum _{k=1}^{n}y(x_0(k),x_i(k))$$ 7

Self-organizing Maps (SOM)

SOM is an unsupervised clustering algorithm based on neural networks, first proposed by Finnish scholar Kohonen. SOM can map complex and nonlinear high-dimensional data to low-dimensional spaces with simple geometric structures and interrelationships for display, thereby reflecting the topological relationships between high-dimensional data. Therefore, the SOM clustering algorithm can achieve tasks such as data visualization, clustering, classification, and feature extraction. SOM consists of two layers of a network structure. The number of neurons in the input layer is determined by the feature dimensions of the input data. The input layer, also known as the competition layer, generally includes two planar structures: rectangular and hexagonal. The number of neurons in the competition layer determines the granularity and scale of the final model. According to experience, the minimum number of nodes in the competitive layer is $5\sqrt{N}$ , where N represents the number of training samples. This study used the Kohonen package in the R language to complete SOM clustering, and the specific implementation process is as follows:

(1) Load the sequencing data matrix and standardize the original matrix with Z-score;

(2) The network shape of the competition layer is defined as a hexagon with a network size of 10 × 10;

(3) Show the training process, the trend of distance decreasing with iterations, determine whether iterations are enough, and finally tend to be stable and better;

(4) Further, select the appropriate number of SOM clusters through the rockfall map;

(5) Calculate the number of genes included in each SOM center point;

(6) Calculate the cohesion and quality of the SOM center point;

(7) Check the proximity of the SOM center point and potential boundary points (darker colors indicate greater differences from surrounding points and are more likely to be boundary points);

(8) View the trend of changes in SOM center points;

(9) Obtain the relevant genes for each SOM center point;

(10) Obtain the cluster where miRNAs are located.

Gradient Boosting+Categorical Features (CatBoost)

CatBoost is a machine learning library opened by the Russian search engine company Yandex in 2017, which is a type of Boosting family algorithm. CatBoost is a Gradient Boosting Decision Tree (GBDT) framework based on objective trees, which have fewer parameters, support categorical variables, and have high accuracy. Its main pain point is efficiently and reasonably processing categorical features. CatBoost consists of Categorical and Boosting. In addition, CatBoost also solves the problems of Gradient Bias and Prediction shifts, thereby reducing overfitting and improving the accuracy and generalization ability of the algorithm. In this study, we screened for characteristic miRNAs related to RCSA and BMD through the following steps:

(1) Classify the RCSA values of 10 patients into the high RCSA group and the low RCSA group based on the median, and classify the T-values of 10 patients into the normal group and the osteoporosis group based on the gold standard;

(2) Using CatBoost to filter the characteristic miRNAs of each module in the SOM cluster based on the group and implement the specific parameters of CatBoost as Table :

1. The Parameters of CatBoost.

Name	Values
Data segmentation	0.7
Data shuffle	No
Cross validation	No
Iterations	100
Learning rate	0.1
L2 regularization term	1
The maximum depth of the tree	10
Overfitting detection threshold	0
The number of iterations to continue after achieving optimization	20

Conservation Analysis of miRNA Species

miRBase (http://www.mirbase.org) is an online miRNA database developed by researchers from the University of Manchester. The database contains information about nearly 40,000 miRNAs from more than 200 species, which is the most comprehensive miRNA database. This study used the database to compare the mature miRNA sequences obtained from the final analysis and to screen for miRNAs with interspecies conservation in humans, rats, and mice.

Cell Culture

In this study, we used 2 mouse cell lines: C2C12 (Cell Bank of the Chinese Academy of Sciences, China) and C3H10 T1/2 (Pricella, China). Both cells were cultured in DMEM (Pricella, China) containing 10% fetal bovine serum (FBS). The cells were maintained in a cell incubator containing 5% CO₂ at 37 °C. When the cells proliferated to 80–90% confluency, they were digested with trypsin and passaged at a ratio of 1:3.

Establishment and Identification of a Cellular Atrophy Model

Cells were cultured in six-well plates (1 × 10⁶/plate) overnight in the incubator, and then the DMEM medium containing 2% horse serum (HS) was replaced for 4 days. After 4 days of induction, Pyrvinium pamoate (Sigma, USA) was added to the atrophy group to achieve a final concentration of 5 μM in differentiation medium, and the control group was treated with the same concentration of DMSO (Sigma, USA), and then cultured for 24 h. Optical microscopy (OM) (Ts2R, Nikon, Japan) and scanning electron microscopy (SEM) (Inspect F50, FEI, USA) were used to observe cell morphology. Immunofluorescence (IF) was used to detect the expression levels of muscle atrophy-related proteins Fbx32, and real-time PCR (q-PCR) was used to detect the expression levels of muscle atrophy-related mRNAs Atrogin-1 and Murf-1. Flow cytometry was used to detect cell apoptosis rates.

