Potentials of BMSCs for regulating osteogenic–vascular–neural–lymphatic coupling in bone regeneration Authors: Nan Zhang, Yuwen Luo, Peng Luo, Jiarui Cao, Yuning Cheng, Mengyao Kang, Jianping Mao, Jing-Jun Nie, Da-Fu Chen Published: 2026 ABSTRACT BackgroundBone regeneration assisted by synthetic bone substitutes largely depends on the integration of the vascular, neural, and lymphatic systems in the bone. Bone marrow mesenchymal stem cells (BMSCs) are the key cells for this process. However, their role in regulating the integration has not been fully characterized.MethodsHuman BMSCs (hBMSCs) were treated with osteogenic induction and collected from 0 to 504 h for bulk RNA sequencing (RNA-Seq). Differentially expressed genes (DEGs) were identified and Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) and Time-Series Transcriptomic Trend Analysis were used to comprehensively analyze the possible pathways and functions associated with these DEGs. Weighted Gene Co-expression Network Analysis (WGCNA) was constructed to identify the modules and hub genes of the process. Quantitative real-time polymerase chain reaction (qRT-PCR) and enzyme-linked immunosorbent assay (ELISA) were performed to validate the expression of key genes identified by RNA-Seq.ResultsTime-series analysis of the hBMSCs transcriptome suggested a dynamic expression trajectory during osteogenic differentiation, which was characterized by four functional patterns: the initial adaptation stage (1–24 h), the proliferation activation stage (24–72 h), the differentiation regulation stage (72–336 h) and the remodeling stability stage (336–504 h). Moreover, 72 h was suggested as a potential key time point in the osteogenic–vascular–neural–lymphatic coupling process based on transcriptomic analysis, with typical activation of BMP, vascular endothelial growth factor (VEGF) and PPAR signaling pathways. Four modules and closely related hub genes such as growth differentiation factor 5 (GDF5), matrix Gla protein (MGP) and pregnancy-associated plasma protein A2 (PAPPA2), whose expressions were validated by qRT-PCR and ELISA were also identified and highlighted.ConclusionsOur study revealed the temporal trends of angiogenesis, lymphangiogenesis, and neurogenesis during BMSCs osteogenic differentiation, which not only supplemented the transcriptional regulation in bone regeneration, but also provided a theoretical basis for the design of synthetic bone substitutes.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12938-026-01594-7. Background Bone defects, particularly critical-sized bone defects, pose significant challenges in clinical practice due to their difficulties in self-healing [1]. Nowadays, a large number of treatment strategies have been developed to address this issue, ranging from traditional methods such as autologous bone graft to the use of xenografts and bioactive scaffolds [2]. However, these methods are constrained by the limitations of the donor area [3] and the risks of immune rejection and incomplete integration [4]. In response to these problems, synthetic bone substitutes have been widely explored, and tissue-engineered bone with stem cells has also experienced significant development [5]. Crucially, the integration of bone marrow mesenchymal stem cells with these biomaterials has emerged as a cornerstone of advanced tissue engineering strategies. Despite these advancements, achieving functional bone regeneration remains a critical challenge because the existing treatment strategies, even those incorporating stem cells, often fail to sustain and orchestrate the complex microenvironmental signals required for complete tissue remodeling and critical-sized defect healing. Regeneration of bone defect is a complicated biological process that coordinates osteogenesis, angiogenesis, and neurogenesis [6]. Moreover, the discovery of lymph in bone tissue also confirmed the necessity of lymphangiogenesis in bone regeneration [7]. Current strategies primarily focus on osteogenesis, often overlooking the critical roles of vascularization, lymphangiogenesis, and neurogenesis in bone regeneration [8–10]. Some studies may have explored vascular regeneration during bone healing [11, 12], and some have even investigated neurovascularized bone regeneration [13]. However, they are often constrained to observe the phenomenon of certain tissue repair at a single time point, neglecting the dynamic changes in vascularization, lymphangiogenesis, and neurogenesis during osteogenesis, as well as their interactions. Moreover, most research has focused on extrinsic factors, such as biomaterial design or growth factor supplementation, and neglected the regulation of bone mesenchymal stem cells (BMSCs) themselves to the defect microenvironment. As the key progenitor cells in bone repair, BMSCs not only differentiate into osteoblasts but also act as crucial regulators of the defect microenvironment. They can secrete exosomes which have been demonstrated to promote angiogenesis-mediated bone regeneration [14], and have beneficial effects on nerve healing [15]. However, how BMSCs reshape the defect microenvironment during the osteogenic differentiation process remains unknown. Specifically, how BMSCs’ commitment to the osteogenic lineage is temporally coupled with the regulation of vascular, lymphatic, and neural remodeling signals remains unclear, and the potential underlying mechanisms warrant further investigation. Bulk RNA sequencing (RNA-Seq) is a technology based on high-throughput sequencing platforms that enables comprehensive analysis of RNA expression in cells or tissues. This method can detect dynamic changes and quantify all transcripts in cells, thereby revealing gene expression regulatory mechanisms, biological processes, and disease-related molecular characteristics [16]. In this work, we studied the time-series transcriptomic changes of human bone mesenchymal stem cells (hBMSCs) during osteogenic differentiation after osteogenic induction for different times with RNA-Seq. This approach enables us to systematically explore key regulatory genes, signaling pathways, and potential osteogenic–vascular–neural–lymphatic coupling in hBMSCs during osteogenic differentiation, offering valuable insights into the multifaceted role of hBMSCs in these processes. In this work, through high-throughput sequencing expression profiling, we identified differentially expressed genes (DEGs) in hBMSCs at different time points after osteogenic induction. By integrating DEGs data, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis, transcriptomic temporal trend analysis and Weighted Gene Co-expression Network Analysis (WGCNA) were carried out to identify the key regulatory genes and signaling pathway changes of hBMSCs involved in osteogenesis, angiogenesis, lymphangiogenesis, and neurogenesis. The expression of hub genes was also verified by quantitative real-time polymerase chain reaction (qRT-PCR) and