Using the GEO database (ncbi.nlm.nih.gov/geo/), three gene expression level profiles [GSE37552(23), GSE9508(24) and GSE19357] were selected based on the platforms GPL570 [(HG-U133_Plus_2) Affymetrix Human Genome U133 Plus 2.0 Array], GPL6076 (Agilent-Whole Human Genome Oligo Microarray G4112A condensed) and GPL6244 (HuGene-1_0-st) Affymetrix Human Gene 1.0 ST Array [transcript (gene) version], respectively.

Using the GEO2R (25) online analysis tool (ncbi.nlm.nih.gov/geo/geo2r/), DEGs were identified between osteosarcoma and normal tissue. The DEGs were calculated with thresholds of P<0.05 and |log fold-change (FC)|>1.00. Intersections between each dataset were obtained using the Venn diagram tool (bioinformatics.psb.ugent.be/webtools/Venn/).

GO (26) annotations were downloaded from GO. KEGG (27) pathway enrichment analysis of DEGs was applied to determine enriched signaling pathways. Metascape tool (metascape.org/gp/index.html) was used for annotation. P<0.01 and gene count >3 were considered to indicate a statistically significant result.

A PPI network was constructed using STRING (V11.5, Swiss) database (https:string-db.org/). Visualization of PPI networks was performed using Cytoscape software (V3.7.2, BioStar team; https://apps.cytoscape.org/apps/mcode). Hub genes from DEGs were further defined according to module connectivity in the PPI network. Highly interconnected nodes indicated greater essential connectivity. CytoHubba (28) add-on to the aforementioned Cytoscape was used to determine the connectivity of each node measured by degree, density of maximum neighborhood component (DMNC), maximal clique centrality (MCC) and mononuclear cell counts (MNC).

A Cytoscape plugin, MCODE (29), was used to identify significant molecular complexes in the PPI network. Corresponding networks of gene modules were annotated using the UniProt database (uniprot.org/).

Clinical and gene expression level data were analyzed using the Kaplan-Meier plotter (kmplot.com/analysis/) database. The prognostic values of hub genes were analyzed in patients with osteosarcoma. All cases were grouped according to the median value of mRNA expression levels. P<0.05 was considered to indicate a statistically significant difference.

GEPIA (gepia.cancer-pku.cn/), was applied to analyze RNA sequencing expression level data. The Oncomine (oncomine.org/resource/main.html) database was used to confirm the expression levels of the eight hub genes.

Osteosarcoma cell lines MG-63 and U-2OS were obtained from Cell Resource Center, Shanghai Institutes for Biological Sciences at the Chinese Academy of Sciences. The MG-63 cells and U-2OS cells were cultured in minimal essential medium (MEM) and McCoy's 5A medium (Hyclone; Cytiva), respectively, supplemented with 10% FBS (Gibco; Thermo Fisher Scientific, Inc.), 1% penicillin and streptomycin (Hyclone; Cytiva) under a controlled atmosphere with 5% CO₂ at 37˚C and relative humidity of 85-95%.

Total RNA in cells was extracted using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.). The concentration of RNA samples was then measured using a spectrophotometer. Total RNA was reverse-transcribed into first-strand complementary DNA using a Primescript RT reagent kit according to the manufacturer's protocol (Takara Bio, Inc.). For qPCR, 2X SYBR Green qPCR Master Mix (Suzhou Yuheng Biological Technology Co., Ltd.) was used on an ABI 7900 system (Applied Biosystems; Thermo Fisher Scientific, Inc.). The primer sequences used were as follows: CENPF forward, 5'-CGTCCCCGAGAGCAAGTTTATT-3', and reverse, 5'-ACTGCCTTTGCTGCTTTTCC-3'; GAPDH forward, 5'-GCTCTCTGCTCCTCCTGTTC-3', and reverse, 5'-CGACCAAATCCGTTGACTCC-3'. Amplification conditions were as follows: 95˚C for 30 sec, followed by 40 cycles at 95˚C for 15 sec and 60˚C for 45 sec. The relative change in mRNA expression levels was calculated according to the 2^-ΔΔCq method (30).