Immunofluorescence

After the cell density reaches 80–90%, remove the culture medium, add precooled PBS (Pricella, China), clean twice, fix with 4% paraformaldehyde (Aladdin, China) for 10 min, then add PBS containing 0.5% Triton X-100 (Aladdin, China), puncture the cell membrane on ice for 10 min, remove PBS Triton, and wash with precooled PBS three times (5 min each time). Then, perform IF according to the following steps: (1) PBS-3% BSA (Sigma, USA) sealed at room temperature for 30 min; (2) After removing the blocking solution, add the Fbx32 primary antibody (Abcam, USA) (1:400) and incubate for 2 h; (3) Removing the primary antibody, wash with PBS-0.35% Tween 20 (Solarbio, China) three times (5 min each time); (4) Adding fluorescent secondary antibody (Abcam, USA) (1:400) and incubate at room temperature in the dark for 1 h; (5) Removing secondary antibodies, wash with PBS-0.35% Tween 20 three times (5 min each time); (6) Adding 100 ng/mL DAPI solution (Sigma, USA) and incubate at room temperature in the dark for 10 min. Remove the DAPI solution and wash with PBS-0.35% Tween 20 three times (5 min each time). Add antifluorescence quenching solution (Solarbio, China) and take photos under a fluorescence microscope (Ts2R, Nikon, Japan).

Apoptosis

Cells were cultured in a 6-well plate at a density of 1 × 10⁶. We discarded the original medium when the cell density reached to 80% and then followed the steps as the protocols of the manufactures (BD, USA) below to detect the apoptosis level: (1) wash the cells twice with cold PBS and then resuspend them in 1× Binding Buffer at a concentration of 1 × 10⁶ cells/mL; (2) transfer 100 μL of the solution (1 × 10⁵ cells) to a 5 mL culture tube; (3) add 5 μL of FITC Annexin V and 5 μL PI; (4) gently vortex the cells and incubate them for 15 min at RT (25 °C) in the dark; (5) add 400 μL of 1× Binding Buffer to each tube. Analyze by flow cytometry within 1 h.

Exosome Isolation, Qualification, and Characterization

After the successful construction of the muscle atrophy model, DMEM with 10% exosome FBS (Cytogen, China) was used, and the cell supernatant was collected 24 h later. In this study, we used the ultracentrifugation method to extract exosomes, and the detailed steps are as follows: (1) the supernatant was harvested and centrifuged at 300g for 10 min, 2000g for 10 min, and then 10,000g for 30 min; (2) the supernatant was centrifuged at 120,000g for 70 min (Optima L-80 XP, Beckman, USA), and the pellet was resuspended in phosphate-buffered saline (PBS); (3) the sample was centrifuged again at 120,000g for 70 min. The supernatant was carefully removed, and the sample was resuspended in PBS. Western blotting (WB) was used to identify the membrane protein of the exosomes, and exosome morphology was analyzed using transmission electron microscopy (TEM) (TECNAI 10, Philips, Holland). To examine exosome size distribution and particle concentration, nanoparticle tracking analysis (NTA) was performed using a ZetaView instrument (Particle Metrix, Germany).

Western Blotting

The RIPA solution was added to the exosomes, followed by blowing and mixing well, and then cracking on ice for 20 min. After centrifuging the obtained cracking solution at 12,000 rpm for 15 min and taking the supernatant, a NanoDrop spectrophotometer was used to detect protein concentration. After quantification, add SDS loading buffer and boil for 10 min. After cooling, the protein was stored at −20 °C. The protein was taken and SDS-PAGE electrophoresis performed (concentrated gel: 80 V, 30 min; separation gel: 120 V, 90 min). After electrophoresis, the membrane was transferred (90 V, 1 h). After the membrane transfer was completed, a sealing solution (5% skimmed milk powder in TBST solution) was added, and the mixture was sealed on a shaker for 1 h. The primary antibody was diluted with TBST solution containing 1% BSA and incubating at 4 °C overnight, afterward, the membrane was washed three times with TBST on a decolorization shaker at room temperature, each time for 10 min. The dilution ratio of Calnexin/CD9/CD63/CD81/HSP70 and TSG101 antibodies (Abcam, USA) at 1:1000; the dilution ratio of β-actin is 1:5000. The secondary antibody was diluted with TBST at a ratio of 1:5000, incubated at room temperature for 2 h, and then washed three times with TBST on a decolorization shaker at room temperature for 10 min each time; Tanon ECL (Beyotime, China) was used for chemiluminescence, and photos were taken under a chemiluminescence imager to obtain images.

Uptake of Exosomes

Inoculating C3H10 T1/2 cells were inoculated into a 3.5 cm Petri dish at a density of 5000/well and cultured overnight at 37 °C. After the cells adhered to the wall, the exosomes were resuspended and incubated with PKH67 (Solarbio, China) at a ratio of 1:250 for 2 h to stain the exosomes. Then, exosome suspension was added to the cells, cleaned, and the solution was changed after incubation for 0, 3, and 6 h. The uptake of exosomes was observed under a fluorescence microscope, and photographs were taken.

Real-Time PCR

The total RNA extracted via the TRIzol reagent was applied to synthesize complementary DNA (cDNA). mRNA expressions were analyzed quantitatively by using a real-time quantitative polymerase chain reaction. The primers are listed in Table .