enzyme-linked immunosorbent assay (ELISA) to confirm their transcriptional and translational consistency. In addition, we further explored the critical time points for regulating these processes during hBMSCs osteogenic differentiation. Our findings supplemented the transcriptional regulation of BMSCs in osteogenic–vascular–neural–lymphatic coupling during bone regeneration, and provided a theoretical basis for the design of synthetic bone substitutes. Results Temporal dynamics of gene transcription To investigate the temporal dynamics of gene transcription during the hBMSC osteogenic differentiation, we performed transcriptome analysis under osteogenic conditions, collecting data at ten distinct time points: 0 h, 1 h, 2 h, 4 h, 12 h, 24 h, 72 h (3 days), 168 h (7 days), 336 h (14 days) and 504 h (21 days) (Fig. 1A). Each time point included three biological replicates. Principal component analysis (PCA) showed that the samples at different time points were gradually distributed along the principal component axis, and different time groups formed an obvious clustering trend, indicating that there was no anomalous data far from the whole population (Supplementary Fig. 3). In addition, both PCA plot and the Pearson correlation coefficient heatmap (Supplementary Fig. 4) demonstrated high consistency among biological replicates.Fig. 1Experiment overview and data quality. A Schematic shows the culturing and sequencing time of hBMSC. Time 0 was defined as the point when the osteogenic induction medium was added to the 6-well plates. B Bar plot illustrates the total number of differentially expressed genes (DEGs) in hBMSC at different time points after osteogenic induction, p < 0.05. C, D Venn diagram showing common and unique genes in different groups (N = 3) Experiment overview and data quality. A Schematic shows the culturing and sequencing time of hBMSC. Time 0 was defined as the point when the osteogenic induction medium was added to the 6-well plates. B Bar plot illustrates the total number of differentially expressed genes (DEGs) in hBMSC at different time points after osteogenic induction, p < 0.05. C, D Venn diagram showing common and unique genes in different groups (N = 3) We used DESeq2 to analyze differentially expressed genes (DEGs) and a total of 6295 DEGs were identified. The bar plot demonstrated the increasing number of DEGs over time, with a significant rise in both up-regulated and down-regulated genes after 24 h (914 down- and 541 up-regulated), peaking at 504 h (2337 down- and 2580 up-regulated), indicating a progressive transcriptional shift (Fig. 1B). Figure 1C indicates that the number of differentially expressed genes (DEGs) was relatively low in the initial 12 h, with limited overlap. The Venn diagram of Fig. 1D demonstrated the emergence of 739 consistently expressed DEGs across late-stage time points. These could represent the core genes involved in osteogenesis, while the most unique genes were detected at 504 h, reflecting the final phases of osteogenesis, extracellular matrix maturation, and mineralization. According to the correlation heatmap and PCA (Supplementary Fig. 3 and Supplementary Fig. 4), the expression patterns of hBMSCs at 0, 1, 2, 4 and 12 h were highly correlated, and the gene expression patterns were consistent. At 24 h, gene expression patterns began to show significant deviation, and the samples from 72 to 168 h formed a new distribution trend. The results indicated that a critical shift in gene expression occurred after 24 h, and 24 h might be a key turning point in osteogenic induction (Supplementary Fig. 4). After 168 h, the samples were less correlated with previous stages in some regions, suggesting that new regulatory mechanisms might begin to dominate. Moreover, the 504 h sample completely deviated from the early and middle samples, and formed independent clusters, indicating a complete shift in gene expression. Thus, we could simply divide the expression pattern of hBMSCs during osteogenic differentiation into 4 stages. The initial stage was from 1 to 24 h and little change in gene expression took place. In the proliferation activation stage (24–72 h), the pattern of gene expression started to show observable changes. Gene expressions were further adjusted in the differentiation regulation stage (72–336 h). At the last stage (336–504 h), gene expressions were different from others and tended to be stable. Together, these results revealed the time-dependent transcriptional changes of hBMSCs during osteogenic differentiation, and initially categorized the various stages. GO-BP enrichment analysis results show time-dependent transcriptional changes of hBMSCs during osteogenic differentiation To investigate the associated gene expression patterns and GO terms, we analyzed the expression profiles of the entire transcriptome and generated a heatmap trend GO pathway plot. This transcriptomic trend analysis heatmap illustrated the dynamic gene changes in hBMSCs over time after osteogenic induction, and distinct temporal clusters were highlighted. Specifically, the C1 gene cluster exhibited a gradual increase in expression after 24 h, and reached peaks at the late stage (Fig. 2). Combined with GO-BP analysis, we could find that the genes were primarily enriched in the processes of bone mineralization, lymphocyte proliferation, neuron cellular homeostasis, and blood vessel remodeling, indicating their involvement in sustained osteogenesis and tissue remodeling.Fig. 2Heatmap trend GO pathway plot showing the temporal trends in gene expression of hBMSC during osteogenic differentiation and their corresponding pathways analyzed with GO-BP enrichment, p < 0.05 Heatmap trend GO pathway plot showing the temporal trends in gene expression of hBMSC during osteogenic differentiation and their corresponding pathways analyzed with GO-BP enrichment, p < 0.05 The C2 gene cluster was highly expressed at the early stage and showed transient activation between 4 and 24 h. However, this cluster had no significant correlation with the processes of interest in our study. The C3 genes, which were highly expressed at the early time points (1–12 h) but significantly downregulated after 24 h, were primarily involved in transient signaling pathways such as early osteogenic signaling, angiogenesis initiation, neurodevelopmental responses, and even some negative pathways. Combined with the results of PCA and the correlation heatmap (Supplementary Fig. 3 and Supplementary Fig. 4), we found that the gene expression at the early stage showed similar patterns. Moreover, key pathways regulating bone mineralization, vascular remodeling, lymphocyte proliferation, and neural differentiation showed activation from 24 h after osteogenic induction, which marked the transition of hBMSC function during osteogenic differentiation, from early signaling events to the establishment of tissue remodeling. GO enrichment analysis of DEGs during