LncRNA GABARAPL2 promotes non‑union long bone fractures by suppressing bone mesenchymal stromal cell osteogenesis

Abstract

Background

Fracture healing is a complex biological process involving precisely coordinated phases of inflammation, cartilage formation, vascularization, and bone remodeling. Growing evidence suggests that non-coding RNAs play crucial regulatory roles in bone metabolism.

Objective

This study systematically investigated the role and molecular mechanisms of lncRNA GABARAPL2 (lnc_GABARAPL2) in fracture healing through integrated bioinformatics analysis, clinical validation, and functional experiments, with the goal of identifying novel biomarkers and therapeutic targets for impaired fracture healing.

Methods

Lnc_GABARAPL2, miR-302a-3p, RUNX2 and OCN mRNA expression levels were quantified by RT-qPCR. Osteogenic differentiation was assessed through Western blot (Collagen I) and spectrophotometric ALP activity measurement and alizarin red S-based mineralization assay. In vitro, lnc_GABARAPL2 expression was modulated in hBMSCs via siRNA knockdown and overexpression vectors, with target interaction validated by dual-luciferase reporter assay. Cell proliferation (CCK-8) and apoptosis (flow cytometry) were assessed. Statistical methods included t-tests, ANOVA, and ROC curve analysis.

Results

Lnc_GABARAPL2 expression was significantly elevated in non-union patients and demonstrated excellent predictive performance. Multivariate regression analysis confirmed lnc_GABARAPL2 as an independent predictor of non-union. Functional studies revealed that lnc_GABARAPL2 impaired osteogenic differentiation by suppressing ALP activity and reducing expression of key osteogenic markers RUNX2 and Collagen I, while modulating bone marrow mesenchymal stem cell proliferation and apoptosis through targeting miR-302a-3p.

Conclusion

These findings establish that lnc_GABARAPL2 may as a regulator of fracture healing and a promising predictive biomarker for non-union and targeting miR-302a-3p to modulate bone marrow mesenchymal stem cell functions, providing new insights for the development of RNA-based therapeutic strategies for impaired bone repair.

Introduction

Fracture healing represents a sophisticated regenerative process involving well-orchestrated biological phases, including inflammatory response, chondrogenesis, angiogenesis, and bone remodeling [1]. This complex cascade is tightly regulated by an array of cytokines, growth factors, and signaling pathways [2–4]. During the initial inflammatory phase, macrophages and mesenchymal stem cells are recruited to the fracture site, where they secrete pro-inflammatory cytokines such as TNF-α and IL-6 to initiate the repair cascade [5]. The subsequent chondrogenesis phase witnesses the differentiation of chondrocytes that produce abundant extracellular matrix components, establishing a provisional scaffold for ensuing ossification. The pivotal bone formation phase is characterized by the differentiation of mesenchymal stem cells into functional osteoblasts, a process that critically determines the quality of bone regeneration [6]. Emerging evidence from epigenetic research has highlighted the crucial regulatory roles of non-coding RNAs in bone metabolism [7]. Particularly, the intricate networks formed by long non-coding RNAs (lncRNAs) have been shown to modulate osteogenic differentiation through key signaling pathways [8, 9]. Recent evidence further underscores the broad involvement of non-coding RNAs (e.g., miRNA, siRNA, circRNA) in musculoskeletal conditions, including tendinopathy [10, 11], osteoarthritis [12], rheumatoid arthritis [13], and osteoporosis [14, 15], where they regulate extracellular matrix composition, inflammatory responses, and metabolic processes. However, the precise mechanistic roles of these non-coding RNAs at distinct stages of fracture healing, especially their regulatory contributions to the pathogenesis of non-union, remain to be fully elucidated.

