Identication of Candidate Genes Associated with Steroid-Induced Osteonecrosis of The Femoral Head by Bioinformatics Based on GEO Database

BACKGROUND(cid:0) Steroid-induced osteonecrosis of the femoral head (SONFH) is a progressive bone disorder and its characterized by femoral head collapse and hip joint dysfunction and the biomarkers of SONFH remain unclear. The purposes of this study are to identify the signicant biological function and pathway involved in SONFH, and further to search the underlying mechanism of this pathway in SONFH. METHODS(cid:0) The GSE123568 dataset obtained from the Gene Expression Omnibus (GEO) database and normalized using Robust Multiarray Averaging (RMA) methods. And the Gene set enrichment analysis (GSEA), Ingenuity pathway analysis (IPA), VarElect online tool, MalaCards database, miRWalk online tool, DIANA too, and Cytoscape were integrated for bioinformatics analyses. RESULTS(cid:0) 6 biological processes and 4 KEGG pathways were enriched by GSEA, and 68 candidate genes were involved in these pathways. Besides, the canonical pathway and molecule function analysis by IPA, the results revealed that 10 canonical pathways and 12 candidate genes were identied, and 20 modules and 101 candidate genes were enriched by molecule function analysis. The above candidate genes were combined and ltered using the VarElect online tool. The ltered candidate genes were overlapped with another cluster of candidate genes from the MalaCards database to identify hub genes ACP5, TNF, MMP8. Based on the hub genes, the miRNAs were screened and overlapped to predict the lncRNAs. Total 7 miRNAs of ACP5, TNF, MMP8 were targeted 956 candidate lncRNAs. CONCLUSIONS(cid:0) In summary, this study identied the hub candidate genes and pathways associated with SONFH progress, and constructed the ceRNA network based on the hub candidate genes. Our ndings might provide the potential biomarkers of SONFH diagnosis and treatment.

Introduction shares microRNAs 10,11 . It has been reported that lncRNAs act as the important regulators in SONFH through functioning as ceRNAs of miRNAs and target genes 12 . For example, lncRNA RP11-154D6 promotes osteogenic differentiation and inhibits adipogenic differentiation in bone marrow mesenchymal stem cells (BMSCs) to contribute SONFH progress 13 . lncRNA RP1-193H18.2, MALAT1, and HOTAIR were proved association with abnormal osteogenic and adipogenic differentiation of BMSCs in the patients with SONFH 14 . However, the functions and mechanisms of lncRNA have been entirely unknown in SONFH.
Taken together, we used bioinformatic methods to identify the differentially expressed genes of SONFH, and to enrich the SONFH related pathways. Based on the candidate genes to construct lncRNA-miRNA-mRNA network in SONFH. Our ndings revealed the candidate genes and key pathways of SONFH and proved the potential biomarkers of SONFH.

Materials And Methods
Data acquiring and processing The mRNA expression pro les of SONFH (GSE123568) were acquired from GEO database. The GSE123568 dataset was performed using Affymetrix EG1.0 array, including the peripheral serum of 30 SONFH patients and 10 non-SONFH patients (following steroid administration). RMA was used for the background correction of raw mRNA expression data, and then processed signals were log2 transform and normalize through quantile normalization. Furthermore, the median-polish probe sets were summarized using affy R package. The quality was assessed by samples clustering based on the distance between different samples in average linkage.
Gene set enrichment analysis (GSEA) The SONFH samples were divided into two groups according to the mRNA expression levels of GSE123568 dataset by the GSEA software. c5.all.v7.1.symbols.gm and c2.cp.kegg.v7.1.symbols.gm from molecular signature database (MSigDB) version 6.2 were selected as the reference gene set. The gene sets > 500 and gene sets < 15 were used as the excluded criteria. The signi cant gene sets were identi ed according to a threshold FDR < 0.25 and P-value < 0.05.

Ingenuity pathway analysis (IPA)
The common DEGs of GSE123568 were uploaded into Qiagen's IPA system for core analysis according to the gene sets of the ingenuity knowledge base. IPA was used to screen canonical pathways associated with common DEGs, and explore the association between diseases and gene function. P-value < 0.05 and | Z-score | > 2 used as the threshold.
VarElect is an online tool that entails specifying a gene symbol list imported from an experimental data le (s), together with disease phenotype and symptom terms related to the studied disorder 15 . The candidate genes which enriched by GSEA and IPA were rstly integrated, then uploaded the candidate genes into VarElect online tool (http://ve.genecards.org) to rank genes, and point out the candidate genes likely to be related to SONFH. Besides, the human disease database MalaCards (http://www.malacards.org/) was used to obtain genes that correlated to SONFH. Moreover, Venn diagrams were applied to calculate the intersections of SONFH related genes mentioned above.