2. Primer Sequences Used for Real-Time PCR Amplification.

Gene	Forward primers	Reverse primers	RT primer
U6	CTCGCTTCGGCAGCACA	AACGCTTCACGAATTTGCGT
mmu-miR-144-3p	GCGCGCGTACAGTATAGATGA	ATCCAGTGCAGGGTCCGAGG	GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACAGTACA
mmu-miR-183-5p	AGCGAGGCTATGGCACTGGTA	ATCCAGTGCAGGGTCCGAGG	GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACAGTGAA
mmu-miR-92a-1-5p	ACCATTCAAGGTTGGGATCGGTT	ATCCAGTGCAGGGTCCGAGG	GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACAGCATT
mmu-miR-484	AGTATCCTCAGGCTCAGTCCCC	ATCCAGTGCAGGGTCCGAGG	GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACATCGGG
Gapdh	ACCCAGAAGACTGTGGATGG	ACACATTGGGGGTAGGAACA
Atrogin-1	AGAGAGGCAGATTCGCAAGCGT	TGCAAAGCTGCAGGGTGACCC
Murf-1	acctgctggtggaaaacatc	cttcgtgttccttgcacatc
Alp	TCCGTGGGCATTGTGACTAC	TGGTGGCATCTCGTTATCCG
Ocn	GGTAGTGAACAGACTCCGGC	GGCGGTCTTCAAGCCATACT
Col1a1	CGATGGATTCCCGTTCGAGT	GAGGCCTCGGTGGACATTAG

In Vitro Osteogenic Induction of C3H10 T1/2 Cells

Cells were cultured in 6-well plates (2 × 10⁶/plate) overnight in the incubator. After the cell density reached 80%, the culture medium was discarded, the cells were washed twice with PBS, and osteogenic induction medium was added for further culture. The induction medium was changed every 2 days. The osteogenic induction medium contained 50 μg /mL ascorbic acid (Sigma, USA), 10 mM β-glycerophosphate (β-GP) disodium (Sigma, USA), and 10 nM dexamethasone (Sigma, USA).

We discarded the original medium before staining and cleaned the cells twice using PBS, and then we used the BCA (bicinchoninic acid) method to prepare the protein samples. The ALP/ARS staining was analyzed quantitatively using a microplate reader with 562/405 nm absorbance (Beyotime, China).

The ALP activity (Beyotime, China) test followed the steps below steps: (1) cells were cultured in 96-well plates at a density of 1 × 10³; (2) the original medium was discarded before staining, and the cells were cleaned twice using PBS; (3) lysis solution was added, and the supernatant was collected after centrifugation , mixed with the detection reagent, and incubated for 30 min at room temperature in a dark environment); (4) absorbance was measured at 405 nm using a microplate reader.

Cell Transfection

Cell transfection was performed as follows: (1) Solution A: 10 nM miRNA (Hanbio, China) was diluted in serum-free culture medium Opti-MEM (Gibco, USA), gently mixed, and incubated at room temperature for 5 min; (2) Solution B: 5 μL Lipo3000 (Invitrogen, China) was diluted in serum-free culture medium Opti-MEM, gently mixed, and incubated at room temperature for 5 minutes. Solutions A and B were gently mixed, incubated at room temperature for 20 min, added to a 6-well plate, and incubated in an incubator for 48 h.

Cell Coculture

Inoculate C3H/10T1/2 cells separately in the lower chamber of the transwell and transfer miRNA-transfected C2C12 cells to the upper chamber of the transwell. After the cells have formed iron walls, coculture the cells. In the intervention group, add the exosome secretion inhibitor GW4869 (Sigma, USA) and coculture for 12 h. Remove the cells from the incubator, remove the culture medium, and gently wash the cells twice with precooled PBS. Add 100 ng/mL Hoechst (Abcam, USA) solution and incubate at room temperature in the dark for 10 min. Remove the Hoechst solution, add PBS, and wash 3 times for 5 min each time. Observe and take photos under a fluorescence microscope.

Cell Proliferation

Cultivating cells in 96-well plates with 1000 cells/well. Before testing, the old culture medium was removed, and the cells were washed twice with PBS. Then, the culture medium with CCK-8 (Beyotime, China) detection reagent (1:10) was added to the well plate and placed in the cell incubator for 1 h. Then, a microplate reader was used to measure the absorbance at 450 nm.

Target-Gene Prediction and KEGG/GO Pathway Enrichment

Predict and intersect the target genes of miRNA through five online prediction databases: RNA22, RNAIntel, TargetMiner, TargetScan, and miRDB. Perform online functional enrichment and prediction of gene sets through the Metascape Web site, with parameters set to (1) min overlap: 3; (2) p value cutoff: 0.01; (3) min enrichment: 1.5.

Time Series Clustering

Perform time series clustering on transcriptome sequencing data of osteogenic differentiation at four time points (days 0, 1, 7, and 14) using STEM and Mfuzz. Bioinformatic analysis was performed using the OmicStudio tools at https://www.omicstudio.cn/tool.

Double-Luciferase Experiment

Cells were cotransfected with luciferase vectors containing wt or mut 3′-UTR of Col1a1 and miR-92a-1-5p mimics or miR-Control by using Lipofectamine 3000 (Invitrogen). Luciferase activity was measured using a Dual-Luciferase Reporter Assay System (Beyotime, China) at 48 h after transfection.

