Cell culture and Infection with a lentivirus encoding luciferase

The human OS cell line 143B was purchased from the American Type Culture Collection (ATCC, Manassas, USA) and cultured in Eagle’s minimum essential medium (EMEM, Sigma Aldrich, UK), supplemented with 10% heat-inactivated fetal bovine serum (FBS, Gibco, Paisley, UK), 0.015 mg/ml 5-bromo-2’-deoxyuridine (BrdU, Sigma Aldrich, St. Louis, MO, USA), 1.0 mM sodium pyruvate (Gibco, Grand Island, USA), and 1% antibiotic-antimycotic (Gibco, Grand Island, USA). The cells were maintained under standard adherent conditions in a humidified incubator with 5% CO2 at 37 °C. Cells were authenticated by immunohistochemistry with antibodies against vimentin (Invitrogen, Thermo Fischer Scientific, Netherlands) and Ki-67 (clone MIB-1; Dako) and haematoxylin and eosin (H&E). The 143B cells were stably transduced with a lentivirus encoding Luciferase as described elsewhere [27].

For secretome (SCR) preparation, 143B cells (25 × 103 cells/cm2) were cultured in an exosome-free culture medium for 24 h. The exosome-depleted medium was prepared by ultracentrifugation of 50% FBS in the appropriate cell culture medium (120 000 g, 16 h). The supernatant was collected after 24 h of incubation, centrifuged to remove cell debris and concentrated with Amicon Ultra-15 Centrifugal Filters (10 kDa molecular weight cut-off, Merck Millipore Ltd, Carrigtwohill, Ireland). Aliquots were stored at -80ºC until usage.

EFEMP1 knockdown

Small interfering RNA (siRNA) knockdown of EFEMP1 in 143B cells was performed using the siGENOME human EFEMP1 siRNA SMARTpool (M-011855-01-0010). At 50% confluency cells were transfected with 20nM siRNA and a nontargeting (NT) control sequence (D-001810-10), both from Dharmacon Inc (Lafayette, USA), using Lipofectamine™ 3000 (Thermo Fisher Scientific, Waltham, USA) according to the manufacturer’s recommendations. After 48 h, the culture medium was replaced by an exosome-free medium, and the cells were maintained in culture for another 24 h afterwards the secretome was collected. The knockdown efficiency was evaluated by western blot and ELISA.

Animal studies

Athymic Swiss nude (Foxn1nu/nu) mice, male or female, (8–12 weeks old, 20–30 g) were purchased from the Institute for Clinical and Biomedical Research (iCBR, Coimbra, Portugal) of the Faculty of Medicine of the University of Coimbra and housed under pathogen-free conditions in individually ventilated cages, with controlled temperature/humidity (22 °C/55%) environment on a 12 h light-dark cycle and with food and water ad libitum.

Premetastatic niche formation

Animals were randomized into two groups. A group was injected subcutaneously into the lower flank with 5 × 105 of 143B-Luc+/100 µL PBS for primary tumour formation. The tumour growth was monitored every two days in two dimensions using a digital calliper, and mice were sacrificed when the tumour reached 50–60 mm3 in volume. Tumour volumes were calculated using the modified ellipsoid formula V = A × B2/3 (A length; B width).

The second group received a daily intraperitoneal (i.p.) injection of 25 µL of the 143B-derived SCR or the vehicle for 1 week, following which they were sacrificed. Animals were euthanized by cervical dislocation, and peripheral blood and lung tissue were harvested for further analysis. Peripheral blood samples were collected to K3EDTA tubes (Greiner bio-one, Kremsmunster, Austria) from anaesthetized animals via cardiac puncture before euthanasia. Plasma was separated by centrifugation at 3000 rpm for 10 min at 4ºC, and aliquots were stored at -80ºC until usage.

Mouse models of Lung Metastasis

For experimental lung metastasis formation, animals were injected intravenously (i.v.) into the lateral tail vein with 1.5 × 106 of 143B-Luc+ cells/100 µL PBS. Another set of animals received daily intraperitoneal injections of 25 µL of the SCR for 1 week prior to the i.v. injection of 143B-Luc+ (1.5 × 106 cells/100 µL PBS). A separate group received the SCR of EFEMP1-knockdown 143B cells, followed by tail vein injection of the 143B-Luc+ cells, as described above.

