ACP5, TNF, and MMP8 were identied as potential biomarkers of steroid-induced osteonecrosis of the femoral head

Steroid-induced osteonecrosis of the femoral head (SONFH) is a progressive bone disorder that is characterized by femoral head collapse and hip joint dysfunction. To elucidate the biomarkers of SONFH, the GSE123568 dataset was downloaded from the Gene Expression Omnibus (GEO) database. A total of 436 differentially expressed SONFH genes were screened in comparison with non-SONFH genes. Six biological processes and four KEGG pathways were enriched in SONFH by GSEA, and 68 candidate genes that were involved in these pathways were selected for subsequent analysis. Moreover, through an ingenuity pathway analysis, we obtained 10 canonical pathways and 20 molecule function modules related to SONFH, and acquired 121 candidate genes. Furthermore, we identified ACP5, TNF, and MMP8 as the genes most related to SONFH according to the VarElect and MalaCards database. Based on these hub genes, the targeted miRNAs and the lncRNAs were predicted. Finally, the ceRNA network was constructed by using ACP5, TNF, MMP8, seven miRNAs, and 956 candidate lncRNAs. In conclusion, the ACP5, TNF, and MMP8 might be potential biomarkers of SONFH. regulation mechanism


Introduction
Osteonecrosis of the femoral head (ONFH), which is also called avascular necrosis (AVN) of the femoral head, is an increasing health problem in the world. ONFH refers to progressive necrosis of osteocytes and bone marrow elements, and the clinical symptoms include grievous pain and limping gait [1] . Steroid administration is the main cause of ONFH, and it occurs in 51%-60% of all cases of ONFH [2,3] . The etiology and pathological process of SONFH are complex. Oxidative stress [1] , osteoclasts activation [4] , and bone-metabolism disorder [5] have been correlated with SONFH. However, the exact molecular mechanisms of SONFH remain unclear.
Although, magnetic resonance imaging (MRI) is a valid method for diagnosing SONFH [3] , some patients were misdiagnosed due to the complexity of their clinical manifestations. Therefore, reliable diagnostic biomarkers of SONFH are urgently needed.
A recent study demonstrated that the osteoclastic-related genes OPG and RANML were abnormally expressed in SONFH, and therefore, they are potential diagnostic markers [6] . However, other processes, including coagulopathy [7] , dysregulated apoptosis [8] , and disordered lipid metabolism [9] were also crucial for SONFH. In addition, inflammatory pathways such as the PDK1/AKT/mTOR signaling pathway and the PERK and Parkin pathways [10,11] are utilized for the function of dysregulated genes in these function pathways might act as potential biomarkers of SONFH.
Non-coding RNAs have been identified as regulators of gene expression and physiological process. In SONFH, studies have shown that miRNAs are aberrantly expressed in necrotic tissue [12] , serum [13] , bone marrow mesenchymal stem cells [14] , and osteoblasts [15] . Abnormal miRNA expression causes dysregulation of genes. ncRNAs and lncRNAs, which act as competing endogenous RNAs (ceRNAs) of miRNAs [16] , have rarely been investigated in SONFH. Xiang et al. found that lncRNA RP11-154D6 impacted the progress of SONFH by promoting osteogenic differentiation and inhibiting adipogenic differentiation in BMSCs [17] . lncRNA RP1-193H18.2, MALAT1, and HOTAIR have been associated with abnormal osteogenic and adipogenic differentiation of BMSCs in SONFH [18] . However, the mechanisms of lncRNA in SONFH have yet to be discovered. Revealing the associations among mRNA, miRNA, and lncRNA in the ceRNA network of SONFH can elucidate activities at the molecular level that would lead to uncovering new targets for therapy and diagnosis.
In our study, we analyzed the function pathways in SONFH based on differentially expressed genes and screened hub genes to predict miRNA and lncRNA to construct a ceRNA network in SONFH. Our study revealed candidate biomarker genes and potential regulated pathways of SONFH.

Data download and processing
We downloaded data on the gene expression profiles of SONFH from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). Dataset GSE123568 was performed on the Affymetrix human gene expression array, including a total of 40 peripheral serum specimens (30 SONFH patients and 10 non-SONFH individuals). The background correction of raw data were processed by robust multiarray average (RMA), and the signals were log2 transformed and normalized through quantile normalization. After that, qualified transcriptome data were used for future analyses.

Gene set enrichment analysis
To explore the functional phenotypes between SONHF and non-SONFH, we first used the "limma"R package to screen the differentially expressed genes (DEGs) between these two groups in the GSE123568 dataset. The screening standard was |log2FC|>=1 and p < 0.05. Then, we performed gene set enrichment analysis (GSEA) by using "c5.all.v7.1.symbols.gm" and "c2.cp.kegg.v7.1.symbols.gm" as the reference gene set. In addition to adopting the default parameter, larger sets > 500 and smaller sets < 15 were excluded, as additional parameter settings for filtering enriched gene sets. A false discovery rate (FDR) < 0.25 and p < 0.05 were considered significant. Genes participating in significant pathways were selected as candidate genes.