Establishment of Animal Models

The construction method of the osteoporosis animal model in this study was to remove the bilateral ovaries of female SD rats (6–8 weeks old), remove the adipose tissue around the ovaries in the control group, and evaluate the bone tissue changes of the L4 vertebral body of the rats using H&E staining after 3 months. The construction method of a muscle atrophy model is to inject botulinum toxin type A (BTXA) (Lanzhou Institute of Biological Products, China) into the erector spinal muscle of rats (5U) and inject physiological saline into the control group. After 3 months, Masson staining and qPCR were used to evaluate the atrophy of the paraspinal muscles in rats.

In Vivo Injection and Verify

Using antago-miR, antago-miR-NC/ago-miR, and ago-miR-NC (Hanbio, China), we injected the rat tail vein (1 μM) once every 2 weeks for a total of 4 injections. After the last 2 weeks of injections, the rats were euthanized. H&E staining was used to detect pathological changes in the L4 vertebral body, and immunohistochemical staining was used to detect the expression changes of COL1A1 (Abcam, USA) in the L4 vertebral body.

Statistical Analysis

The data are expressed as mean ± standard deviation. Statistical differences between three or more groups were analyzed by one-way analysis of variance (ANOVA) and the Bonferroni post hoc test. The significance of the difference between two groups was determined through an unpaired Student’s t-test. Statistical analysis was conducted using SPSS software 19.0. p < 0.05 was considered statistically significant.

Results

BMD and RCSA Have a High Gray Correlation Degree

The demographic data of 10 donors are shown in Figure A. From the cross-sectional MRI of the L2–L3 level, donors with high BMD have a higher RCSA compared to those with low BMD (Figure B). To further analyze the relationship between the two, we used GRA to analyze the correlation between BMD, RCSA, and other clinical data (Figure C) In this study, based on the formula for calculating the gray correlation degree, we obtained: ① Y (x ₀, x ₁) = 0.808; ② Y (x ₀, x ₂) = 0.781; ③ Y (x ₀, x ₃) = 0.638; ④ Y (x ₀, x ₄) = 0.696. The correlation value ranges from 0 to 1, and the larger the value, the stronger the correlation between it and the “reference value” (parent sequence), which means its evaluation is higher. For the four evaluation items in this study, RCSA scored the highest (correlation degree: 0.808), followed by age (correlation degree: 0.781), BMI (correlation degree: 0.696), and finally E2 (correlation degree: 0.638).

1.

Screening crucial miRNAs associated with muscle and bone phenotype simultaneously based on clinical and sequencing data. (A) Demographic data of donors included in the study. (B) The MRI image of two typical donors (B1) has a higher RCSA than (B2). (C) GRA was used to analyze the relation among age, BMI, E2, and RCSA with the BMD. (D–J) Reduce dimensionality and cluster exosome miRNA data matrix using SOM: (D) schematic diagram of SOM; (E) training progress showed whether the number of iterations is enough; (F) to select the appropriate number of clusters through the rockfall map further; (G) the introversion and quality of SOM center node; (H) the number of genes included in the SOM center node; (I) proximity distance of SOM center node; (J) classified SOM clusters. (K–M), Screening for species conserved miRNAs associated with both BMD and RCSA by CatBoost simultaneously: (K) the top 5 miRNAs among different clusters based on the high and low BMD grouping; (L) the top 5 miRNAs among different clusters based on the high and low RCSA grouping; (M) candidate miRNAs with interspecific conservation that intersect the top 5 feature values in both BMD and RCSA phenotypes within the same cluster.

There Is a Correlation between Four Exosomal miRNAs and Clinical Phenotypes of BMD and RCSA

The extraction and quality inspection of exosomal miRNAs from serum samples of all patients were qualified. There are 485 miRNAs by sequencing totally. The training process showed that as the number of iterations decreased, the curve eventually stabilized, proving that the number of iterations met the requirements (Figure E). The drop plot of the sum of squares within clusters with the number of clusters showed that when the number of clusters reached 7, the curve tended to stabilize. Therefore, it was more appropriate to divide the sequencing data into 7 cluster modules (Figure F). As shown in Figure G–J, based on the above judgment, the sequencing data was ultimately divided into 7 different clusters, and the miRNAs in each cluster may have similar trends in biological functions. Cluster 1 contains 72 miRNAs, Cluster 2 contains 15 miRNAs, Cluster 3 contains 41 miRNAs, Cluster 4 contains 30 miRNAs, Cluster 5 contains 166 miRNAs, Cluster 6 contains 100 miRNAs, and Cluster 7 contains 61 miRNAs. Grouping by BMD, the top 5 miRNAs with feature contribution values for each module in the 7 clusters (Figure K) were obtained. Grouping with RCSA, the top 5 miRNAs with feature contribution values for each module in the 7 clusters were obtained (Figure L). The top 5 miRNA molecules with feature contribution values in both the BMD and RCSA groups were intersected in each cluster. In Cluster 1, no intersecting miRNAs were identified. In Cluster 2, a total of four intersecting miRNAs were obtained: hsa-miR-1-3p, hsa-miR-206, hsa-miR-9-5p, and hsa-miR-184. In Cluster 3, one intersecting miRNA was identified: hsa-miR-183-5p is obtained; In Cluster 4, three intersecting miRNAs were obtained: hsa-miR-92a-1-5p, hsa-miR-181b-5p, and hsa-miR-3190-3p. In Cluster 5, one intersecting miRNA was identified: hsa-miR-144-3p is obtained. In Cluster 6, no intersecting miRNA were identified. In Cluster 7, a total of two intersecting miRNAs were identified: hsa-miR-484 and hsa-miR-422a. Species conservation analysis was performed on the miRNA molecules obtained from SOM and CatBoost analyses, resulting in four miRNAs (Figure M) with their sequences shown in Table .