For spontaneous lung metastasis formation, animals were injected subcutaneously in the lower flank with 5 × 105 of 143B-Luc+ cells/100 µL PBS. After reaching a maximum volume of 60–100 mm3, subcutaneous tumours were surgically excised and the skin incision was sutured with the animals under anaesthesia (10 mg/kg; Rompun 2%, Kiel, Germany).

Animals were monitored weekly for lung metastasis formation during a maximum of 60 days by bioluminescence imaging (BLI) on an IVIS Lumina XR (Caliper Life Sciences Inc., PerkinElmer, Massachusetts, USA). Images were acquired after i.p. injection of D-Luciferin (150 mg/kg), with the animals anaesthetized with 2.5% of isoflurane (Virbac, Carros, France) in 100% O2. Bioluminescent images were analysed using the Living Image software version 4.10 (Xenogen, Alameda, California). A region of interest was drawn around the lesions for the quantification of the bioluminescent signal. Values are expressed photons/sec/cm2/sr. At the end of the study, animals were euthanized by cervical dislocation and peripheral blood and metastatic lungs were collected for histopathological analysis.

Clinical samples

Paraffin-embedded biopsy specimens were obtained retrospectively from 29 patients diagnosed with OS. At the time of diagnosis, all patients had strictly localized disease and were naïve to any chemotherapy. Of the 29 patients, 26 were classified as having conventional high-grade OS while the remaining 3 were classified as low-grade. Except for those with low-grade, all patients underwent chemotherapy prior to surgery, and all received adjuvant chemotherapy. The follow-up period ranged from 1 year to at least 10 years, or until death.

The specific characteristics of the patients are summarized in Table 1.

Table 1 Clinicopathological features of osteosarcoma patientsScanning electron microscopy (SEM)

Small fragments of resected lung tissues and decellularized matrices were fixed with 2% glutaraldehyde and examined in a scanning electron microscope (Flex SEM 1000, HITACHI), under variable pressure scanning using an accelerating voltage of 5–10 kV.

Transmission electron microscopy (TEM)

Resected lung tissues were sectioned into small fragments (1 mm3) and fixed for 2 h in 2.5% glutaraldehyde buffered with 0.1 M cacodylate buffer (pH 7.4), followed by a post-fixation in 1% osmium tetroxide (OsO4) for 1.5 h. After washing, samples were incubated with 1% aqueous uranyl acetate for 1 h, for contrast enhancement. Samples were then dehydrated in a graded acetone series (30–100%) followed by resin embedding using an epoxy embedding kit (Fluka Analytical, Sigma Aldrich, Germany). Ultra-thin sections were obtained with a Leica EM UC6 (Leica Co; Austria) ultramicrotome, mounted on copper grids and stained with lead citrate 0.2% for 7 min. Images were acquired on a Tecnai G2 Spirit Bio Twin electron microscopy at 100 kV (FEI) and AnalySIS 3.2 software.

Histopathological analysis and immunostaining

For murine models, lung tissues and primary tumours were fixed in 4% paraformaldehyde (PFA) and processed for paraffin embedding. Sections of 4 μm were stained with haematoxylin and eosin (H&E, Sigma-Aldrich, St. Louis, MO, EUA) or antibodies against fibronectin (ab2413, Abcam, USA), alpha-smooth muscle actin (α-SMA, ABT 1487, Millipore, Darmstadt, Germany) and vimentin (V9; Ventana, Arizona, EUA). Antigen retrieval was performed by immersing slides in EDTA-Tris buffer (pH 8) for 8 min at 95 ºC and then blocked with a buffered hyper protein solution for 4 min to avoid nonspecific bonds. Immunostaining was performed using a Ventana Marker Platform Benchmark Ultra IHC/ISH with the resource of a multimeric indirect free biotin detection system - Optiview DAB IHC Detection Kit (Ventana Medical Systems, Arizona, EUA), according to the manufacturer instructions. A Gordon’s and Sweet silver staining was performed for the detection of reticulin fibers. Slides were observed under a light microscope Nikon Eclipse 50 I and images were captured with a Nikon-Digital Sight DS-Fi1 camera. Architectural changes were evaluated, and inter-alveolar septal thickness was measured in randomly selected H&E-stained sections.