hBMSC osteogenic differentiation Based on the heatmap GO pathway plot, the gene expression patterns associated with osteogenesis, angiogenesis, lymphangiogenesis and neurogenesis could be classified into two temporal clusters, with 24 h as the demarcation time point. We further analyzed the DEGs with GO enrichment analysis during the osteogenic induction, and found that no clear regulatory trend was observed during 1–12 h, suggesting that early-stage changes might not establish a stable pattern (Supplementary Fig. 5). By focusing on the 24–504 h induction window, we reduced the impact of transient gene fluctuations while capturing the critical regulatory mechanisms driving coordinated bone, vascular, lymphatic, and neural development. As shown in Fig. 3A, the osteogenesis pathway was activated early (24 h), with osteoblast differentiation and matrix deposition peaking during 168–336 h, followed by bone maturation at 504 h. The enriched signaling pathways including Wnt signaling pathway, BMP and MAPK signaling pathway were activated in the early period. Angiogenesis initiated concurrently with osteogenesis at 24 h, reached its peak between 72 and 168 h, and stabilized during 336–504 h (Fig. 3B). Lymphangiogenesis was found to be more delayed compared to angiogenesis. It began at 72 h with the activation of the vascular endothelial growth factor-related signaling pathway, a pathway associated with both lymphangiogenesis and angiogenesis. VEGF signaling pathway could facilitate lymphangiogenesis by activating the PI3K/Akt/mTOR, Ras/MEK/ERK, and Src/FAK/Paxillin pathways, thereby promoting cell proliferation, migration, differentiation, and lumen formation [17]. This lymphangiogenesis became most active during 168–336 h, with the regulation of lymphocyte migration, indicating a role in bone remodeling and immune regulation (Fig. 3C). Neurogenesis followed a similar delayed pattern, with neural differentiation and axon guidance becoming significant after 72 h, peaking during 168–336 h, and stabilizing at 504 h (Fig. 3D). Above all, we could conclude that the 72 h could be the key time point for initiating all processes and the 168–336 h period was the critical time for the entire process.Fig. 3DEGs of hBMSC during osteogenic differentiation analyzed with GO enrichment analysis at different time points, p < 0.05. A Osteogenesis-related pathways and their changes at specific time points. B Angiogenesis-related pathways and their changes at specific time points. C Lymphangiogenesis-related pathways and their changes at specific time points. D Neurogenesis-related pathways and their changes at specific time points DEGs of hBMSC during osteogenic differentiation analyzed with GO enrichment analysis at different time points, p < 0.05. A Osteogenesis-related pathways and their changes at specific time points. B Angiogenesis-related pathways and their changes at specific time points. C Lymphangiogenesis-related pathways and their changes at specific time points. D Neurogenesis-related pathways and their changes at specific time points WGCNA of hBMSC during osteogenic differentiation To identify the modules potentially involved in these processes, we conducted a relevant analysis using WGCNA. Based on 19792 genes with standardized RPKM values >0.3, a total of 11 modules were obtained by WGCNA, containing 27 to 1290 genes in each module. Four modules were found to have strong correlation with osteogenesis, angiogenesis, lymphangiogenesis, and neurogenesis: Module 3 (R = 0.76, P = 0.0009, 44 genes, 72 h), Module 6 (R = 0.64, P = 0.01, 150 genes, 72 h), Module 7 (R = 0.7, P = 0.004, 29 genes, 168 h) and Module 1 (R = 0.74, P = 0.001, 294 genes, 504 h) (Fig. 4).Fig. 4Relationships between modules and hBMSC osteogenic differentiation at different time points. The color scale on the right shows module–trait correlation coefficients from − 1 (blue) to 1 (red), indicating correlations from negative to positive. The numbers in modules represent correlation coefficients and those in brackets represent significance, p < 0.05 Relationships between modules and hBMSC osteogenic differentiation at different time points. The color scale on the right shows module–trait correlation coefficients from − 1 (blue) to 1 (red), indicating correlations from negative to positive. The numbers in modules represent correlation coefficients and those in brackets represent significance, p < 0.05 GO and KEGG pathways in each modules indicated that genes in the modules were involved in different processes. In Module 1, the pathways related to extracellular matrix (ECM) remodeling, bone mineralization, vascular adaptation and ECM–receptor interaction were mainly enriched (ossification, regulation of vasculature development, ECM–receptor interaction). This profile strongly aligned with the final remodeling and stabilization phase of osteogenesis, indicating that the bone tissue structure was dynamically regulated and stabilizing in the late stages. Furthermore, the strong enrichment in pathways like focal adhesion and endothelial cell migration was crucial for vessel integrity and lymphatic lumen formation, suggesting that vascular and lymphangiogenesis maturation, hence coupling completion, occurred during the late stage (Fig. 5A, B). Module 3 was mainly involved in amino acid metabolism, cell adhesion and mTOR signaling (alpha-amino metabolic process, focal adhesion assembly and mTOR signaling pathway), suggesting osteoblast activation and metabolic reprogramming in the early stages (Fig. 5C, D). The activation of mTOR signaling was a central checkpoint for cell growth and is essential for the transition from mesenchymal progenitor to the osteoblast lineage. Moreover, the intense activation of cell adhesion and focal adhesion pathways might provide the necessary anchor points and chemical gradients for the initial invasion and alignment of endothelial cells and neural fibers toward the differentiating BMSCs. Module 6 was associated with neural regulation (regulation of neuron death, glycine, serine and threonine metabolism) (Fig. 5E, F). Module 7 was involved in lipid metabolism, PPAR signal and TGF-β signal (steroid metabolic process, PPAR signaling pathway, TGF-beta signaling pathway), suggesting that the lipid metabolism of the osteogenic cell membrane and the switches in the regulatory patterns during the mid-to-late stages (Fig. 5G, H). Both PPAR and TGF-β have bidirectional regulatory effects on the osteogenic process. PPARγ inhibits osteogenesis but PPARδ might promote it. And TGF-β plays different roles in different stages, promoting in the early stage but inhibiting it in the later stage. However, they both contributed to angiogenesis, lymphangiogenesis, and neurogenesis.Fig. 5Enrichment analysis of representative modules from WGCNA analysis. GO enrichment and top 10 pathways in the KEGG enrichment analysis results for Module 1 (A, B), Module 3 (C, D), Module 6 (E, F), and Module 