Small interfering (si)RNA1 (si-CENPF-1), siRNA2 (si-CENPF-2) and negative control (NC) siRNA (non-targeting sequence) were purchased from Sangon Biotech Co., Ltd. MG-63 and U-2OS cells were incubated at 37˚C with 5% CO₂ for 24 h, and subsequently plated and transfected with 3.125 nm of si-NC, si-CENPF-1 or si-CENPF-2 at a density of 3,000 cells/well for use in the CCK-8 assay. Moreover, a total of 1x10⁵ cells/well were plated and transfected with 50 nm si-NC, si-CENPF-1 or si-CENPF-2 for the wound healing and Transwell assays. All transfections were carried out using LipoHigh transfection reagent (Sangon Biotech Co., Ltd.) for 4 h after plated for 16-24 h. The cells were collected after 48 h for use in subsequent experiments. The sequences were as follows: siRNA NC, 5'-TTCTCCGAACGTGTCACGTdTdT-3'; si-CENPF-1 (Sense), 5'-GCAGAATCTTAGTAGTCAA-3'; and si-CENPF-2 (sense), 5'-GCAACCATCTACTTGAAGA-3'. The level of transfection efficiency was higher for si-CENPF-2 compared with that of si-CENPF-1, and it was therefore selected for subsequent assays.

The protein expression levels of CENPF were detected by western blotting. The transfected MG-63 and U-2OS cells were washed with pre-cooled PBS on ice and then boiled for 10 min in SDS-sample buffer (Beyotime Institute of Biotechnology). The concentration of protein was determined using a BCA assay according to the manufacturer's instructions (Sangon Biotech Co., Ltd.), and the absorbance was read at 570 nm using Thermo Multiskan FC (version, US6111636; Thermo Fisher Scientific, Inc.). A total of 30 µg protein lysates/per lane were separated by electrophoresis with 10% SDS-PAGE. Afterwards, the samples were transferred to PVDF membranes. Membranes were blocked for 1 h with 5% skimmed milk at room temperature, and then incubated at 4˚C overnight with the following primary antibodies from ABclonal Biotech Co., Ltd.: GAPDH (1:1,000; cat. no. AC002) and CENPF (1:1,000; cat. no. A18644). Subsequently, membranes were incubated at 37˚C for 1 h with the GAPDH HRP goat anti-mouse IgG (1:10,000; cat. no. AS003) and CENPF goat anti-rabbit IgG (1:10,000; cat. no. AS014) secondary antibodies from ABclonal Biotech Co., Ltd. Bands were visualized using chemiluminescence HRP Substatet (MilliporeSigma). The Bio-Rad gel Doc XR + system (Bio-Rad Laboratories, Inc.) was used to scan the gel images. ImageJ (version, 1.53e; National Institutes of Health) was used for grayscale analysis. GAPDH was used as an internal control.

Cell Counting Kit-8 (CCK-8) assays (Beyotime Institute of Biotechnology) were performed according to the manufacturer's protocol. MG-63 and U-2OS cells were seeded in 96-well plates at a density of 3,000 cells/well and transfected at 16-24 h, and subsequently cultured at 37˚C in a 5% CO₂ incubator. CCK-8 was added to each well at 0, 24, 48 and 72 h after transfection. Following incubation for 1 h at 37˚C, the absorbance at 450 nm was read using Thermo Multiskan FC (version, US6111636; Thermo Fisher Scientific, Inc.). A total of three independent repeats were performed.