3. miRnas with Interspecies Conservation and Their Sequences.

ID	Sequence
miR-484	TCAGGCTCAGTCCCCTCCCGAT
miR-92a-1-5p	AGGTTGGGATCGGTTGCAATGCT
miR-183-5p	TATGGCACTGGTAGAATTCACT
miR-144-3p	TACAGTATAGATGATGTACT

Successful Construction of a Myotube Atrophy Cell Model

After constructing the cell model, we first observed the morphological characteristics of the atrophic myotubes and the control group myotubes under OM and SEM. The results showed that the control group had full cells and a complete myotube structure under the microscope, while the intervention group had cell atrophy and structural integrity destruction under the microscope (Figure B). Under the light microscope, the muscle tube diameter of the atrophy group was lower than that of the control group (Figure C). The IF results showed that the muscle atrophy characteristic protein FBX32 was highly expressed in the cytoplasm of the atrophy group (Figure D,E). In addition, the apoptosis rate of cells in the atrophy group was significantly higher than that in the control group (Figure F,G). qPCR results showed that compared to the control group, the characteristic genes of muscle atrophy (Atrogin-1, Murf-1) were highly expressed in the atrophy group (Figure H).

2.

Screening and validation of candidate miRNAs through experiments. (A), Schematic diagram of the experiment (graphic created using Figdraw). (B–H) Establishment of muscle atrophy cell model: (B) morphology of muscle fibers in the control and atrophy groups under OM and SEM; (C) comparison of myotube diameter between the control and atrophy groups under OM; (D) using IF technology to detect the expression levels of FBX32 in the cytoplasm of the control and atrophy groups; (E) comparison of average fluorescence intensity of IF detection between the control and atrophy groups; (F) flow cytometry scatter plots; and (G) comparison of cell apoptosis rates between the control and the atrophy groups; (H) after constructing the cell model, real-time PCR technology was used to detect the content of Atrogin-1 and Murf-1 in the cells. The average expression of mRNA in the control group was standardized with Gapdh as the internal reference (n = 3). (I–L), Identification of exosomes in the supernatant of cell models and screening of candidate miRNAs through real-time PCR: (I) WB detection of protein expression of exosomes characteristic membrane proteins in control group, atrophy group, and cells (J) TEM was used to detect the morphology of exosomes in the control group and the atrophy group; (K) NTA detection of exosomes particle size in the control group and atrophy groups; (L) real-time PCR technology was used to detect the expression levels of exosomes candidate miRNAs in the control group and atrophy group, standardize the mean miRNA expression in the control group, and use U6 as the internal reference (n = 3). (M,N) PKH67 marked exosomes are ingested by recipient cells: (M) fluorescence intensity of exosomes in the control group and atrophy groups at 0, 3, and 6 h after the addition to receptor cells; (N) the fluorescence expression levels at 0, 3, and 6 h after the addition of exosomes to receptor cells in the control group and atrophy groups. ns = No statistical significance, *p < 0.05, **p < 0.01, ***p < 0.001.

miR-92a-1-5p Is Highly Expressed in Exosomes of Atrophic Cells

Exosomes were extracted from the cell supernatants of the control group and atrophy group and validated using WB, TEM, and NTA. The WB results showed that specific membrane proteins CD9, CD81, CD63, HSP70, and TSG101 were highly expressed in exosomes, while Calnexin was lowly expressed (Figure I). The TEM results showed that the exosomes exhibited a typical “tea cup” shape (Figure J), and the NTA results showed that the extracted exosomes peaked at 100 nm (Figure K). Total RNA was extracted from the exosomes of the atrophy group and control groups and validated using qPCR with four candidate miRNAs: miR-92a-1-5p, miR-484, miR-183-5p, and miR-144–3p. The results showed that miR-92a-1-5p was differentially expressed in the atrophy group’s exosomes, while there was no difference in the expression levels of miR-484, miR-183-5p, and miR-144-3p (Figure L). Subsequently, we labeled the exosomes with PKH67 and added them to the culture medium of receptor cells (C3H10 T1/2). At two time points of 3 and 6 h, the exosomes extracted from the control group and atrophy group cell models entered the receptor cells, and there was no statistically significant difference in fluorescence intensity between the two groups (Figure M,N).

miR-92a-1-5p Can Enter Receptor Cells through the Exosome Pathway to Exert Regulatory Functions