Results
DEGs identi cation and pathway enrichment by GSEA GSEA was used to enrich the pathways of these genes. The enriched gene sets with a threshold FDR < 0. 25 , table 2). Besides, the molecule function analysis indicated that the differentially expressed genes were enriched in different pathways in SONFH ( gure 2B), and the enriched genes were involved in 48 functional modules with a threshold |z-score| > 2 and P-value <0.05, 20 modules were signi cantly associated with SONFH ( gure 2C, table 3). The function among these modules involved in microvascular injury and intravascular coagulation, oxidative stress, immunoin ammation, and bone marrow cell activity. Among them, microvascular injury and intravascular coagulation included cell-to-cell signaling and interaction, cellular movement, cardiovascular system development and function, cell-to-cell signaling and interaction, cell morphology. Oxidative stress included free radical scavenging, cardiovascular system development and function, cell-to-cell signaling and interaction, free radical scavenging, molecular transport. Immuno-in ammation included in ammatory response, cell death and survival, cardiovascular system development and function, cell-tocell signaling and interaction, and cell-mediated immune response. bone marrow cell activity included cellular movement (table 3). And the 101 candidate genes involved in 20 function modules, including chemokine receptors, and immune-related genes, etc. Such as CCR1, CCR2, IGHG3, IGKV1-39 (supplementary table 1).

Identi cation of candidate genes of SONFH
In order to identify the hub genes, the candidate genes were combined and ranked. Then we identi ed the most likely genes related to SONFH by VarElect online tool. We found that TNF with the highest correlation score, and TNFSF10, FAS, TNFSF13B, FASLG exhibited higher correlation scores than other candidate genes (Table 4). Moreover, the 32 candidate genes related to SONFH were obtained from the MalaCards database (Table 5). Furthermore, the intersections of SONFH related hub genes were identi ed by Venn diagrams including ACP5, TNF, MMP8 ( gure 3).