7 (G, H) Enrichment analysis of representative modules from WGCNA analysis. GO enrichment and top 10 pathways in the KEGG enrichment analysis results for Module 1 (A, B), Module 3 (C, D), Module 6 (E, F), and Module 7 (G, H) Co-expression network analysis was performed to identify the hub or key genes in each module. There were 4 genes (SMOC2, MGP, LTBP2, TSPAN8) out of 50 hub genes in Module 1 (Fig. 6A), 4 genes (FERMT2, SLC7A5, JAG1, VEGFA) out of 45 hub genes in module 3 (Fig. 6B), 12 genes (SLC38A1, PAPPA2, ADM2, RIMS3, TRAK2, TGFB2, RAP1B, NCOA7, SLFN5, COL15A1, P4HA1, ACTA2) out of 50 hub genes in Module 6 (Fig. 6C) and 3 genes (EFNB2, NOTCH3, GDF5) out of 29 hub genes in module 7 (Fig. 6D) that were related to the osteogenesis, angiogenesis, lymphangiogenesis, and neurogenesis processes.Fig. 6The co-expression networks of hub genes from representative modules. A Module 1 includes 50 genes and 630 edges. B Module 3 includes 44 genes and 472 edges. C Module 6 includes 50 genes and 628 edges. D Module 7 includes 29 genes and 203 edges. The dots colored green, red, blue and red–yellow interwoven represent the hub genes related to osteogenesis, angiogenesis, neurogenesis and vascular–lymphatic development The co-expression networks of hub genes from representative modules. A Module 1 includes 50 genes and 630 edges. B Module 3 includes 44 genes and 472 edges. C Module 6 includes 50 genes and 628 edges. D Module 7 includes 29 genes and 203 edges. The dots colored green, red, blue and red–yellow interwoven represent the hub genes related to osteogenesis, angiogenesis, neurogenesis and vascular–lymphatic development The expression of key genes at different time points To validate whether the expression of key genes screened from RNA-Seq at different time points is consistent with the transcriptomic trends, the expressions of key genes related to osteogenesis, angiogenesis, neurogenesis and vascular–lymphatic development were further evaluated by using qRT-PCR. Among the 23 related tested key genes, the expressions of GDF5 and MGP basically conformed to the transcriptomic trends, and PAPPA2 displayed minor discrepancies, but the overall trend remained consistent with its biological role (Fig. 7A). During the BMSCs’ osteogenic differentiation process, the expression of GDF5 increased and peaked at 72 h, then gradually declined during the subsequent differentiation period, indicating that it might play a role in the early stage of proliferation in the osteogenic–vascular–neural–lymphatic coupling. The corresponding GDF5 protein expression determined by ELISA also confirmed the expression trend (Fig. 7B). As for MGP, the transcriptional level gradually increased during osteogenic differentiation, and the corresponding protein level also seemed to show a continuous increase. The expression of PAPPA2 showed some difference. The mRNA expression of PAPPA2 gradually decreased during the osteogenic differentiation and the ELISA results indicated that its protein level showed a transient increase at 72 h, then decreased over time and reached the lowest level at 504 h.Fig. 7Validation of the expression of hub genes and proteins. A Quantitative real-time PCR analyses of GDF5, MGP, PAPPA2. B Enzyme-linked immunosorbent assay (ELISA) analyses of GDF5, MGP, PAPPA2. (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). C Regulatory interaction network of GDF5 axis in osteogenic–vascular–neural coupling. D A schematic regulation network diagram shows the potential regulation ways of GDF5, MGP and PAPPA2 Validation of the expression of hub genes and proteins. A Quantitative real-time PCR analyses of GDF5, MGP, PAPPA2. B Enzyme-linked immunosorbent assay (ELISA) analyses of GDF5, MGP, PAPPA2. (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). C Regulatory interaction network of GDF5 axis in osteogenic–vascular–neural coupling. D A schematic regulation network diagram shows the potential regulation ways of GDF5, MGP and PAPPA2 Construction of regulatory interaction network and system-wide interaction networks To further investigate the potential regulatory relationships underlying osteogenic–vascular–neural–lymphatic coupling, we constructed transcriptional correlation networks centered on GDF5, PAPPA2, and system-wide signaling interactions using the time-series transcriptomic dataset. As shown in Fig. 7C, GDF5 exhibited a strong positive correlation with SMAD1 (R = 0.70). And SMAD1 exhibits strong positive correlations with angiogenic markers VEGFA (R = 0.89) and KDR (R = 0.42), neurogenic markers TUBB3 (R = 0.61) and NRP1 (R = 0.55), as well as osteogenic markers ALPL (R = 0.84), RUNX2 (R = 0.45), and SP7 (R = 0.72). This evidence suggests that GDF5 may contribute to the initiation of the coupling process through the canonical BMP/SMAD1 signaling axis. As for PAPPA2, it demonstrated close associations with multiple components of the IGF signaling pathway (Supplementary Fig. 6). PAPPA2 showed strong positive correlations with PIK3CA (R = 0.76) and PROX1 (R = 0.51), suggesting that PAPPA2 may be involved in osteogenic and lymphatic regulation through IGF-related signaling and PI3K/AKT-associated pathways. Furthermore, system-wide interaction network analysis (Supplementary Fig. 7) revealed extensive transcriptional interactions among BMP, VEGF, IGF, and PPAR-related signaling pathways. SMAD1 exhibited widespread positive correlations with osteogenic genes and angiogenic factors, while VEGFA displayed extensive associations with osteogenic and neural-related markers. In addition, IGF-related signaling components, including IGF1, IGF1R, PIK3CA, and AKT1, formed interconnected regulatory relationships with BMP- and VEGF-associated genes. Notably, PIK3CA showed multiple connections with osteogenic, vascular, and neural markers, suggesting that PI3K/AKT signaling may serve as an integrative regulatory node during the coupling process. Discussion BMSCs are widely applied in constructing synthetic bone substitutes for bone repair due to their self-organizing differentiation and exosome-based regulatory abilities [18–20]. However, the temporal regulation of BMSCs on angiogenesis, lymphangiogenesis, and neurogenesis in the defect area during osteogenic differentiation after implantation was unclear. Thus, hBMSCs under osteogenic induction at different times were prepared and detected by using RNA-seq. The PCA and correlation analysis showed that there were significant differences in the gene expression patterns of hBMSCs at different times after osteogenesis induction (Supplementary Fig. 3 and 4). Osteogenic induction in hBMSCs used to be recognized to be divided into three sequential phases according to phenotypic characteristics: the early proliferation phase (1–168 h), the intermediate differentiation phase (168–336 h), and the late mineralization phase (336–504 h) [21]. However, as shown in Supplementary Fig. 3 and Fig. 2, little changes took place within 1–24 