LncRNAs constitute a class of non-coding transcripts exceeding 200 nucleotides in length. Despite their lack of protein-coding potential, they participate in sophisticated gene expression regulation through multiple mechanisms [16]. At the molecular level, lncRNAs can function as competitive endogenous RNAs to sequester miRNAs, modify chromatin architecture, influence mRNA stability, or directly regulate transcriptional processes [17]. Recent studies have demonstrated that lncRNAs exhibit remarkable tissue specificity and spatiotemporal specificity, with their expression patterns closely associated with cellular differentiation states [18]. In the skeletal system, several lncRNAs have been implicated in bone metabolism regulation. For instance, H19 enhances osteogenic differentiation by activating the Wnt/β-catenin signaling pathway, and its expression level shows a negative correlation with the severity of osteoporosis [19]; MALAT1 modulates bone homeostasis through the β-catenin-OPG/Jagged1 axis and exhibits abnormal expression patterns in senile fractures [20]; whereas TUG1 promotes fracture repair via the miR-22-5p/Anxa8 pathway [21], with its overexpression demonstrated to accelerate the healing process in murine fracture models. Intriguingly, these lncRNAs often form complex interaction networks that collectively regulate downstream target genes to influence osteogenic differentiation. Nevertheless, a comprehensive understanding of lncRNA dynamics during fracture healing, particularly their specific roles in non-union development, requires further investigation.

Research suggests that lnc_GABARAPL2 may be involved in inflammatory responses via the neuroactive ligand-receptor interaction signaling pathway, indicating its potential as both a biomarker and therapeutic target for dry eye disease (DED) [22]. This study aims to systematically elucidate the regulatory role and molecular mechanisms of lnc_GABARAPL2 in fracture healing through bioinformatics analysis, clinical sample analysis, and cellular experiments, which may provide theoretical foundations and potential therapeutic targets for early prediction of non-union and targeted treatment of non-union.

Materials and methods

Data acquisition and differential expression analysis

The GSE213891 dataset was downloaded from the GEO database (https://www.ncbi.nlm.nih.gov/geo/), which provides transcriptome sequencing data of lncRNA and mRNA expression profiles in bone nonunion tissues compared with normal fracture healing samples. The log2FPKM matrix was subsequently extracted for further analysis. Differential expression analysis was performed using the R package limma with thresholds of|log₂FC| >1, p < 0.05, and RNA annotation containing “lnc”. Results were visualized via a volcano plot and heat map. Potential miRNA binding partners of the candidate lncRNAs were predicted using the lncRNASNP2 online tool (https://guolab.wchscu.cn/lncRNASNP/#!/) to explore their functional mechanisms.

Clinical sample collection

This study identified 45 patients with long bone fracture non-union from the medical records of Shanxi Bethune Hospital, and collected 45 matched controls with healed fractures. Serum samples were collected in an outpatient clinical setting before surgery and clinical data were documented for all participants, encompassing basic information such as age and gender, as well as health status, injury severity, and treatment regimens. As shown in Table 1, participants were aged between 20 and 50 years, with generally comparable baseline characteristics between the two groups. The study protocol was approved by the Institutional Ethics Committee of Shanxi Bethune Hospital, and written informed consent was obtained from all participants.

Table 1.

Clinical parameters

	Union (n = 45)	non-union (n = 45)	P value
Clinical parameters
General
Age (yeas)	34.16 ± 12.82	36.96 ± 12.44	0.296
Gender (female/ male)	15/30	16/29	0.824
General health status
American College of Anaesthesiologists (ASA) classification system	1.29 ± 0.55	1.36 ± 0.57	0.505
Injury severity
Charlson comorbidity index	0.58 ± 0.69	0.62 ± 0.78	0.846
Injury Severity Score (ISS)	12.51 ± 7.38	11.09 ± 5.92	0.462
Glasgow coma scale (GCS)	13.91 ± 1.52	14.00 ± 1.57	0.530
Abbreviated Injury Scale (AIS) head	0.89 ± 0.83	1.04 ± 0.85	0.382
Abbreviated Injury Scale (AIS) thorax	0.51 ± 0.82	0.51 ± 0.92	0.850
Concomitant injuries (yes/no)	20/25	22/23	0.673
Open/closed fracture	14/31	10/35	0.340
Clinical course
Duration of hospital stay (days)	17.84 ± 10.62	17.91 ± 12.14	0.856
In-hospital complications (yes/no)	10/35	11/34	0.803
ETC vs. DCO	24/21	14/31	0.033*