MG-63 and U-2OS cells were seeded in 6-well plates at a density of 1x10⁵ cells/well, cultured for ~16-24 h and transfected. Cells were used at ~80-90% confluence on the day of transfection. LipoHigh transfection reagent (5 µl/well; Sangon Biotech Co., Ltd.) and corresponding RNA (100 pmol/well) were added. After transfection for 4-6 h, cells were uniformly scratched using a sterile pipette, washed with PBS and then respectively cultured in MEM and McCoy's 5A medium (Hyclone; Cytiva) with 2% FBS for 72 h at 37˚C in a 5% CO₂ incubator. The wound images were observed and captured using an optical microscope at (magnification, x40; version, CKX41; Olympus Corporation) at 0, 24, 48 and 72 h. The quantification data for wound closure assay was calculated using the following formula: Distance wound closed=initial wound width-final wound width; wound closure rate=distance wound closed/initial wound width.

Transwell migration assays were performed, and a total of 1x10⁴ MG-63 and U-2OS cells were seeded into the top chamber in serum-free medium, and medium containing 10% FBS was placed in the bottom chamber. Migrated cells were fixed in 4% polyformaldehyde for 30 min at room temperature, and subsequently stained with 0.1% crystal violet for 20 min at room temperature. Five views per group were selected and images were captured at x100 magnification using a light microscope. A total of three independent repeats was performed.

Transwell invasion assays were also performed using Millipore Transwell chambers. Transfected cells were resuspended in serum-free MEM or McCoy's 5A medium and placed in the upper chamber, previously coated with Matrigel (BD Biosciences) at 37˚C for 6 h, while MEM or McCoy's 5A medium containing 10% FBS was added to the lower chamber. Following incubation for 24 h at 37˚C in a 5% CO₂ incubator, cells in the upper membrane were removed with cotton swabs, whereas invaded cells were fixed in 4% polyformaldehyde for 30 min at room temperature and stained with 0.1% crystal violet for 20 min at room temperature. Five views per group were selected and images were captured at x100 magnification using an optical microscope. A total of three independent repeats was performed.

All data were analyzed with GraphPad Prism 7.0 software (GraphPad Software, Inc.) and are expressed as the mean ± standard deviation. Paired student's t-test was used to analyze differences between two groups. One-way ANOVA followed by Dunnett's post hoc multiple comparisons test were used to analyze differences between >2 groups. All experiments were performed in triplicate. P<0.05 was considered to indicate a statistically significant difference.

A total of three gene expression level profiles were selected. GSE37552 consisted of two metastatic samples and two non-metastatic samples; GSE9508 included 13 samples from non-metastatic patients and 21 samples from metastatic patients; GSE19357 contained five frozen bone tumor biopsies (two osteochondromas, two cases of unusual parosteal osteochondromatous proliferation and one subungual exostosis). In total, 52 DEGs were identified in accordance with the criteria of P<0.05 and |logFC|>1: Seven overlapping genes in GSE37552 and GSE19357; 13 overlapping genes in GSE37552 and GSE9508; 31 overlapping genes in GSE19357 and GSE9508; and one overlapping gene in GSE37552, GSE9508 and GSE19357. Venn diagram analysis presents the intersection of DEGs (Fig. 1). Subsequently, seven overlapping DEGs were selected from three gene expression profiles.

The online analysis tool Metascape was applied to identify GO categories and KEGG pathways for DEGs. The results indicated that the terms ‘regulation of muscle tissue development’, ‘positive regulation of cell migration’, ‘bone development’ and ‘resolution of sister chromatid cohesion’ were primarily enriched with DEGs (Fig. 2).