Clinical samples were formalin-fixed for 24 h and then decalcified with 10% nitric acid for 1–5 days based on tissue hardness, with daily integrity checks. The samples were paraffin-embedded and sectioned into 4 μm tissue slices for immunostaining with anti-EFEMP1/fibulin 3 monoclonal antibody (1:500, ab256457, Abcam). Hematoxylin-eosin staining was performed using standard methods as a counterstain. Immunohistochemistry positivity was considered when at least 1% of the viable neoplastic cells showed cytoplasmic and/or membrane expression of any intensity (weak, moderate or strong). The sections were assessed by an experienced pathologist.

Primary lung fibroblasts isolation and culture

Lungs were harvested, washed in PBS, minced with scissors, and enzymatically digested with 0.1% collagenase A (Roche, Mannheim, Germany) and dispase II (2.4 U/mL; Gibco, Japan) for 90 min at 37 ºC, and then filtered through a 70 μm filter strainer and washed with a physiological saline solution containing 0.05 M EDTA. The obtained cell suspension was plated into 1% gelatin pre-coated dishes and maintained in RPMI 1640 (Sigma-Aldrich, UK) supplemented with 15% FBS and 1% antibiotic-antimycotic at 37 ºC under 5% CO2. All experiments were performed with early passages fibroblasts (P2-P4).

DecellularizationLung fragments

Small lung fragments (2 × 2 mm) were incubated in hypotonic buffer (10 mM Tris HCl/0.1% EDTA, pH 7.8) for 18 h at room temperature, and then washed in PBS (3x, 1 h) and immersed in a detergent solution (0.2% SDS/10 mM Tris HCl, pH 7.8) for 24 h at room temperature. Fragments were washed in hypotonic buffer (3x, 20 min) and incubated in DNAse solution (50 U/ml DNAse/ 10mMTris HCl, pH 7.8) for 3 h at 37 °C. Decellularized matrices were washed in PBS (2x, 20 min) to remove the residual detergent and DNAse solution and were maintained under sterile conditions at 4ºC in PBS until use. All steps were performed under agitation at 165 rpm.

Fibroblasts

Fibroblast cell sheets were incubated with buffer I (1 M NaCl, 5 mM EDTA, 10 mM Tris/HCl pH 7.4) for 1 h at room temperature, washed with PBS (3x, 10 min) and then exposed to buffer II (0.5% [w/v] SDS, 25 mM EDTA, 10 mM Tris/HCl pH 7.4) for 30 min with agitation.

Immunofluorescence

Primary fibroblasts were fixed with 4% paraformaldehyde (PFA, Sigma-Aldrich, St. Louis, MO, USA) for 20 min, permeabilized with 0.2% TritonX-100 (Sigma-Aldrich, St. Louis, MO, USA) for 10 min, and blocked with 3% bovine serum albumin (BSA, Sigma-Aldrich, St. Louis, MO, USA) for 1 h at room temperature. Afterwards, the cells were incubated with the primary antibodies against α-SMA (1:200, ABT 1487; Millipore, Darmstadt, Germany), fibroblast activation protein (FAP) (1:150, PA5-99313, Thermo Fisher Scientific, USA), vimentin (1:200, SP20, Thermo Fisher Scientific, USA) and fibronectin (1:200, ab2413, Abcam, Cambridge, UK) overnight at 4 ºC. The fibroblasts were then incubated with secondary antibody Alexa Fluor 568 or 488 (1:200, Invitrogen, USA) for 1 h at room temperature in the dark, and nuclei were counter with 2 mg/mL Hoechst 33,342 (Sigma-Aldrich, USA).

For immunofluorescence staining with Alexa Fluor 555 phalloidin, cells were fixed with PBS containing 4% sucrose and 4% paraformaldehyde for 10 min, permeabilized with 1% TritonX-100 for 10 min and blocked with 1% BSA for 45 min at room temperature , and then incubated with Alexa Fluor 555 phalloidin (1:40, Abcam, Cambridge, MA, USA) for 30 min at RT. Nuclei were stained with 2 mg/mL Hoechst 33,342 (Sigma-Aldrich, Buchs, Switzerland). The chamber slides were mounted in the Vectashield Mounting Medium (Vectashield, Vector Laboratories, United States).