*Statistical significance, p < 0.05

Peripheral venous blood (1.5 mL) was collected in the morning after an 8-hour fast using sterile serum separator tubes (SST). After clotting at room temperature for 30 min, samples were centrifuged at 3000 rpm for 15 min (4 °C) to isolate serum. The supernatant was aliquoted into pre-chilled RNase/DNase-free tubes to prevent freeze-thaw cycles, flash-frozen in liquid nitrogen, and stored at -80 °C with continuous temperature monitoring. All processing steps were completed within 2 h post-collection to minimize pre-analytical variability. Quality control included documenting freeze-thaw cycles and excluding improperly stored samples.

Cell culture

Human bone marrow-derived mesenchymal stem cells (hBMSCs, Cyagen Biosciences, China) were thawed and cultured in complete medium for adult human bone marrow mesenchymal stem cells (Cyagen, China) supplemented with 10% fetal bovine serum (FBS, Gibco), 1% penicillin-streptomycin, and 2 mM L-glutamine. The cells were maintained at 37 °C with 5% CO₂ and subcultured every 3 days using a trypsin-based dissociation reagent. Cells at passages 3–5 in the logarithmic growth phase were utilized for subsequent experiments.

Cell transfection

To investigate the function of lnc_GABARAPL2, RNA knockdown and overexpression experiments were performed using siRNA and plasmid vectors, respectively. hBMSCs were cultured in 6-well plates until 70%-80% confluency and transfected with either siRNA targeting lnc_GABARAPL2 (lnc_GABARAPL2 siRNA) or negative control siRNA (NC siRNA). For overexpression, the lnc_GABARAPL2 overexpression vector (lnc_GABARAPL2 pcDNA) and empty vector control (NC pcDNA) were used. Additionally, cells were transfected with miR-302a-3p mimics, mimics negative control (mimics NC), miR-302a-3p inhibitor, and inhibitor negative control (inhibitor NC). Transfection was conducted using Lipofectamine 3000 (Invitrogen) following the manufacturer’s protocol. Cells were harvested 48 h post-transfection for subsequent functional assays and RNA extraction.

Osteogenic induction

Transfected cells (including knockdown, overexpression, and corresponding control groups) were seeded in 6-well plates at a density of 5 × 10⁴ cells/well. When cell confluence reached 60%-70%, the medium was replaced with osteogenic induction medium (Cyagen, China), consisting of basal medium supplemented with 10% FBS, 0.1 µM dexamethasone, 10 mM β-glycerophosphate, 50 µM ascorbic acid, and 1% penicillin-streptomycin. The induction medium was refreshed every 3 days for 14 days, followed by subsequent analysis.

Real time quantitative PCR (RT-qPCR)

Total RNA was isolated employing Trizol Reagent (Thermo Fisher Scientific). Subsequently, reverse transcription was conducted to generate first-strand cDNA using a PrimeScript™ RT reagent kit (TaKaRa, Japan). Using cDNA as template, RT-qPCR was performed on an ABI 7900HT Real-Time PCR System (Applied Biosystems, USA) to detect the expression of lnc_GABARAPL2 and miR-302a-3p in serum from patients with bone union and non-union, and to quantify RUNX family transcription factor 2 (RUNX2) and osteocalcin (OCN) mRNA. All reactions were run in quadruplicate. Gene expression levels were calculated using the 2^−ΔΔCt method [23], with GAPDH as the endogenous control for RUNX2 or lnc_GABARAPL2 normalization.

Western blotting

After 14 days of osteogenic induction, cells from each group were collected. Total proteins were extracted using RIPA lysis buffer containing 1% protease inhibitor cocktail. Protein samples were separated by electrophoresis on 12% polyacrylamide gels and subsequently transferred to PVDF membranes (Millipore). The membranes were blocked with 5% skim milk for 1 h at room temperature, followed by incubation with primary antibodies against Collagen I and GAPDH overnight at 4 °C. After washing three times with TBST, the membranes were incubated with horseradish peroxidase (HRP) conjugated secondary antibodies for 1 h at room temperature. Protein bands were visualized using an electrochemiluminescence detection kit (Amersham).