h, and the variations began at 24 h. The 24–504 h period observed a significant activation of genes associated with osteogenesis, angiogenesis, lymphangiogenesis, and neurogenesis. Further analysis indicated that the gene expression profiling suggested the existence of four characteristic transcriptional phases or patterns (Figs. 2, 3 and Supplementary Fig. 5). At the initial adaptation stage (1–24 h), some genes showed significant upward or downward adjustment. The DEGs of hBMSCs screened during the stage of the osteogenic induction were analyzed with GO enrichment analysis, and no clear regulatory trend was observed (Supplementary Fig. 5). As shown in Fig. 3A, Wnt signaling pathway [22], BMP signaling pathway [23], TGF-β signaling pathway [24] and mechanical stimulation response [25] occurred between 24–72 h, indicating the initial activation of osteogenesis. During 72–336 h, enhanced extracellular matrix deposition, angiogenesis, and neural projection were significantly enriched. And bone mineralization, vascular homeostasis, lymphangiogenesis and nerve repair were gradually enhanced until 504 h. This time-dependent trend in gene regulation suggested that the regulation of osteogenesis, angiogenesis, lymphangiogenesis, and nerve regeneration occurred synergistically and played different roles during BMSCs osteogenic differentiation assisted bone regeneration. Thus, based on the characteristics of these stages, we classified them into the initial adaptation stage (1–24 h), characterized by minimal changes and unclear trend; the proliferation activation stage (24–72 h), marked by the initial activation of signaling pathways; the differentiation regulation stage (72–336 h), featuring formation processes and the remodeling stability stage (336–504 h) during which the four processes reached a stable state (Fig. 8). Here, the stability refers to the consolidation of osteogenic cell identity and functional state rather than a reduction in transcriptional activity, as late-stage transcriptional changes were primarily associated with extracellular matrix remodeling and tissue maturation. To further analyze the regulation of osteogenic–vascular–neural–lymphatic coupling in hBMSCs during osteogenic differentiation, we analyzed the transcriptome changes in the last three stages (Figs. 3, 4, 5 and 6), and a total of 6136 DEGs were screened.Fig. 8Graphical summaries of four transcriptional stages and key events Graphical summaries of four transcriptional stages and key events Previous studies have shown that osteogenesis and angiogenesis exhibit a high degree of synergy [26–28]. Our study also proved that the regulation of osteogenesis and angiogenesis in hBMSCs initiated simultaneously within 24 h. The positive regulation of vascular endothelial growth factor production, Wnt signaling pathway and BMP signaling pathway increased together after 24 h, and benefited both the osteogenesis and angiogenesis processes (Fig. 3A, B) [29, 30]. The angiogenesis also peaked at 168 to 336 h, triggering the activation of additional signaling pathways regulating this process, meaning that the regulation of osteogenesis and angiogenesis regulation occurred simultaneously. WGCNA analysis also revealed that a number of osteogenic genes (e.g., TGFB2 [31], ACTA2 [32]) and vascular-related genes (e.g., ADM2 [33], COL15A1 [34]) clustered in the same module (Fig. 6). Furthermore, some genes contributed to both osteogenesis and angiogenesis. For example, PAPPA2 could degrade IGFBP-5 and release IGF-1, which activates the PI3K/Akt and MAPK signaling pathways, regulating osteoblast differentiation and bone remodeling [35, 36]. At the same time, PAPPA2 could serve as a biomarker of angiogenesis and may influence placental and neonatal vascular development through IGF-1 [37]. JAG1, serving as a key regulator of Notch signaling pathway, promotes osteoblast differentiation and regulates blood vessel formation, thereby accelerating bone repair and regeneration [38, 39]. SMOC2 also plays a similar role. These results further confirmed the osteogenesis–angiogenesis coupling induced by hBMSCs during osteogenic differentiation [40, 41]. Traditionally, it was thought that there were no lymphatic structures in bone. In 2023, researchers from the University of Oxford discovered an active network of lymphatic vessels in bone using high-resolution light-sheet imaging technology. Furthermore, they identified key signaling interactions among lymphatic vessels, hematopoietic stem cells, and skeletal stem cells, which played a crucial role in both bone regeneration and hematopoiesis [7]. To date, no studies have specifically investigated the temporal dynamics of lymphatic regeneration during bone regeneration. However, following acute spinal cord injury, the number of LYVE-1⁺ (a lymphatic endothelial cell marker) cells increased in the first week and became more pronounced in the second week, with stable lymphatic vessels being observed. In contrast, during acute kidney injury, lymphatic remodeling persisted even in the late phase (> 30 days) [42]. Therefore, lymphatic regeneration might be an early event that completes rapidly, but is followed by sustained structural modifications over time. In our study, the gene expression involved in the lymphangiogenesis-related signaling pathway began to increase after 72 h and reached its peak at 168 h (Figs. 2 and 3C). The results indicated that the hBMSCs could regulate lymphangiogenesis during osteogenic differentiation, and the regulation was initiated in the early transition stages of osteogenesis. The osteogenic microenvironment may activate lymphangiogenesis during 72–168 h of osteogenesis, corresponding to the key time points of lymphatic regeneration after injury. And some hub genes in WGCNA modules associated with revascularization also contributed to lymphangiogenesis such as EFNB2 [43], ADM2 [33] and VEGFA. Among them, VEGFA was reported to bind to VEGFR-2 and VEGFR-3, synergistically VEGF-C/VEGF-D regulate vascular and lymphatic formation in tumors [44, 45]. However, time-course analysis revealed that, compared to angiogenesis, the initiation of lymphangiogenesis was temporally delayed. This meant that the initiation signals may occur simultaneously but lead to different effects in lymphangiogenesis and angiogenesis, which may be attributed to the different roles of blood vessels and lymphatic vessels in tissue and needs further investigations. The neurogenesis-related pathways were significantly upregulated after 72 h and maintained regulatory effects throughout the later stages of the process (Fig. 3D). Hub gene RIMS3 [46] interacted physically and functionally with presynaptic voltage-dependent Ca2+ channels, which might increase neurotransmitter release, promote synaptic plasticity, and support neuronal growth (Fig. 6C). Hub gene SLC38A1 in module 6, which served as a glutamine/glutamate transporter, could affect nerve signaling and maintain