Decellularized lung fragments and fibroblast cell sheets were processed for immunofluorescence as described above and stained with the primary antibodies against fibronectin (1:200) and collagen IV (1:200 ab19808; Abcam, Cambridge, UK) and then incubated with secondary antibody Alexa Fluor 488 (1:200, Invitrogen, USA). Images were captured in Carl Zeiss Axio Observer Z1 inverted microscope (Carl Zeiss, Thornwood, NY) and processed using the ImageJ 1.52p software (National Institutes of Health, USA).

Cell migration

Cell migration was analysed using the wound-healing assay. Confluent monolayers of fibroblasts were manually scratched with a 200 µL sterile pipette tip. Photographs were taken immediately after scratching (baseline) and at 6 and 24 h in a Carl Zeiss Axio Observer Z1 inverted microscope (Carl Zeiss, Thornwood, NY). The wound closure at each time-point was quantified using ImageJ 1.52p software (National Institutes of Health, USA) and normalized to the baseline.

Lung transcriptomic analysis

Total RNA was extracted from lung tissues using TRIzol reagent (Ambion by Life Technologies, USA) according to the manufacturer’s instructions. RNA was eluted in 30–40 µL RNase-free water. The concentration and quality of extracted RNA were determined with a NanoDrop spectrophotometer (Thermo Scientific, Wilmington, USA). RNA was stored at 80 ºC until use. Single-end sequencing was performed using the library prep kit TruSeq and the sequencing kit NovaSeq6000 SP Flowcell 100 cycles (Illumina, Inc.) for Mus musculus (Ensembl.GRCm38.82). The transcriptomic sequencing of RNA was performed on the Illumina NovaSeq 6000 instrument, at the VIB Nucleomics Core (www.nucleomics.be). The reads pre-processing was performed by VIB Nucleomics Core and involved: quality trimming (FastX 0.0.14, HannonLab), adapter trimming (cutadapt 1.15) [28], quality filtering (FastX 0.0.14 and ShortRead 1.40.0) [29] and removal of contaminants (bowtie 2.3.3.1). The pre-processed reads were then aligned to the reference genome of Mus_musculus.Ensembl.GRCm.38.82 using STAR 2.5.2b [30] and SAMtools 1.5 [31] Bioconductor package. The expression levels of the read overlapping genes were computed using the EDASeq package [32] for the within-sample and between samples normalizations. The differentially expressed genes were estimated by fitting a negative binomial generalized linear model using the edgeR 3.24.3 package [33]. The resultant p-value was corrected for multiple tests with Benjamini-Hochberg to control the false discovery rate (FDR).

Unless stated otherwise, the following bioinformatics analyses were conducted using the R programming language (version 4.1.1) in the RStudio integrated development environment. The volcano plots were achieved using the ggplot2 [34] (version 3.3.5) and ggrepel (version 0.9.1) packages. The Venn diagrams were generated with the webtool available at bioinformatics.psb.ugent.be/webtools/Venn/. The clusterProfiler [35] version 4.0.5) package was used to perform the gene set enrichment analyses (GSEA) for the Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Genes showing a p-value higher than 0.05 were excluded from the gene set. Heat maps were made with the ComplexHeatmap [36] (version 2.8.0) package. Protein-protein interaction (PPI) network of DEGs was visualized using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (http://string-db.org/). Graphics were refined and figures were assembled in Adobe Illustrator 2019 (version 23.1.1).

Flow cytometry

Fresh lung tissue was minced with fine scissors and subjected to mechanical dissociation (GentleMACS, Miltenyi Biotech, Germany). The tissue homogenates were filtered gently through a 40 µm cell strainer and centrifuged at 600 g for 15 min at RT. The isolated single-cell suspension and peripheral leukocytes were fluorescently stained with anti-mouse CD11b-FITC (1:200, BioLegend, 101205) and Ly6G-PE (1:80, BioLegend, 127607), CD68-APC (1:300, BioLegend, 137007), CD19-PerCP (1:200, BioLegend, 115531) and CD45-PB (1:200, BioLegend, 103125). FACS Lysing solution (BD Sciences) was added to the samples for lysing red blood cells. Data were collected on a FACS Canto™ II (BD Biosciences, USA) and analysed using Infinicyt V.1.8 software (Cytognos, Salamanca, Spain).