Alkaline phosphatase (ALP) activity assay

Following a 14-day osteogenic induction, culture media were carefully removed, and cell monolayers were rinsed twice with sterile phosphate-buffered saline. Cellular lysis was performed using 1% NP-40 detergent solution (Nacalai Tesque), followed by centrifugation at 12,000×g for 3 min at 4 °C to obtain clarified supernatants. ALP enzymatic activity was determined spectrophotometrically at 405 nm wavelength after 30-minute incubation using a commercial alkaline phosphatase detection kit (Takara Bio, Shiga, Japan) with measurements conducted in quartz cuvettes using a UVmax Colorimeter (Molecular Devices), strictly following the supplier’s recommended procedures. ALP enzymatic activity was quantified and normalized to total protein content, with results expressed as optical density (OD) units per milligram of cellular protein.

Mineralization assay

After transfection and osteogenic induction, cells were seeded in 6-well plates and then fixed in 10% neutral-buffered formalin and stained with 2% alizarin red S (Sigma-Aldrich, USA). Calcium deposition was quantified by destaining with 10% cetylpyridinium chloride (Sigma-Aldrich). Absorbance of the eluted dye was measured at 550 nm.

Dual-luciferase reporter assay

To validate the lnc_GABARAPL2-miR-302a-3p interaction, the wild-type (WT) or mutant (MUT) binding sequences of lnc_GABARAPL2 (predicted by lncRNASNP2) were cloned into pmirGLO vectors (Promega). hBMSCs cultured in 24-well plates at 70%-80% confluency were transfected with the following combinations: pmirGLO-WT or pmirGLO-MUT reporter vectors together with miR-302a-3p mimics, mimics negative control (mimics NC), miR-302a-3p inhibitor, or inhibitor negative control (inhibitor NC). After 48 h, luciferase activity was measured using a luciferase detection system (Promega). Data were normalized and presented as Relative Luciferase Activity, with four independent replicates performed.

Cell proliferation assay by CCK-8

To systematically evaluate the regulatory effect of lnc_GABARAPL2 on hBMSC proliferation, cells were divided into four transfection groups: (1) lnc_GABARAPL2 siRNA, (2) negative control siRNA (NC siRNA), (3) lnc_GABARAPL2 siRNA + miR-302a-3p inhibitor, and (4) lnc_GABARAPL2 siRNA + inhibitor negative control (lnc siRNA + NC inhibitor). Cells were seeded in 96-well plates at 5 × 10³ cells/well and assessed at 24, 48, and 72 h post-transfection. After adding 10 µL CCK-8 reagent (Dojindo) per well and incubating for 2 h at 37 °C, absorbance at 450 nm was measured using a microplate reader (BioTek). Quadruplicate technical replicates were performed for each group, with four independent biological replicates conducted.

Flow cytometric analysis of apoptosis

To systematically evaluate the regulatory effect of lnc_GABARAPL2 on hBMSC apoptosis, the same experimental groups as in the cell viability assay were used. After 48 h of transfection, cells were harvested and stained using Annexin V-FITC/PI Apoptosis Detection Kit (BD Biosciences), followed by analysis with a flow cytometer (BD FACSCanto II). Quadruplicate technical replicates were performed for each group in four independent experiments.

Statistical analysis

Continuous variables were expressed as mean ± SD and compared by two-tailed Student’s t-test or ANOVA. Categorical variables were described as frequencies and analyzed by chi-square. ROC curves were generated using GraphPad Prism (v10.1) to evaluate the predictive value of lnc_GABARAPL2 for non-union of bone after fracture. Multivariate logistic regression (SPSS v27) identified independent predictors of non-union, with statistical significance set at p < 0.05.