synaptic stability [47]. TRAK2 [48], NCOA7 [49] and SLC7A5 [50] in WGCNA modules also could regulate the neurogenesis via promoting mitochondrial trafficking, affecting synaptic homeostasis, and modulating neuronal survival. Therefore, hBMSCs also participated in regulating the growth, development, and stability of nerves during the osteogenesis process (Fig. 6B, C). To further confirm the expression of the 23 key genes analyzed by WGCNA related to osteogenesis, angiogenesis, neurogenesis and vascular–lymphatic development, qRT-PCR and ELISA were performed. Among them, the GDF5, MGP and PAPPA2 showed similar expression trends with the RNA-Seq results. GDF5 is a secreted ligand–protein from the bone morphogenetic protein (BMP) family and the TGF-β superfamily. It promotes the formation of cartilage and bone, and participates in skeletal development through BMP signaling pathway [51]. GDF5-centered transcriptional network analysis suggest that GDF5 is highly correlated with the transcription factor SMAD1 (R = 0.70), which subsequently orchestrates a multi-system coupling effect by targeting key markers (Fig. 7C). GDF5 influences lymphatic vessel formation by regulating extracellular matrix synthesis and cell differentiation, and it also plays a regulatory role in neural development [52]. We confirmed that the GDF5 protein was highly expressed around 72 h during the osteogenic period, which suggested its importance in regulating the proliferation activation stage by osteogenic–neural–lymphatic coupling. The transcription and protein expression of MGP were also relatively synchronous, peaking at 504 h, suggesting its role in regulating matrix mineralization especially in preventing aberrant calcification during osteogenic differentiation [53]. PAPPA2 mRNA expression declined during osteogenic differentiation, while the protein level peaked at 72 h. It has a relatively long biological half-life once translated and PAPPA2 may be subject to regulated secretion or temporary storage within the extracellular matrix (ECM) or the endoplasmic reticulum/Golgi apparatus [54, 55], which lead to a delayed peak time in PAPPA2 protein levels. Such mRNA–protein discordance may also reflect post-transcriptional and post-translational regulatory mechanisms commonly observed in secreted metalloproteinases [56]. The expression trend further supported the notion that PAPPA2 may support early osteogenesis and angiogenesis [57] (Fig. 7D). In addition, as a modulator of the IGF axis, PAPPA2 may be influenced by feedback regulation associated with IGF signaling, which could further contribute to the observed temporal discrepancy between mRNA and protein expression [58]. Correlation analysis revealed the positive correlation between PAPPA2 and PIK3CA (R = 0.76), a key downstream effector of IGF-1 signaling, is consistent with the hypothesis that PAPPA2 contribute to the functional activation of the IGF pathway. Given that PIK3CA encodes the catalytic subunit of PI3K, this observation further suggests that PAPPA2 may exert its effect through the PI3K/AKT signaling axis downstream of IGF-1 (Supplementary Fig. 6). Our network analysis (Supplementary Fig. 7) suggests that key components of BMP, VEGF, IGF signaling, and PPAR-related pathways are highly interconnected, forming a dense regulatory network characterized by both positive and negative correlations. Within this network, SMAD1 shows strong associations with osteogenic and angiogenic genes, while VEGFA is broadly correlated with osteogenic and neural markers, supporting its role as a downstream effector linking vascularization with multi-lineage differentiation. In addition, IGF-related signaling components exhibit widespread interactions with both BMP- and VEGF-associated genes, with PIK3CA indicating a potential role for the PI3K/AKT pathway as an integrative node. Based on these observations, we propose a preliminary regulatory model in which BMP signaling (GDF5–SMAD1) may act as an upstream regulatory axis, VEGF signaling may function as a downstream effector, and IGF signaling together with PPAR-related pathways may contribute to pathway crosstalk, potentially via the PI3K/AKT pathway. The coexistence of both positive and negative correlations further suggests that this coupling process is governed by a dynamic and tightly regulated network rather than a simple linear cascade. Further experimental studies will be required to validate the direct regulatory relationships among these pathways. Currently, most studies focus on the relationship between osteogenesis and angiogenesis. However, our study further reveals the dynamic changes in hBMSC-mediated regulation of lymphangiogenesis and neurogenesis during osteogenesis. Recent studies have demonstrated that lymphatic vessels actively participate in bone regeneration by regulating immune cell trafficking, inflammatory resolution, and microenvironmental homeostasis, rather than functioning solely as passive drainage structures. In addition, lymphangiocrine signaling may contribute to progenitor cell recruitment and synergistically coordinate osteogenesis and angiogenesis during tissue regeneration [59, 60]. We also found that 72 h may be the key time for the osteogenic–vascular–neural–lymphatic coupling. Interestingly, the delayed activation pattern of lymphangiogenesis observed in our study may reflect its predominant regulatory role during the later stages of bone remodeling and regenerative microenvironment stabilization. We validated the early upregulation of the skeletal development-related gene GDF5 and the angiogenesis marker PAPPA2. Moreover, the expression of the bone mineralization-related gene MGP increased progressively over time, peaking at 504 h. These time-dependent molecular changes further support the identified critical time windows. The dynamic change patterns of angiogenesis, lymphangiogenesis and neurogenesis during osteogenesis could provide key time nodes and regulatory targets for osteogenic material design, tissue-engineered bone construction and regenerative medicine research. For example, bioactive materials may be designed to release GDF5, PAPPA2 and other factors at 24–72 h after implantation, enhancing the early osteogenesis and vascularization process together with hBMSCs. As for the materials focused on neural integration and lymphangiogenesis, the release time point falls within the 72–168 h window. In regenerative medicine, the time trend provides the sequential regulation law of bone, blood vessel, lymphatic vessel and nerve regeneration, helping to optimize the regeneration strategy. The drug delivery systems can adjust the release of different factors based on time windows. For example, our data demonstrate that the vascularization process is initiated early (around 24 h), while the regulation of lymphangiogenesis and neurogenesis is relatively delayed, peaking between 72 and 168 h. This differential timing offers a framework