RNA extraction and qRT-PCR

Total RNA from lung tissue was extracted using TRIzol reagent (Ambion by Life Technologies, USA) according to the manufacturer’s instructions. The concentration and quality of extracted RNA were determined using a NanoDrop spectrophotometer (Thermo Scientific, Wilmington, USA). First-strand complementary deoxyribonucleic acid (cDNA) was synthesized from 2 µg of total RNA using the NZY first-strand cDNA synthesis kit. Quantitative real-time PCR analysis was performed using a Xpert SYBR Green MasterMix (GRISP, Porto, Portugal) on a CFX Connect Real-Time PCR Detection System (Bio-Rad Laboratories, Inc). Gene expression was normalized to the housekeeping genes glyceraldehyde 3-phosphate dehydrogenase (GADPH) and hypoxanthine-guanine phosphoribosyltransferase 1 (HPRT-1) ranked as stably expressed by the RefFinder algorithm (https://www.heartcure.com.au/reffinder/). The relative expression of target genes was calculated based on the 2−∆∆Ct method, where ΔΔCt=(Ct target−Ct housekeeping)Sample−(Ct target−Ct housekeeping)Control. Primer sequences (Eurofins Genomics, Lisboa, Portugal) are listed in Table 2. A Primer-BLAST search was performed to evaluate primer specificity and self-complementarity values of already validated and published primer sequences.

Table 2 Primer sequences used in qRT-PCRWestern blot

Total protein extracts were prepared using standard lysis buffer, separated by sodium dodecyl sulfate/polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to PVDF membranes (Bio-Rad, Hercules, CA, USA). Membranes were blocked with 5% of bovine serum albumin in Tris-buffered saline-tween 20 (TBS-T) and were incubated with appropriate primary antibodies fibronectin (1:500; ab2413; Abcam, Cambridge, UK), collagen IV (1:500; ab19808; Abcam, Cambridge, UK), EFEMP1 (1:1000, ab256457, Abcam), ITG β1 (1:1000, D6S1W, Cell Signalling), ITGα6 (1:1000, ab181551, Abcam) and CD44/HCAM (1:200,sc-7297, Santa Cruz Biotechnology), overnight at 4 °C. Afterward, the membranes were washed in 0.1% TBS-Tween and incubated with the secondary antibody (1:10000) for 1 h at RT and revealed by chemiluminescence (Clarity ECL western blotting substrates, BioRad) using ImageQuant LAS 500 chemiluminescence CCD camera (GE Healthcare Bioscience AB). Images were analysed using ImageJ software (National Institutes of Health).

Adhesion assays

The 96-well plates were coated with fibronectin or collagen at concentrations ranging from 1 to 10 µg/mL for 2 h. The 143B-luc+ cells were seeded onto the substrates and allowed to adhere for 10 min at 37 °C. To remove the non-adherent cells, the wells were washed 3 times with PBS containing 1 M CaCl2.H2O and MgCl2.6H2O, and then 100 µL of EMEM culture medium containing 0.3 mg/mL of D-Luciferin was added to each well. The plate was read in the IVIS optical imaging system, and the bioluminescent signal was quantified using Living Image Software 4.10 (Xenogen, Alameda, CA, USA).

To evaluate the 143B cell adhesion to the decellularized lung scaffold, the decellularized fragments were placed in the 96-well plates and the 143B cells were seeded on top. The cells were allowed to adhere for 10 min at 37 °C and the same protocol as described above was followed to quantify cell adhesion.

Mass spectrometryProtein extraction

Proteins were reduced and alkylated with 100 mM Tris pH 8.5, 1% sodium deoxycholate, 10 mM tris(2-carboxyethyl)phosphine (TCEP), and 40 mM chloroacetamide for 10 min at 95ºC at 1000 rpm (Thermomixer, Eppendorf). 100 µg of protein were processed for proteomic analysis following the solid-phase-enhanced sample-preparation (SP3) protocol as described by Hughes et al. [37]. Enzymatic digestion was performed with Trypsin/LysC (2 µg) overnight at 37ºC at 1000 rpm. The resulting peptides were cleaned-up and desalted with C18 micro columns and further quantified.