Protein interactions between DEGs were determined using STRING in the PPI network. A total of 52 nodes and 80 edges were included (Fig. 3A). Hub genes were identified based on the connectivity of degree, DMNC, MCC and MNC in the PPI network (Fig. 3B). Venn analysis was performed to determine the intersection of the DEG profiles, which indicated that five genes were overlapping in those four analysis methods (degree, DMNC, MCC and MNC; Fig. 3C). Additionally, three module networks were analyzed by MCODE. The most important module with six genes was selected (Fig. 3D). Collectively, the results of the union of set were calculated according to the methods of hub genes selection and eight genes [DEP domain containing 1 (DEPDC1), TIMELESS interacting protein (TIPIN), shugoshin 1 (SGOL1), origin recognition complex subunit 6 (ORC6), IGF-binding protein 5 (IGFBP5), minichromosome maintenance 10 replication initiation factor (MCM10), MET proto-oncogene, receptor tyrosine kinase (MET) and CENPF] were identified. Thus, eight hub genes were obtained using Cytohubba and MCODE analysis.

Kaplan-Meier plotter online database provided the prognostic values of eight hub genes. Overall survival analysis of 259 cases indicated that with the exception of IGFBP5, high expression levels of the hub genes (DEPDC1, TIPIN, SGOL1, ORC6, MCM10, MET and CENPF) were significantly associated with unfavorable overall survival in patients with osteosarcoma (Fig. 4). The survival curve of eight hub genes exhibited significant differences between low and high expression levels.

In order to confirm the differential hub gene expression levels, expression level profiles were constructed based on GEPIA, which demonstrated that the hub genes ORC6 and CENPF were significantly upregulated in osteosarcoma (Fig. 5A). In order to verify the results, the expression levels of hub genes were collected in Oncomine database, which indicated that IGFBP5 and MET were downregulated in tumor tissues, whereas the other hub genes [DEPDC1, TIPIN, SGOL1(SGO1), ORC6, MCM10 and CENPF] were upregulated in tumor tissues, compared with the adjacent non-tumorous tissues (Fig. 5B). The expression of eight hub genes exhibited a significant difference between tumors and adjacent tissues.

As aforementioned, ORC6 and CENPF were upregulated in osteosarcoma. Furthermore, Kaplan-Meier plotter indicated that the P-value of CENPF was smaller compared with that of ORC6. Therefore, CENPF was selected for further experiments analyses.

In order to select the most effective siRNA for CENPF, transfection efficiency was assessed using RT-qPCR and western blotting. MG-63 and U-2OS cells were transfected with NC, si-CENPF-1 or si-CENPF-2. The results demonstrated that CENPF expression at both the mRNA and protein levels were significantly downregulated in both cell lines transfected with si-CENPF-2 compared with the NC (^*P<0.05, ^**P<0.01; Fig. 6A-D). Thus, si-CENPF-2 was selected for the following functional experiments.

In order to analyze the functional role of CENPF in osteosarcoma, CCK-8 assays were performed in MG-63 and U-2OS cell lines. NC and si-CENPF-2 were separately transfected into osteosarcoma cell lines. Proliferation ability was significantly decreased in the si-CENPF-2 group compared with the corresponding NC group (^*P<0.05; Fig. 7A and B). This indicated that CENPF knockdown inhibited the proliferation ability of osteosarcoma cell lines.

Wound healing and Transwell migration assays were performed to investigate the effect of CENPF on migration in osteosarcoma cell lines. Cells were transfected with NC and si-CENPF-2 and migration ability was then investigated. In osteosarcoma cell lines, the wound healing rate was significantly decreased in the si-CENPF-2 groups compared with the corresponding NC groups (^*P<0.05, ^**P<0.01; Fig. 8B and D). The results demonstrated that CENPF knockdown decreased migration ability in osteosarcoma cell lines.

In order to measure invasion ability, Transwell invasion assays were performed in osteosarcoma cell lines using Matrigel-coated Transwell inserts. Cell lines were transfected with NC or si-CENPF-2. The results demonstrated that cells transfected with si-CENPF-2 exhibited significantly lower invasion abilities compared with the corresponding NC groups (^**P<0.01, ^***P<0.001; Fig. 9). This indicated that downregulated CENPF significantly decreased invasion ability in osteosarcoma cell lines.