for staged, sequential factor delivery in advanced biomaterials. Bioactive scaffolds designed for immediate bone repair should prioritize the early and rapid release of factors targeting the initial coupling phase (24–72 h). Specifically, the observed early upregulation and peaking of GDF5 at 72 h, along with angiogenesis-related PAPPA2, suggests that these two factors can be co-loaded and delivered in a burst-to-early-sustained release mode, enabling an initial high local concentration followed by a sustained exposure to support early hMSC commitment and vascular–osteogenic coupling. In terms of implementation, this early-phase delivery can be achieved using a spatiotemporally controlled strategy based on 3D printing-assisted layer-by-layer assembly, where GDF5 and PAPPA2 are preferentially incorporated into the outer layers of the construct to facilitate rapid or early sustained release upon implantation. Conversely, factors aimed at regulating neural integration and lymphangiogenesis should be delivered with a sustained release profile optimized for the 72–168 h window. Accordingly, these later-stage factors can be encapsulated within the inner layers to achieve delayed and prolonged release, thereby forming a sequential and partially overlapping delivery architecture that mirrors the temporal progression of tissue regeneration. This sustained delivery design is consistent with the delayed activation pattern of these lineages, as reflected by their later transcriptional upregulation in our time-course analysis. The progressive increase of the mineralization regulator MGP, peaking at 504 h, dictates that mineralization-inhibiting or -regulating factors should maintain a long-term presence within the scaffold microenvironment until the remodeling phase. In interpreting our findings, it is essential to acknowledge certain limitations. First, our bulk RNA-seq analysis offers a comprehensive view of transcriptional dynamics, yet it might conceal the underlying cellular heterogeneity within the hBMSC population. A more refined analysis, such as single-cell RNA sequencing, would be beneficial for future work. Crucially, this study utilized an in vitro two-dimensional (2D) culture system for osteogenic differentiation. Although it is highly controlled, this setup cannot fully recapitulate the complex three-dimensional (3D) microenvironment and biomechanical cues present during in vivo bone regeneration. Consequently, the precise molecular mechanisms and time windows identified here require further validation in more sophisticated 3D culture models and in vivo studies. Conclusions In conclusion, our study conducted time-series RNA-seq analysis on hBMSCs during osteogenic differentiation. By integrating DEGs data, GO and KEGG enrichment analysis, transcriptomic temporal trend analysis, and WGCNA, we found that the gene expression profile of hBMSCs indicated the existence of four characteristic transcriptional phases or patterns: the initial adaptation stage, proliferation activation stage, differentiation regulation stage, and remodeling stability stage. Moreover, 72 h may represent a potential key time window associated with the early transcriptional activation of osteogenic–vascular–neural–lymphatic coupling during BMSC-mediated bone regeneration, as suggested by the activation of BMP, vascular endothelial growth factor (VEGF), and PPAR signaling pathways. Four modules and closely related hub genes such as GDF5, MGP and PAPPA2 involved in these processes were also identified and highlighted. The study supplemented the transcriptional regulation in BMSC bone regeneration and provided reference time points for the design of biomaterials in regulating bone regeneration. Further research is needed to confirm the expression and regulation patterns of key genes at different time points. Materials and methods Cell culture Human bone mesenchymal stem cells (hBMSCs) were purchased from Procell Life Science & Technology Co. Ltd and cultured in Minimum Essential Medium α (MEM α) (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 1% streptomycin and penicillin (Solarbio, Beijing, China) and 10% fetal bovine serum (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) at 37 ℃ in a 5% CO2 atmosphere. The cells were passaged and the p3 to p5 cells were used in the subsequent experiments. Osteogenesis induction of hBMSCs The hBMSCs were seeded in 6-well plates at a density of 5 × 104 cells/well. Osteogenic induction medium which was composed of complete growth medium supplemented with 50 μg/mL ascorbic acid (CR Double Crane, Beijing, China), 0.1 μmol/L dexamethasone (Tianjin Pharmaceutical Group Co., Ltd., Tianjin, China) and 10 mmol/L β-glycerophosphate (Sigma-Aldrich, St. Louis, MO, USA), was added to replace the medium in the wells after culturing 24 h (as shown in Fig. 1A). Total RNA was extracted from the hBMSCs using Trizol (Invitrogen, Carlsbad, CA, USA) reagent according to the manufacturer’s protocol after 0 h, 1 h, 2 h, 4 h, 12 h, 24 h (1 day), 72 h (3 days), 168 h (7 days), 336 h (14 days) and 504 h (21 days) of osteogenic induction. A total of three (N = 3) independent biological replicates were used for the bulk RNA-sequencing at each time point. RNA-seq of hBMSCs during osteogenic differentiation RNA-sequencing was used to evaluate mRNA profiles in hBMSCs. Data analysis of 30 samples was completed by Beijing Easyresearch Technology Co., Ltd., (Beijing, China). NanoPhotometer® spectrophotometer was used to detect the purity of the samples (IMPLEN, CA, USA), and the Agilent 2100 RNA Nano 6000 Assay Kit (Agilent Technologies, CA, USA) was used to assay the concentration and integrity of RNA samples. The subsequent steps, including library building, library quality inspection, clustering and sequencing were executed following the established protocols recommended by the manufacturer, and 150 bp double-ended sequencing reads were obtained. Raw Data were stored in FASTQ file format. To ensure the quality of the data used for analysis, the useful Perl script was used to filter the original data (Raw Data) to obtain the high-quality sequences (Clean Data). The reference genomes and the annotation file were downloaded from the ENSEMBL database (http://www.ensembl.org/index.html). Bowtie2 v2.2.3 was used for building the genome index, and Clean Data were then aligned to the reference genome using HISAT2 v2.1.0. Then the IGV (Integrative Genomics Viewer) was used to view the mapping results by the heatmap, histogram, scatter plot or other style. Reads count for each gene in each sample was counted by HTSeq v0.6.0, and FPKM (fragments per kilobase million mapped reads) was then calculated to estimate the expression level of genes in each sample. Differential gene expression analysis and function enrichment analysis DESeq2 was designed for differential gene expression analysis between two samples with biological replicates. The theoretical basis of this