After miR-92a-1-5p labeled with Cy3 was transfected into C2C12 and cocultured with recipient cells for 12 h, the results showed a significant decrease in fluorescence intensity in recipient cells in the GW4869 group compared to the DMSO group (Figure B,C). To verify the regulation of miR-92a-1-5p on the function of receptor cells, we first transfected miR-92a-1-5p mimics, inhibitors, and corresponding controls into the receptor cells. The qPCR results showed that after transfection with mimics, the expression level of miR-92a-1-5p was higher than that of the mimics control group, while after transfection with inhibitors, the expression level of miR-92a-1-5p was lower than that of the inhibitor control group, and the transfection system met our expectations (Figure D). Subsequently, absorbance was measured using the CCK-8 method at 5 time points of 0, 24, 48, 72, and 96 h. The results showed that compared to the mimic control group, mimics could inhibit the proliferation ability of receptor cells, while compared to the inhibitor control group, the inhibitor could promote the proliferation ability of receptor cells (Figure E). At 48 h, flow cytometry was used to detect the apoptosis level of receptor cells. The results showed that compared to the mimic control group, mimics could increase the apoptosis level of receptor cells, while compared to the inhibitor control group, the inhibitor could reduce the apoptosis level of receptor cells (Figure F,G). After 7 days of induction with an osteogenic induction solution, the qualitative and quantitative results of ALP staining showed that the staining depth of mimics was weakened relative to the mimic control, and the ALP activity was lower. The staining results of the inhibitor were deeper compared to the inhibitor control, and the ALP activity was higher (Figure H,J,K). After induction with osteogenic induction solution for 14 days, the qualitative and quantitative results of ARS staining showed that the staining depth of mimics relative to the mimic control was weakened, and the staining results of the inhibitor relative to the inhibitor control were deeper (Figure I,L). After being incubated with osteogenic induction medium for 14 days, the total RNA of receptor cells was extracted, and the expression levels of bone characteristic genes Alp, Col1a1, and Ocn were detected for mimics, inhibitors, and corresponding controls, and blank controls, respectively. The results showed that mimics had relatively low expression compared to the mimic control for osteogenic characteristic genes, while inhibitors had relatively high expression compared to the inhibitor control for osteogenic characteristic genes (Figure M–O).

3.

The effect of C2C12-derived exosomal miRNA on osteogenic differentiation of C3H10 T1/2 cells. (A), Schematic diagram of the experiment (graphic created using Figdraw). (B,C) miR-92a-1-5p can enter receptor cells through the exosome pathway: (B) fluorescence expression levels of DMSO and GW4869 groups in receptor cells and (C) average fluorescence intensity between two groups in receptor cells. (D) Real-time PCR technology was used to detect the expression levels of miR-92a-1-5p after transfection into recipient cells, standardize the mean miRNA expression in the DMSO group, and use U6 as the internal reference (n = 3). (E–O) The effects of miR-92a-1-5p transfection into recipient cells on cell proliferation, apoptosis, and osteogenic ability: (E) CCK-8 assay was applied to measure the cell proliferation at 0, 24, 48, 72, and 96 h; (F) flow cytometry scatter plots and (G) comparison of cell apoptosis rates among 5 groups; (H) the ALP staining result under the light microscope at 7 days; (I) the ARS staining result under the light microscope at 14 days; (J) the quantify analysis result of ALP; (K) the ALP activity result; (L) the quantify analysis result of ARS. (M–O) After transfection, real-time PCR technology was used to detect the content of Col1a1, Alp, and Ocn in the cells. The average expression of mRNA in the mock group was standardized with Gapdh as the internal reference (n = 3). ns = No statistical significance, *p < 0.05, **p < 0.01, ***p < 0.001.

miR-92a-1-5p Can Target Col1a1 to Inhibit Receptor Cell Osteogenic Function

Using online data sets to predict and intersect the target genes of miR-92a-1-5p, a total of 77 target genes were obtained (Figure A). Then, Metascape was used to perform GO and KEGG enrichment functional analyses on these 77 target genes (Figure B). GO enrichment analysis revealed a total of 20 functional pathways, among which there were skeletal-related pathways: skeletal system morphogenesis (GO: 0048705), including a total of 13 mRNAs (Figure C): Hoxa7, Palld, Mapk1, Acsl4, Sox21, Col1a1, Tfap2b, Il10, Hoxc13, Trps1, Six4, Ybx3, and Pax5, and KEGG enrichment analysis revealed a total of 4 pathways. To further determine the target genes of miR-92a-1-5p, we first used time series-related bioinformatics analysis methods Mfuzz and STEM to cluster transcriptome sequencing data at 4 time points: 0, 1, 7, and 14 days of osteogenic induction. Finally, we obtained 2 intersecting mRNAs Palld and Col1a1 (Figure D–G) by intersecting the Mfuzz, STEM, and skeletal system morphogenesis pathways, where Col1a1 is closely related to osteogenic differentiation. For this reason, we chose Col1a1 as the target gene for miR-92a-1-5p. We found a binding site between miR-92a-1-5p and Col1a1 through TargetScan prediction (Figure H). Subsequently, we validated this binding through a dual-luciferase assay, and the results showed that the relative fluorescence intensity of the wt Col1a1 plasmid and miR-92a-1-5p plasmid groups was lower than that of the wt Col1a1 plasmid and miR-92a-1-5p control plasmid groups. The relative fluorescence intensity of the group transfected with the mut Col1a1 plasmid and miR-92a-1-5p plasmid showed no difference compared to the group transfected with the mut Col1a1 plasmid and miR-92a-1-5p control plasmid (Figure I).