Results

Identification of differentially expressed lnc_GABARAPL2

To identify lncRNAs potentially associated with non-union long bone fractures, differential expression analysis was performed using the log2FPKM matrix from the GSE213891 dataset (available in GEO. Differentially expressed lncRNAs and genes (DEGs) were identified with a p < 0.05,|log₂FC| >1, and gene annotation containing “lnc” as thresholds. Analysis revealed three significantly dysregulated lncRNAs: lnc_AL391421.1_3:1, lnc_GABARAPL2_1:1, and lnc_VCL_1:1 (Fig. 1A). Notably, lnc_GABARAPL2 demonstrated the most substantial upregulation (log₂FC = 1.55), which was further corroborated by hierarchical clustering analysis/ heat map (Fig. 1B).

Fig. 1.

Identification of differentially expressed lnc_GABARAPL2. (A) Volcano plot and (B) heatmap of differentially expressed lncRNAs in GSE213891 dataset (|log₂FC| >1 and p < 0.05)

Analysis of demographic and clinical parameters

The analysis of baseline characteristics revealed no statistically significant differences between the union and non-union groups in terms of age, gender distribution, general health status (American College of Anaesthesiologists (ASA) classification system), or injury severity (Injury Severity Score (ISS), Glasgow coma scale (GCS), and Abbreviated Injury Scale (AIS) scores) (all p > 0.05). Similarly, fracture characteristics (open/closed type, concomitant injuries) and clinical outcomes (hospital stay duration, complication rates) showed no significant intergroup differences (p > 0.05). The treatment strategy (early total care, ETC vs. damage control orthopaedics, DCO) was associated with fracture healing status (p = 0.033; Table 1).

Predictive potential of lnc_GABARAPL2 for non-union

To evaluate the clinical potential of lnc_GABARAPL2 for non-union prediction, we measured its expression levels in the serum of fracture patients with bone either union or non-union. The results demonstrated that lnc_GABARAPL2 expression was significantly higher in non-union patients compared to union patients (p < 0.001, Fig. 2A). Furthermore, ROC curve analysis indicated that serum lnc_GABARAPL2 exhibited high predictive value for non-union occurrence, with an area under the curve (AUC) of 0.9220 (Fig. 2B). The multivariate logistic regression analysis, after adjusting for age, gender, ASA classification, and other potential confounders, identified lnc_GABARAPL2 expression as an independent predictor of nonunion (OR = 11.661, 95%CI: 3.794–35.840, p < 0.001; Table 2). These findings suggest that lnc_GABARAPL2 may serve as a potential predictive biomarker for non-union bone fracture.

Fig. 2.

Predictive potential of lnc_GABARAPL2 for non-union after fracture. (A) Expression of lnc_GABARAPL2 in serum from bone union or non-union individuals at fracture. (B) ROC curve of serum lnc_GABARAPL2 for predicting the occurrence of bone non-union. ***p < 0.001

Table 2.

Multivariate regression analysis referring to nonunion analysing lncGABARAPL2, age, gender, ASA, CCI, ISS, GCS, aishead, aisthorax, concomitant injuries, open/closed fracture, duration of hospital stay and in-hospital complications and ETC vs. DCO as potential predictors

Predictor	Odds ratio (OR)	95% confidence interval (95%-CI)	p value
lncGABARAPL2	11.661	3.794–35.840	< 0.001*
Age	1.296	0.455–3.689	0.628
Gender (male)	1.095	0.342–3.508	0.879
ASA	1.181	0.123–11.341	0.885
CCI	0.513	0.041–6.335	0.602
ISS	1.275	0.445–3.653	0.651
GCS	2.008	0.556–7.247	0.287
AIShead	1.739	0.538–5.618	0.355
AISthorax	0.896	0.135–5.955	0.909
Concomitant injuries	0.753	0.254–2.232	0.609
Open/closed fracture	1.022	0.305–3.423	0.972
Duration of hospital stay (days)	0.457	0.145–1.441	0.181
In-hospital complications	0.983	0.281–3.443	0.979
ETC vs. DCO	0.457	0.149–1.396	0.169