analysis obeys the hypothesis of negative binomial distribution for the count value. Genes with p ≤ 0.05 and |log2_ratio|≥ 1 were identified as differentially expressed genes (DEGs). The GO (Gene Ontology, http://geneontology.org/) enrichment analysis of DEGs was implemented by the hypergeometric test to identify GO terms that are significantly enriched in DEGs compared to the whole genome. GO terms with p < 0.05 were significantly enriched. KEGG (Kyoto Encyclopedia of Genes and Genomes, http://www.kegg.jp/) is a database resource containing a collection of manually drawn pathway maps representing our knowledge of the molecular interaction and reaction networks. The KEGG enrichment of DEGs was implemented by the hypergeometric test. Terms with p < 0.05 were significantly enriched. Time-series transcriptomic trend heatmap and GO-BP enrichment analysis The heatmap indicating the temporal transcriptomic trend and the GO enrichment matrix dot plot were plotted by https://www.bioinformatics.com.cn (last accessed on 6 Feb 2025), an online platform for data analysis and visualization. The time-series gene expression data were divided into different clusters using mfuzz or kmeans algorithm. Then GO-BP enrichment was performed and the annotation information in each cluster was obtained. The ClusterGVis package was used for analysis. This online platform was also used for GO enrichment and KEGG analysis also used this online platform. Weighted gene co-expression network analysis Weighted gene co-expression network analysis (WGCNA) was performed to identify gene modules associated with osteogenesis, angiogenesis, lymphangiogenesis, and neurogenesis during the osteogenic differentiation of hBMSC. Genes with standardized RPKM values below 0.3 were excluded prior to performing WGCNA using the WGCNA v1.7.1 package in R v4.2.3 [61]. Module classification was conducted using a minimum soft-thresholding power of 30 (Supplementary Fig. 1), a merge cut height of 0.15, and a minimum module size of 20 genes. Genes exhibiting similar expression patterns were grouped into co-expression modules based on hierarchical clustering (Supplementary Fig. 2). Then Pearson correlation coefficients for relationships between the module and sample trait were calculated. Modules with Pearson correlation coefficient ≥ 0.3 and P < 0.05 were associated with osteogenesis. GO and KEGG enrichment analyses were further conducted to investigate the functions of genes in the modules, and then four modules were selected. We used the WGCNA package to calculate the intramodular connectivity (kWithin) between genes and intermodular (kOut) correlations and construct a gene co-expression network using Cytoscape v3.10.0 (Institute for Systems Biology, Seattle, WA, USA). Quantitative real-time PCR (qRT-PCR) analysis Extracted total RNA was also reverse transcribed by HiScript III RT SuperMix for qRT-PCR (Vazyme, Nanjing, China). The Tag Pro Universal SYBR gPCR Master Mix (Vazyme, Nanjing, China) along with Archimed X (Rocgene, Beijing, China) was used for real-time PCR according to the manufacturers’ instructions. The comparative Ct (2 − ΔΔCt) approach was used to analyze the data and determine the relative gene expression of key genes in WGCNA. The sequences of the primer used are shown in Table. S1. Enzyme-linked immunosorbent assay (ELISA) The hBMSCs were seeded in 6-well plates at a density of 5 × 104 cells/well and lysed after osteogenic induction for 24 h (1 day), 72 h (3 days), 168 h (7 days), 336 h (14 days) and 504 h (21 days). The cell culture supernatants were collected. ELISA kits for GDF5, MGP and PAPPA2 were purchased from BlueGene Biotech (Shanghai, China). The protocol was performed as follows: Preparation and sample addition: appropriately coated microwells were secured, and 100 µL of standard or sample was added to the respective wells, and 100 µL of PBS (pH 7.0–7.2) was used as the blank control. For the culture supernatant samples, 10 µL of Balance Solution was dispensed into the 100 µL specimen and thoroughly mixed. Conjugate incubation: 50 µL of conjugate was added to each well (excluding the blank control). The plate was covered and incubated at 37 °C for 1 h. Washing: after incubation, the plate underwent five washing cycles with the provided 1× wash solution. Complete removal of the liquid was ensured between each wash. Substrate reaction: 50 µL of Substrate A and 50 µL of Substrate B were added sequentially to all wells, including the blank control. The plate was covered and incubated in the dark at 37 °C for 10–15 min. Stopping and measurement: the reaction was stopped by adding 50 µL of Stop Solution to each well. The optical density (OD) was immediately measured at 450 nm using a microplate reader. The concentration of the target factor in the samples was calculated based on the standard curve generated from the serial dilutions. Construction of regulatory interaction network and system-wide interaction networks Regulatory interaction network and global regulatory network analyses were performed by integrating transcriptomic co-expression profiles with functional evidence from the STRING database. Potential molecular targets of GDF5 within the BMP/SMAD1 pathway were identified based on Pearson correlation coefficients (|r|> 0.85, p < 0.05) calculated via Hmisc v5.1 package, with the resulting GDF5–SMAD1 axis visualized in Cytoscape v3.10.1. Expression patterns of the PAPPA2–IGF signaling axis were analyzed using limma v3.58 package and pheatmap v1.0.12 package. The synergistic coupling between IGF components and lineage-specific markers (e.g., RUNX2, PROX1) was evaluated using pairwise Pearson correlation analysis (|r|> 0.85, p < 0.05). Finally, a system-wide coupling network was established by utilizing the pre-identified core hub genes and their first-order neighbors (|r|> 0.7, p < 0.05) extracted from key WGCNA modules. These nodes were mapped against the STRING database v12.0 with a high-confidence interaction score threshold (> 0.7) to integrate documented physical interaction evidence with transcriptomic co-expression data. Data wrangling and topological refinements, including the calculation of node degree and betweenness centrality, were executed via tidygraph v1.2 and dplyr. The osteogenic–vascular–neural–lymphatic axis was rendered using a circular hierarchical layout in Cytoscape to illustrate the synergistic crosstalk and multi-lineage coordination within the BMSC niche. Statistical analysis All experiments were performed at least three times and statistical analyses were conducted using GraphPad Prism 9.5 (San Diego, USA). T-tests, one-way ANOVA, or two-way ANOVA were applied according to the data characteristics. Differences were considered statistically significant at p < 0.05, and the significance levels were indicated as follows: ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Supplementary Information Supplementary Material 1 Supplementary Material 1