Introduction

Osteoporosis is a progressive bone disease that is featured by a decrease in bone mass and density. The disease can be classified as primary osteoporosis and secondary osteoporosis. Osteoporosis and associated fractures are a cause of mortality and morbidity worldwide [1]. According to the 2003–2006 survey of China Ministry of Health, almost 69.4 million Chinese above the age of 50 years suffered from osteoporosis and the economic burden of osteoporosis to Chinese patients was heavy [1, 2].

The underlying mechanism of osteoporosis is an imbalance between bone resorption and bone formation. The three major mechanisms by which osteoporosis develops are an inadequate peak bone mass (PBM), excessive bone resorption, and inadequate formation of new bone during remodeling [3]. It has been reported that the overall rates of both bone formation and bone resorption remain high in elderly women [4]. Meanwhile, estrogen plays a fundamental role in the maintenance of skeletal homeostasis [5] and estrogen deficiency in elderly women could induce bone loss and hence cause osteoporosis [6]. Estrogen and its receptor are then considered to be the major factors in the pathogenesis of osteoporosis [7, 8]. Besides, aging, which is related to a loss of sex hormone in both women (menopause) and men (andropause), is also an important risk factor of osteoporosis [9]. In women, estrogen is implicated as playing a critical role in aging [10]. Furthermore, PBM, the amount of bone present at the end of skeletal maturation, is obtained in early adulthood and then bone loss aggravated with increasing age [11, 12]. Oxidative stress is reported to be a fundamental mechanism of the age-dependent decrease of bone mass [13, 14]. The fat and bone connection also take part in the pathophysiology of age-related bone loss [15]. Age-related collagen cross-linking is observed in osteoporosis, which can affect the mechanical properties of bone via influence on mineralization process and microdamage formation [16]. Manolagas et al. [17] propose a revise on the perspective of the pathogenesis from estrogen-centric to aging and oxidative stress. Obviously, aging causes a range of changes in the processes of osteoporosis, which makes microarray technology an ideal tool to unveil the complicated molecular mechanisms.

Moreover, human mesenchymal stem cells (MSCs), with high self-renewal capacity, are mainly differentiated into adipocytes and osteoblasts in adult bone marrow [18]. In addition, MSCs are the cellular sources of fracture healing and closely related to osteoporosis [19]. In the present study, we used microarray analysis to identify the differentially expressed genes (DEGs) in MSCs from elderly patients suffering from osteoporosis compared with those of young and old non-osteoporotic donors. Furthermore, the significantly enriched functions and pathways of DEGs were screened and the protein–protein interaction (PPI) network was constructed, which might provide a deeper insight into the role of aging in the molecular mechanisms of osteoporosis.

Materials and methods

Microarray data

Microarray data set GSE35956 and GSE35958 [20] was downloaded from Gene Expression Omnibus. A total of 14 human mesenchymal stem cells (MSC) samples were acquired (Table 1), five samples from elderly patients (females; 86.2 + 5.89 years old) suffering from osteoporosis, five controls from young non-osteoporotic donors (four females and one male; 57.6 + 9.56 years old) and four controls from old non-osteoporotic donors (three females and one male; 81.75 + 4.86 years old). MSCs of patients suffering from osteoporosis were obtained from femoral heads after low-energy fracture of the femoral neck. Meanwhile, the vertebrae fractures acts as an additional criteria to confirm the primary osteoporosis in the patients. MSCs of non-osteoporotic donors were isolated from bone marrow of femoral heads after total hip arthroplasty. The experiments have been approved by the local Ethics Committee of the Medical Faculty of the University of Wuerzburg and performed in accordance with the ethical standards. Affymetrix Human Genome U133 Plus 2.0 Array was used. Chi square test was used to compare the gender between any two groups and a P value less than 0.05 was considered to be significantly different.

Table 1 Summary of the 14 transcriptome data
Full size table

Pre-treatment of raw data

Raw data were in CEL format and processed using package affy [21] as well as RefSeq annotation files, which was followed by normalization with robust multi-array average (RMA) method. Finally, expression levels were determined for 26,739 genes.

Screening of DEGs

Gene expression data were divided into three groups: young control group, old control group and osteoporosis group. Three comparisons were conducted among the 3 groups. Differential analysis was performed using package limma [22]. P value ≤0.05 and log|Fold change| > 1 were set as the cut offs to screen out DEGs.

Functional enrichment analysis of DEGs

Kyoto encyclopedia of genes and genomes (KEGG) pathway enrichment analysis and gene ontology (GO) enrichment analysis were carried out using DAVID [23]. P value <0.05 was set as the threshold.