*** Statistical significance, p < 0.001

Lnc_GABARAPL2 modulates osteogenic differentiation markers

Transfection experiments confirmed that lnc_GABARAPL2 siRNA significantly reduced the expression level of lnc_GABARAPL2 (p < 0.001), while lnc_GABARAPL2 pcDNA markedly increased its expression (p < 0.001, Fig. 3A). ALP, a crucial biomarker for assessing early-stage osteogenic differentiation, demonstrated activity levels that strongly correlated with the initial differentiation status of osteoblasts. Experimental data revealed that suppression of lnc_GABARAPL2 expression significantly enhanced ALP activity (p < 0.001, Fig. 3B), while its overexpression substantially inhibited ALP activity (p < 0.001). Analysis of osteogenesis-related genes showed that lnc_GABARAPL2 knockdown significantly increased expression levels of both the RUNX2 and Collagen I (p < 0.001, Fig. 3C-D), whereas lnc_GABARAPL2 overexpression produced significant inhibitory effects on these molecular markers (p < 0.001). Moreover, an increase was observed in late-stage marker (OCN) after lnc_GABARAPL2 knockdown, while a reduction after lnc_GABARAPL2 overexpression (p < 0.001, Fig. 3E). Also, knockdown of lnc_GABARAPL2 led to enhanced activity of cell mineralization, while lnc_GABARAPL2 overexpression had opposite effects (p < 0.001, Fig. 3F).

Fig. 3.

Lnc_GABARAPL2 suppressed osteogenesis of hBMSCs. (A) Relative expression level of lnc_GABARAPL2 after transfection. (B) Activity of ALP (OD/mg). (C) Relative expression level of RUNX2 mRNA (D) Relative expression level of Collagen I protein. (E) Relative expression level of OCN. (F) Alizarin Red S staining absorbance at 550 nm. ***p < 0.001

Lnc_GABARAPL2 promotes non-union by targeting miR-302a-3p

Analysis of clinical samples revealed significantly lower miR-302a-3p levels in non-union patients compared to the union group (Fig. 4A), with a strong negative correlation to lnc_GABARAPL2 expression (r = -0.798, p < 0.001; Fig. 4B). The lncRNASNP2 database predicted a complementary binding sequence between lnc_GABARAPL2 and miR-302a-3p (Fig. 4C). Dual-luciferase assays demonstrated that miR-302a-3p mimics markedly suppressed the activity of wild-type lnc_GABARAPL2 reporters (p < 0.001), while mutant constructs remained unaffected (Fig. 4D-E). RT-qPCR experiments further confirmed that lnc_GABARAPL2 overexpression significantly reduced miR-302a-3p levels (p < 0.001), whereas siRNA-mediated knockdown of lnc_GABARAPL2 upregulated miR-302a-3p (Fig. 4F), confirming a direct binding and repressive relationship.

Fig. 4.

LncRNA GABARAPL2 directly targets miR-302a-3p. (A) Relative expression level of miR-302a-3p in union and non-union fractures. (B) Negative correlation between lnc_GABARAPL2 and miR-302a-3p expression. (C) Predicted binding sequence between lnc_GABARAPL2 and miR-302a-3p. (D-E) Dual-luciferase reporter assay. (F) Overexpression of lnc_GABARAPL2 suppressed miR-302a-3p levels, while siRNA of lnc_GABARAPL 2 increased miR-302a-3p expression. ***p < 0.001

Lnc_GABARAPL2/miR-302a-3p axis modulates hBMSC proliferation and apoptosis

CCK-8 assays showed that lnc_GABARAPL2 knockdown enhanced hBMSC proliferation, an effect reversed by co-transfection with miR-302a-3p inhibitor (Fig. 5A). Flow cytometry revealed that lnc_GABARAPL2 silencing significantly reduced apoptosis (p < 0.001), which was rescued upon miR-302a-3p inhibition (p < 0.05; Fig. 5B). These findings suggest lnc_GABARAPL2 suppresses hBMSC proliferation and promotes apoptosis via miR-302a-3p inhibition, contributing to impaired fracture healing.

Fig. 5.