4.

Prediction and validation of miR-92a-1-5p target genes. (A,) Venn plot for predicting miR-92a-1-5p target genes. (B), GO and KEGG enrichment analyses of target gene functions and pathways. (C) The genes in the pathway of skeleton system morphogenesis. (D,E) Gene modules related to osteogenic induction in C3H10 T1/2 cells through bioinformatics analysis: (D) Mfuzz and (E) STEM time series analysis were used to analyze the sequencing data at different time points in osteogenic differentiation. (F) Intersecting genes among the characteristic cluster of Mfuzz, STEM analysis, and skeletal system morphogenesis genes. (G) Schematic diagram of bone induction at different time points (graphic created using Figdraw). (H) Prediction of binding sites between miR-92a-1-5p and Col1a1. (I) Verification of binding sites by dual luciferase assay. ns = No statistical significance, *p < 0.05, **p < 0.01, ***p < 0.001.

In Vivo Experiments Have Shown That miR-92a-1-5p Promotes Osteoporosis by Targeting COL1A1

To verify that muscle atrophy can accelerate the rate of bone loss, we constructed 6 groups of rat animal models. The degree of bone changes in the rat model was evaluated through H&E staining, while the degree of fibrosis in the paraspinal muscles was evaluated using Masson staining and qPCR. The results showed that compared to the EM+N C., the EM+BTXA group showed a higher degree of muscle fibrosis, and the trabecular structure of the rat vertebral body became thinner. There was no significant difference in the degree of paravertebral muscle fibrosis between the OVX shape group and the OVX group rats, and compared to the OVX shape group, the bone trabecular structure of the vertebrae in the OVX group rats became thinner and the number decreased. Compared to the OVX-EM+NC group, the OVX-EM+BTXA group of rats had a higher degree of paravertebral muscle atrophy, and the vertebral trabecular structure of the OVX-EM+BTXA group of rats became thinner, the number decreased, and the gap increased (Figure A–E).

5.

Establishment of animal models and validation of miR-92a-1-5p and Col1a1 in vivo experiments (graphics in the schematic section using Figdraw). (A) H&E staining of vertebrae in different animal models. (B–E) Verification of the degree of erector spinae muscle atrophy in different animal models: (B) Masson staining of erector spinae muscle in different animal models; (C) comparison of CVF(%) among different animal models; (D) after constructing the animal model, and real-time PCR technology was used to detect the content of Atrogin-1 and Murf-1. (E) The average expression of mRNA in the control group was standardized with Gapdh as the internal reference (n = 3). (F–I) In vivo experimental verification of the effect of miR-92a-1-5p on osteogenic function: (F) H&E staining of different groups of vertebrae after tail vein injection and (G) the COL1A1 IHC among different groups of vertebrae after tail vein injection (H). The quantitative analysis between Groups A and B. (I) The quantitative analysis between Groups C and D. ns = No statistical significance, *p < 0.05, **p < 0.01, ***p < 0.001.

Through the construction of the above animal model, we believe that muscle atrophy can to some extent lead to pathological changes in the trabecular structure of rat vertebrae, and in osteoporotic rats, this change may accelerate bone loss and ultimately lead to the progression of osteoporosis. To further investigate the role of miR-92a-1-5p–COL1A1 in the muscle–skeleton system, we selected the OVX-EM+BTXA group and the OVX-EM+N C. group to perform subsequent validation. The experimental results showed that compared to Group B, the vertebral trabecular structure of Group A rats was denser and more numerous and the gap between bone trabeculae was smaller. The IF results showed that the proportion of COL1A1-positive cells in Group A was higher. Compared to Group C, the bone trabecular structures in Group D were denser and more numerous, and the IF results showed a higher proportion of COL1A1-positive cell areas in Group D (Figure F–I).

Discussion

In this study, we have found that compared to the low BMD patients, the high BMD patients have a larger RCSA. Then, the GRA result verified our speculation that these two clinical phenotypes had a potential correlation. Due to the relationship, we screened 4 potential exo-miRNAs as the bridge that communicates between muscle and skeleton using SOM and CatBoost methods based on serum exosome miRNA sequencing. Moreover, in vitro experiments showed that miR-92a-1-5p was the key molecule from the atrophy cell model exosomes, and miR-92a-1-5p can be ingested by receptor cells and regulate their proliferation, apoptosis, and osteogenesis ability. To explore the mechanism further, we first selected the most related mRNAs by time-series analysis based on transcriptome sequencing data including 4 time points during the osteogenesis period of osteoblasts. We then merged these mRNAs with the genes that from the pathway of skeletal system morphogenesis that belong to miR-92a-1-5p target genes. Finally, we chose Col1a1 as a target mRNA, and in vitro and in vivo experiments confirmed the miR-92a-1-5p–Col1a1 relation. These results exposed that miR-92a-1-5p secreted from atrophy muscle cells can regulate osteogenesis function target Col1a1 by the transport ship-exosomes.