Construction of protein–protein interaction (PPI) network

A PPI network was constructed for the protein products of DEGs using STRING [24] and then was visualized using Cytoscape [25].

Results

Differentially expressed genes in the three groups

A total of 3179 mRNAs were detected. Compared with young control group, 271 genes were up-regulated (termed as ost_C_up genes) and 560 genes down-regulated (termed as ost_C_down genes) in osteoporosis group. Compared with old control group, 781 genes were up-regulated (termed as ost_old_up genes) and 1528 genes down-regulated (termed as ost_old_down genes) in osteoporosis group. In comparison with young control group, 180 genes were up-regulated (termed as old_C_up genes) and 294 genes were down-regulated (termed as old_C_down genes) in old control group. In addition, 158 DEGs were shared in ost_C_up and ost_old_up groups; three DEGs were shared in ost_C_up and old_C_up groups; 269 DEGs were shared in ost_C_down and ost_old_down groups; five DEGs were shared in ost_C_down and old_C_down groups. Venn diagram of DEGs is shown in Fig. 1. Meanwhile, Chi square test showed that there were no significant differences (P > 0.05) in gender between any two groups.

Fig. 1
figure 1

Venn diagram of up-regulated genes (a) and down-regulated genes (b). ost_C_up: genes up-regulated in osteoporosis group compared with young control group; ost_old_up: genes up-regulated in osteoporosis group compared with old control group; old_C_up: genes up-regulated in old control group compared with young control group; ost_C_down: genes down-regulated in osteoporosis group compared with young control group; ost_old_down: genes down-regulated in osteoporosis group compared with old control group; old_C_down: genes down-regulated in old control group compared with young control group

Full size image

Functional enrichment analysis results

Five significant KEGG pathways were identified in ost_C_up genes and 17 pathways were revealed in ost_old_up genes (Fig. 2a). Four terms were shared by both and they were extracellular matrix (ECM)–receptor interaction (hsa04512), focal adhesion (hsa04510), mammalian target of rapamycin (mTOR) signaling pathway (hsa04150) and bladder cancer (hsa05219). Significant pathways of down-regulated genes in osteoporosis group are shown in Fig. 2b.

Fig. 2
figure 2

Significant KEGG pathways in up-regulated genes (a) and down-regulated genes (b). Significant terms were in red, while insignificant ones were in black

Full size image

PPI networks of DEGs

Interactions with coefficient >0.4 were retained in following analysis. Twenty-six interactions were disclosed among ost_C_up genes, from which four subnetworks were extracted using ClusterONE. All genes in subnetwork 1 (Fig. 3a) were associated with ECM–receptor interaction. Besides, integrin beta 2 (ITGB2), sorbin and SH3 domain containing 3 (SORBS3) and zyxin (ZYX) from subnetwork 2 (Fig. 3b) were involved in cell adhesion.

Fig. 3
figure 3

Subnetwork 1 (a) and 2 (b) from the whole PPI network for ost_C_up genes

Full size image

Ninety-seven interactions were identified among ost_old_up genes and five subnetworks were obtained. Genes from subnetwork 1 of ost_old_up genes were enriched in focal adhesion (Fig. 4a) and genes from subnetwork 2 were significantly enriched in cell-substrate adhesion (Fig. 4b).

Fig. 4
figure 4

Subnetwork 1 (a) and 2 (b) from the whole PPI network for ost_old_up genes

Full size image

A total of 768 interactions were observed among ost_C_down genes and 11 subnetworks were acquired. In addition, DNA replication (hsa03030), cell cycle (hsa04110), nucleotide excision repair (hsa03420) and mismatch repair (hsa03430) were found to be significantly enriched in subnetwork 1 of ost_C down genes. Meanwhile, 501 interactions were found among ost_old_down genes.

Two interactions were disclosed among old_C_up genes and 54 interactions were uncovered among old_C_down genes, from which three subnetworks were obtained. Synaptosomal-associated protein, 23 kDa (SNAP23), syntaxin 5 (STX5) and vesicle-associated membrane protein 3 (VAMP3) from subnetwork 1 were implicated in SNARE interactions in vesicular transport (hsa04130). Coatomer protein complex, subunit alpha (COPA), coatomer protein complex, subunit epsilon (COPE) and adaptor-related protein complex 2 mu 1 subunit (AP2M1) from subnetwork 2 were linked to vesicle-mediated transport.