Functional effects of lnc_GABARAPL2/miR-302a-3p axis on hBMSCs. (A) CCK-8 assay showing cell viability. (B) Apoptosis rate of cells detected by flow cytometry. #p < 0.05, ## p < 0.01, ***p < 0.001

Discussion

Through integrated bioinformatics analysis and clinical sample validation, this study further confirms the critical regulatory role of lnc_GABARAPL2 in fracture healing. Compared with previous studies, our work represents the systematic evaluation of its predictive value in serum samples from nonunion patients, demonstrating excellent predictive performance (AUC = 0.922). Multivariate logistic regression analysis revealed that lnc_GABARAPL2 expression serves as an independent predictor of fracture nonunion (OR = 11.661). Clinical observations indicate the biological relevance of lnc_GABARAPL2 with bone non-union, showing significant correlations between lnc_GABARAPL2 expression patterns and fracture healing outcomes.

The regulatory network of lnc_GABARAPL2 uncovered in this study exhibits distinct characteristics compared to canonical lncRNA mechanisms. Unlike established pathways involving growth factor activation (e.g., ENST00000563492/VEGF axis [24]), Wnt signaling modulation (e.g., H19/β-catenin axis) [25, 26], or bone homeostasis maintenance (e.g., Malat1/OPG-Jagged1 axis [27]), lnc_GABARAPL2 exerts its negative regulatory effects through suppression of ALP activity and RUNX2/Collagen I expression in hBMSCs. Current evidence for fracture healing interventions is limited: Pharmacological agents (e.g., bisphosphonates, statins) largely fail to improve bony healing [28], while physical modalities like intermittent pneumatic compression (IPC) and low-intensity pulsed ultrasound (LIPUS) show promise but require robust clinical validation [29, 30].

Our findings align with emerging reports on RNA-mediated fracture pathology: LINC00339 and CASC11 impair healing by sponging miR-16-5p and miR-150-3p respectively [31, 32], while miR-1271-5p delays pilon fracture repair via ZBTB7A suppression [33]. Non-RNA factors like stress-induced hyperglycemia parameters also predict delayed union [34], and agents such as Jintiange may improve osteoporotic fracture healing [35]. The current findings contribute new evidence to this field, suggesting that distinct lncRNAs may cooperatively regulate fracture healing through unique molecular mechanisms, although it should be noted that the preoperative timing of serum sampling, while predictive, necessitates future validation within narrower post-fracture time windows due to dynamic molecular changes.

This study provides new perspectives for the clinical prediction of non-union and treatment of fracture healing. Future large-scale clinical studies should establish threshold levels of lnc_GABARAPL2 correlating with healing outcomes, enabling the development of rapid detection kits for early identification and intervention in high-risk patients. Furthermore, systematic mapping of interaction networks between lnc_GABARAPL2 and other non-coding RNAs (e.g., circular RNAs) during bone repair may uncover critical synergistic regulatory mechanisms. Compared to single-target therapies, combinatorial interventions targeting multiple nodal molecules within this regulatory network could substantially enhance therapeutic efficacy. These research directions hold promise for establishing precision medicine frameworks in fracture management through integrated theoretical and technological advancements.

Conclusion

This study evaluated the predictive value of lnc_GABARAPL2 for bone non-union, confirming its high accuracy as an independent predictor (AUC = 0.922). We found that lnc_GABARAPL2 promotes non-union by suppressing osteogenic markers (e.g., ALP, RUNX2, and Collagen I) and targeting miR-302a-3p to modulate human bone marrow mesenchymal stem cell proliferation and apoptosis. These findings provide theoretical basis for developing lnc_GABARAPL2-based early diagnostic tools and precise therapeutic strategies for bone non-union.

Author contributions

Study concept and design: H.R. D., and W. L.; analysis and interpretation of data: H.R. D., W.L. Z., and Z.Y. L.; drafting of the manuscript: H.R. D.; critical revision of the manuscript for important intellectual content: W. L.; statistical analysis: W.L. Z., and Z.Y. L.

Funding

No funding was received to assist with the preparation of this manuscript.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Institutional Ethics Committee of Shanxi Bethune Hospital, and written informed consent was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.