Discussion
SONFH is a progressive bone disorder caused by excessively administrating glucocorticoids and resulted in vascular damage, mechanism stress damage, intraosseous pressure increasing, adipocyte dysfunction, apoptosis, and coagulation dysfunction 5 . In our study, we rst identi ed the biological function of SONFH, which includes programmed necrotic cell death, intracellular lipid transport, necrotic cell death, lymph vessel morphogenesis, hydrogen peroxide catabolic process, and lymph vessel development. It has been reported that the nal step in osteonecrosis is vascular insu ciency to the femoral head, resulting in apoptosis and necrosis 16 . Several recent studies have reported that apoptosis relates to the pathogenesis of osteonecrosis of the femoral head 17,18 . Besides, previous studies have shown that steroid treatment implied the intra-osteoblastic lipid droplets pathology and corresponded to low bone mass with increased bone marrow adiposity 19,20 . Around lymph vessel morphogenesis and development, several studies have indicated that different stem and progenitor cells reside in distinct cellular niches in bone marrow, such as hematopoietic stem cells occupy a perivascular niche and early lymphoid progenitors occupy an endosteal niche 21 .
Besides, we screened signi cant pathways related to SONFH including leishmania infection, glycosaminoglycan biosynthesis chondroitin sulfate, and T cell receptor signaling pathway. Several studies proved that mice infected with Leishmania showed osteonecrosis. In addition, the histopathological analysis demonstrated that mononuclear cells in ltrated in plasma cells richly as well as parasitism of intra-medullary and extra-medullary macrophages intensely, also with bone necrosis areas and discrete cartilaginous tissue involvement 22,23 . Okazaki previously reports that the toll-like receptor (TLR) 4 signaling pathway, which induces in ammatory status, contributes to the pathogenesis of non-traumatic ONFH in rats [24][25][26] . In Okazaki's another study, it has shown that corticosteroid treatment after the administration of TLR7 or TLR9 ligands causes ONFH in rats, whereas corticosteroids alone failed to induce ONFH in healthy animals. In addition, IRF7 and NF-κB are activated in the liver induced by corticosteroid treatment to trigger the development of ONFH 27 . Taken together, normalization of in ammatory status when treating underlying in ammatory diseases may potentially prevent ONFH in the future.
In addition, Gessner's study has illustrated that the differential expressed IL-9 between the susceptible and resistant mice which infected with Leishmania 28 . Moreover, Geng's study also reveals that the production of IL-9 may trigger the cartilage degeneration and destruction in ONFH patients. Il-9 promotes cartilage degeneration, and the effect of IL-9 on cartilage is alleviated by blocking JAK-STAT signaling pathway in a human primary chondrocyte culture model 29 . Furthermore, Chen's study has demonstrated IL-21 promotes cartilage degradation by activating cartilage in ammation through JAK-STAT signaling pathway in ONFH patients 30 . The studies above indicated that immune-related genes act as the critical role in ONFH progression through modulating the cartilage degeneration and destruction.
Furthermore, the IPA was used to further investigate the canonical pathway and molecule function of SONFH. The results revealed that 9 pathways were activated, which includes interferon signaling, production of nitric oxide and reactive oxygen species in macrophages, iNOS signaling, Fcγ receptormediated phagocytosis in macrophages and monocytes, TREM1 signaling, Gαi signaling, neuroin ammation signaling pathway, Tec kinase signaling, and Dendritic cell maturation. In contrast, Sirtuin signaling pathway was suppressed. Besides, the molecule function analysis showed that oxidative stress, microvascular injury and intravascular coagulation, immune in ammation, and myeloid cells movement were signi cantly involved in SONFH. Here, we found the production of nitric oxide and reactive oxygen species in macrophages showed the strongest correlation with SONFH based on the highest score. Macrophages play the proin ammatory promoter in necrotic bone, Naga Suresh Adapala et al have reported that the numbers of proin ammatory M1 macrophages are increased in the repair bone tissue, which reveals high expression of proin ammatory cytokines IL-1β, TNF-α, and IL-6 and the pattern recognition receptor TLR4 31 . Another study has proved that TNF-a-mediated alteration of M1/M2 macrophage polarization plays a vital role in the pathogenesis of steroid-induced osteonecrosis, with a dominant position that M1 macrophages in early stage and M2 macrophages in the late stage of osteonecrosis 32 . In our result, production of nitric oxide and reactive oxygen species in macrophages was activated. Generally, macrophages maintain organism homeostasis by receptor-mediated recognition and phagocytic uptake of pathogenic damaged or apoptotic host cells. The necrosis bone tissues are degraded by phagocytosis, which activated by proteolytic enzymes and oxidative burst through the formation of reactive oxygen species (ROS) and nitric oxide (NO). The reaction of superoxide with NO results in the formation of peroxynitrite, which interacts with lipids, DNA, and proteins via direct oxidative reactions or indirect radical-mediated mechanisms 33 . All the studies suggest that oxidative stress signi cantly associates with necrosis bone may inducing macrophages polarization. In the present study, the oxidative stress related pathways were strongly related to the SONFH, and total 12 genes, such as ARG2, IFNGR1, which were associated with the production of nitric oxide and reactive oxygen species in macrophages pathway. Our nding consists of previous studies. ARG2 plays a vital role in nitric oxide and polyamine metabolism through repressing nitric oxide synthesis and inhibiting in ammatory genes levels in macrophages to interrupt M1 macrophage phagocytosis activity 34,35 .
Based on the above research, the SONFH related candidate genes were ltered by VarElect online tool and then overlapped with the candidate genes from MalaCards database. ACP5, TNF, MMP8 were identi ed as the hub genes of SONFH.
Tartrate-resistant acid phosphatase 5 (ACP5 ), a metalloprotein enzyme that belongs to the acid phosphatase family and is known to be expressed by osteoclasts. Furthermore, it has been demonstrated that ACP5 acts as a classic marker for bone resorption and osteoclast differentiation 36 . In Yin's study, ACP5 has increased in human ONFH tissues, and high expression of miR-410 and low expression of Wnt-11 inhibit ACP5 and MMP9 expression in ONFH rats 37 . Fang's study has shown the effects of TNFα on proliferation, angiogenesis, and osteogenesis, and osteogenesis of rat bone mesenchymal stem cells To further study the function and mechanism of hub genes in SONFH, the 372 miRNAs of ACP5, 795 miRNAs of TNF, 4 miRNAs of MMP8 were screened. And hsa-miR-7845-5p, hsa-miR-6772-3p, hsa-miR-5010-3p, hsa-miR-4653-5p, hsa-miR-1587 were intersected in ACP5, TNF, MMP8. And the lncRNAs were predicted by combination of miRNAs and miRNAs. Then, 7 miRNAs, and ACP5, TNF, MMP8 and lncRNAs.