Furthermore, 17 genes were shared in both ost_C_up and ost_old_up groups, including transforming growth factor beta 1 (TGFB1), neuregulin 1 (NRG1), cyclin-dependent kinase inhibitor 1A (CDKN1A), SORBS3, talin 1 (TLN1), integrin-binding sialoprotein (IBSP), ZYX, insulin–insulin-like growth factor 2 (INS-IGF2), parvin beta (PARVB), mitogen-activated protein kinase kinase 2 pseudogene (LOC407835), insulin-like growth factor 2 (IGF2), integrin beta 2 (ITGB2), vascular endothelial growth factor B (VEGFB), iduronidase, alpha-l-(IDUA), mitogen-activated protein kinase kinase 2 (MAP2K2), insulin (INS) and collagen type VI alpha 1 (COL6A1).

Discussion

Currently, aging is believed to be involved in the development of osteoporosis [17, 26]. In the present study, pairwise comparisons were performed among the three groups of transcriptomes: osteoporosis group, young control group and old control group. A great deal of DEGs was screened in each group. In addition, several important pathways were identified, such as ECM–receptor interaction, focal adhesion, mTOR signaling pathway and bladder cancer. Finally, a range of genes (e.g., TGFB1, IGF2, and ZYX) were selected to be osteoporosis-related by analyzing the pathways and subnetworks.

ECM–receptor interaction, focal adhesion and mTOR signaling pathway were enriched in both ost_C_up genes and ost_old_up genes in this study. ECM provides structural and biochemical support to the surrounding cells. ECM–receptor interaction is implicated in a range of biological process, such as cell differentiation, proliferation and apoptosis via the role in signal transduction. The involvement of ECM in bone development has been widely reported [27, 28]. MalaCards [29] has predicted that ECM–receptor interaction is associated with osteoporosis. Moreover, focal adhesion is a specialized attachment site where the cell makes close contact with either ECM or to other cell surface molecules [30]. Salasznyk et al. [31] have reported that focal adhesion kinase signaling pathways regulate the osteogenic differentiation of human MSC. Young et al. [32] have found that focal adhesion kinase is important for fluid shear stress-induced mechanotransduction in osteoblasts. mTOR, an atypical serine/threonine protein kinase, is able to integrate extracellular signals, cause phosphorylation of downstream target proteins, and thus participate in the regulation of cell growth, proliferation and other processes. Previous studies have indicated its role in osteoclast survival [33, 34]. Xian et al. [35] have found that matrix insulin-like growth factor 1 maintains bone mass by activation of mTOR in mesenchymal stem cells. Besides, VEGF signaling pathway (hsa04370), ErbB signaling pathway (hsa04012) and MAPK signaling pathway (hsa04010) were the significantly enriched pathways in ost_old_up genes and they might be also related to osteoporosis [36, 37]. Therefore, our results suggest that aging may be involved in the occurrence and development of osteoporosis by participating in ECM–receptor interaction, focal adhesion and mTOR signaling pathway.

Based upon the six osteoporosis-related pathways and subnetworks, a total of 49 genes were screened out. Meanwhile, seventeen genes were shared in both ost_C_up and ost_old_up groups, including TGFB1, NRG1, CDKN1A, SORBS3, TLN1, IBSP, ZYX, INS-IGF2, PARVB, LOC407835, IGF2, ITGB2, VEGFB, IDUA, MAP2K2, INS and COL6A1. TGFB1 is an important regulator of homeostasis of bone metabolism. The association between its polymorphisms and bone mineral density as well as osteoporosis has been investigated [38, 39]. IGF2, an aging-related gene [40], can lead to osteogenic differentiation and bone formation together with bone morphogenetic protein 9 [41]. Additionally, SORBS3, a subnetwork-related gene, encodes a cell adhesion molecule and reflects acceleration of age-related changes [42]. Therefore, these genes may be related to aging and act important roles in elderly osteoporosis. By further analyzing these genes and related pathways, the exactly involved signaling pathways of these genes will be figured out and more targets will be excavated for drug development.

However, there is a limitation in our study. Although the levels of estrogen in hMSCs are limited, the hormone levels in all subjects are unknown and the interference of estrogen may be not completely excluded. The larger population and further experiments are needed to confirm our results.

Overall, our study unveiled a range of pathways such as ECM–receptor interaction, focal adhesion and mTOR signaling pathway that were associated with osteoporosis by bioinformatics analysis. Furthermore, the 17 shared DEGs such as, TGFB1, IGF2I and ZXY, which were up-regulated both in ost_C_up and ost_old_up groups, may be aging-related genes and involved in osteoporosis. These findings might benefit future researches in discovering biomarkers and developing new therapies for osteoporosis.