Skeletons and muscles both develop from the mesoderm and are part of the motor system, thus complementing each other in development, differentiation, maturation, and motor function. As age increases, both bones and muscles undergo irreversible degenerative changes, which can lead to osteoporosis and sarcopenia in severe cases. The biomechanical and biochemical interactions in musculoskeletal units are of great significance in the regulation and maintenance of tissue function: (i) As functional units, muscles and bones adapt to metabolic and mechanical needs in health, and due to the same biomechanical and biochemical connections between these two tissues, they degrade with aging. Frost’s “mechanical stability” theory suggests that bones self-regulate to maintain strain within the physiological window: if greater strain (i.e., physical activity) is required, bones are formed, while lower strain (e.g., inactivity) promotes bone absorption. Therefore, the level of muscle mass will increase or decrease the formation of bones, respectively; , (ii) Biochemical communication between muscles and bones: Skeletal muscles release hundreds of proteins and peptides that can affect bone health. For example, the myocytokine irisin can promote osteoblast formation and reduce osteoclast formation by targeting key signaling proteins. In addition, myostatin, a negative regulator of muscle growth, and interleukin-6 (IL-6) have an impact on bones. At present, studies have also shown that extracellular vesicles derived from muscles play a regulatory role in bones by delivering molecules such as nucleic acids and proteins. −

Almost all tissues and organs in the human body can secrete exosomes, which contain a rich variety of molecules that have potential regulatory effects on the occurrence and development of various diseases. In osteoporosis, exosomes derived from bone marrow mesenchymal stem cells, osteoblasts, macrophages, osteoclasts, and endothelial cells, as well as exosomes derived from body fluids, such as serum and plasma, can regulate the proliferation and differentiation of osteoblasts. Under certain periods and interventions, this regulatory mechanism may be bidirectional, such as miR-135b contained in exosomes derived from bone marrow mesenchymal stem cells miR-148a, miR-299-5p, and other molecules are upregulated during the development and differentiation of osteoblasts, while miR-155 and miR-211 molecules are downregulated. − In addition, we found that previous studies have also explored whether exosomal miRNAs are simultaneously associated with decreased muscle mass and BMD, and may play important roles in the development of muscle and bone. Fulzele et al.’s study found that miR-34a accumulates in mouse skeletal muscle with age, and this phenomenon occurs simultaneously in human skeletal muscle tissue. Overexpression of miR-34a in muscle cells leads to the accumulation of ceramides, triggering the output of miR-34a through vesicles. These vesicles may transmit miR-34a to other tissues such as bones through local or circulatory transmission, leading to bone loss; Sun et al. reviewed the research on miR-214 in bone and muscle tissue and found that miR-214 not only mediates skeletal muscle generation and vascular smooth muscle cell proliferation, migration, and differentiation but also regulates the function of osteoblasts and osteoclasts by targeting specific molecular pathways and the expression of various related genes. It is also correlated with bone diseases such as osteoporosis. Zhang et al.’s study found that knocking out muscle-specific miR-23a clusters in skeletal muscle tissue can inhibit bone remodeling in mice, indicating that muscle-derived miRNAs may participate in bone metabolism regulation through exosomes in muscle–bone interactions.

Current research shows that miR-92a-1-5p plays a key role in regulating the occurrence, development, and metastasis of tumor-related diseases, such as hematological tumors, colorectal cancer, melanoma, breast cancer, renal cancer, and prostate cancer. − It is also associated with nontumor diseases such as female hair loss, the development of myocardial tissue, spermatogenic disorders caused by heat stress, autism, and lactation disorders. − In our study, miR-92a-1-5p plays a crucial role in the muscle–exosome–skeleton system and regulates bone maturation by targeting the hub gene Col1a1. Although some studies showed that the muscle–skeleton system is a bidirectional regulation system in causal analysis, our result showed that muscles seem to have a greater contribution to the development of bones than bones do to muscle. This may require larger sample studies to conduct more in-depth research and discussion of the correlation and causal relationship between the two. Although our study suggests that miR-92a-1-5p serves as an information mediator in exosomes involved in the regulation of muscle–bone interactions and provides evidence, our research still has certain limitations in terms of the process of bone remodeling as it does not take into account the important role of exosomes in osteoclast-mediated bone resorption. Based on previous studies, miRNA molecules mediated by exosomes can also enter osteoclasts, regulate their differentiation and maturation in vitro, and ultimately lead to bone resorption, causing an imbalance between the osteogenic and osteoclast systems. In vivo experiments have shown significant bone loss. In summary, our research suggests that miR-92a-1-5p contributes to the regulation of muscle–bone function, which may provide new perspectives for the prevention, diagnosis, and treatment of osteoporosis.

All data generated or analyzed during this study are included in this published article.

Studies with human subjects: The study was approved by the ethics committee of Jinling Hospital , Medical College, Nanjing University (grant number 2023DZGZR-018), and the study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. Informed consent was obtained from all patients for being included in the study. Studies with animal subjects: The study was approved by the ethics committee of Jingling Hospital, Medical College, Nanjing University (grant number 2023JLHGZRDWLS-00019), and all institutional and national guidelines for the care and use of laboratory animals were followed.

The authors declare no competing financial interest.