Ingenuity pathway analysis
The ingenuity pathway analysis (IPA) system was used for core analysis based on the DEGs to determine the function pathways involved in SONFH. Analyses including screening canonical pathways were used to explore the relationship between the gene function and diseases, | z-score | > 2 and p < 0.05 were set as the standard. Genes that were involved in significant pathways were obtained as candidate genes.

Identification of SONFH-related genes
To identify SONFH-related genes, we first integrated candidate genes. Then, the VarElect online tool (http://ve.g.,enecards.org) was implemented to select the genes most strongly associated by score.oreover, MalaCards database (http://www.malacards.org/) was utilized to acquire genes correlated to osteonecrosis.
Finally, SONFH-related genes were identified by the two intersecting methods, and mutual genes were obtained.

DEG identification and pathway enrichment by GSEA
After data preprocessing and screening under the threshold of |log2FC|>=1 and p were significantly up-regulated in SONFH, whereas porphyrin and chlorophyll metabolism were down-regulated (Figures 2A-D). Sixty-eight genes involved in these pathways were selected as candidate genes of SONFH (Table 1).

Candidate genes identified by IPA
To obtain candidate SONFH-related-genes, DEGs were also employed for IPA core analysis with a filter of p < 0.05 and |z-score| > 2. We found that nitric oxide ( Moreover, based on the disease and function analysis, we found that the DEGs were collected in different pathways ( Figure 3B). Of 48 functional modules, 20 were filtered by p <0.05 and |z-score| > 2, and identified as strongly correlated to SONFH ( Figure 3C).

Identification of SONFH-related genes
To identify the hub genes of SONFH, we first integrated the candidate genes screened by GESA and PIA and uploaded them into the VarElect online tool. The result showed that TNF acquired the highest correlation score; TNFSF10, FAS, TNFSF13B, and FASLG obtained a higher correlation score than that of the other candidate genes (Table 3). Then, 32 osteonecrosis-related candidate genes were selected from the MalaCards database (Table 4). Finally, ACP5, TNF, and MMP8 as SONFH-related genes were selected using the Venn diagram ( Figure 4).

Construction of the ceRNA network of SONFH-related genes
To explore the regulatory mechanism of ACP5, TNF, and MMP8 in SONFH, ceRNAs networks including genes, miRNA, and lncRNA were established. Then, 372, 795, and 265 miRNAs were targeted to ACP5, TNF, and MMP8, respectively.

Discussion
SONFH is a progressive bone disease caused by steroid treatment that results in mechanism stress damage, vascular injury, intraosseous pressure enhancing, adipocyte activated, coagulation, and apoptosis dysfunction [3] . The credible biomarker of SONFH is still unknown. In the present study, we enriched the biological function and pathway related to SONFH by clusterProfiler, ClueGO, IPA, and GSEA analyses. In the future, SONFH-related genes were screened by the VarElect and the MalaCards databases. Finally, ACP5, TNF, and MMP8 were identified as the most SONFH-related genes, and a ceRNA network was constructed for searching for the regulation mechanism in SONFH Previous studies indicated that ACP5 is expressed in many types of differentiated cells, such as granulocytes, dendritic cells, macrophages, and osteoclasts [19][20][21] . ACP5 is a histochemical marker for osteoclasts [22] . Furthermore, ACP5 is a multifunctional protein that is necessary for novel bone development, osteoclast differentiation, bone resorption, and osteoclast activity [23] . Since ACP5 has a physiological expression, its pathological expression in different human conditions appears to be reasonable [23] . Recently, a study revealed that ACP5 declined in ONFH tissues and was regulated by Wnt-11 and miR-410 [24] . This was in accord with our results that demonstrated that ACP5 was down-regulated in SONFH. Therefore, ACP5 is a potential biomarker for SONFH. Interestingly, TNF, which is one of the biomarker genes that is involved in the RANKL-NFATc1/c-FOS signaling pathway, induced ACP5 transcription to accelerate osteoclastogenesis [25] . In defining the function mechanism, scientists have reached an agreement that TNF-α inhibits collagen synthesis, AKP activity, and osteocalcin synthesis [26] . Moreover, TNF exhibits elevated expression and is positively correlated with ADAMTS-7 to exaggerate cartilage degeneration in ONFH [27] . After further exploration, we found that TNF was enriched in SONFH and showed a significant interaction of SONFH.
Matrix metalloproteinase-8 (MMP-8) is included in a family of zinc-dependent proteolytic enzymes and is intensely expressed along to promote and improve osteoblast development into osteocytes [28] . A previous study [29] suggested that genetic variations of the MMP/TIMP system could induce aberrant activation of osteoclasts and cause ONFH. Jieli Du's report considered that genetic variants of MMP8 are conductive to steroid-induced ONFH susceptibility in northern China [30] .
We did not detect genetic polymorphisms of MMP8 in our study, but the decreased expression of MMP8 might be associated with genetic variants.