Proteomics data acquisition

Protein identification was performed by nanoLC-MS/MS using an Ultimate 3000 liquid chromatography system coupled to a Q-Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Scientific, Bremen, Germany). Peptides were loaded onto a trapping cartridge (Acclaim PepMap C18 100Å, 5 mm x 300 μm i.d., 160,454, Thermo Scientific) in a mobile phase of 2% ACN, 0.1% FA at 10 µL/min. After 3 min loading, the trap column was switched in-line to a 50 cm by 75 μm inner diameter EASY-Spray column (ES803, PepMap RSLC, C18, 2 μm, Thermo Scientific, Bremen, Germany) at 300 nL/min. Separation was achieved by mixing A: 0.1% FA, and B: 80% ACN, with the following gradient: 5 min (2.5% B to 10% B), 120 min (10% B to 30% B), 20 min (30% B to 50% B), 5 min (50% B to 99% B) and 10 min (hold 99% B). Subsequently, the column was equilibrated with 2.5% B for 17 min. Data acquisition was controlled by Xcalibur 4.0 and Tune 2.9 software (Thermo Scientific, Bremen, Germany).

The mass spectrometer was operated in data-dependent (dd) positive acquisition mode alternating between a full scan (m/z 380–1580) and subsequent HCD MS/MS of the 10 most intense peaks from the full scan (normalized collision energy of 27%). ESI spray voltage was 1.9 kV. Global settings: use lock masses best (m/z445.12003), lock mass injection Full MS, chrom. peak width (FWHM) 15s. Full scan settings: 70k resolution (m/z 200), AGC target 3e6, maximum injection time 120 ms. dd settings: minimum AGC target 8e3, intensity threshold 7.3e4, charge exclusion: unassigned, 1, 8, > 8, peptide match preferred, exclude isotopes on, dynamic exclusion 45s. MS2 settings: microscans 1, resolution 35k (m/z 200), AGC target 2e5, maximum injection time 110 ms, isolation window 2.0 m/z, isolation offset 0.0 m/z, spectrum data type profile.

Data analysis

The raw data was processed using the Proteome Discoverer software (Thermo Scientific) and searched against the UniProt database for Homo sapiens Proteome. The Sequest HT search engine was used to identify tryptic peptides. The ion mass tolerance was 10 ppm for precursor ions and 0.02 Da for fragment ions. The maximum allowed for missing cleavage sites was set to 2. Cysteine carbamidomethylation was defined as a constant modification. Methionine oxidation and protein N-terminus acetylation were defined as variable modifications. Peptide confidence was set to high. The processing node Percolator was enabled with the following settings: maximum delta Cn 0.05; decoy database search target FDR 1%, validation based on q-value. Imputation of missing values was performed only when a peptide was detected in at least one of the replicates analyzed. Quantitative evaluation was performed by pairwise comparisons of all detected peptides and the median ratio was used for protein level comparison. Significance assessment was performed using the background-based ANOVA method implemented in Proteome Discoverer 2.2 and multiple comparison adjustment of the p-values was performed.

Protein functional enrichment analysis

The protein functional enrichment analyses for Gene Ontology (GO) terms and Reactome pathways were performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) 6.8 bioinformatics tool (https://david.ncifcrf.gov/summary.jsp). The data analysis and visualization of differently expressed proteins were conducted in the R programming language (version 4.1.1) in the RStudio integrated development environment. The GO and pathways plots were achieved using the ggplot2 (version 3.3.5). A basic circle packing chart with parckcircles package (available at https://r-graph-gallery.com/305-basic-circle-packing-with-one-level.html) was used for GO – biological processes. The Venn diagrams were generated with the webtool available at bioinformatics.psb.ugent.be/webtools/Venn/.

Enzyme-linked immunosorbent assay (ELISA)

Levels of EFEMP1 ELISA in mouse serum samples and in the SCR of wild-type, MG-63 cells, 143B cells and EFEMP1-knockdown 143B cells were determined using the Human EFEMP1 ELISA kit (Abcam, ab269552) according to the manufacturer’s instructions.

Kaplan-Meier analysis

The Kaplan-Meier analysis was performed using the R2 database (R2: Genomics Analysis and Visualization Platform – http://r2platform.com) which contains genome-wide gene expression data of OS patient samples (dataset: Mixed OS (Mesenchymal) – Kuijjer – 127 – vst – ilmnhwg6v2). The chondroblastic, fibroblastic and osteoblastic OS patient samples (n = 73) or with metastatic disease (n = 37) in the database were divided into high and low EFEMP1 (ID: 2202) expressions based on scan cut-off.

Statistical analysis

All data are expressed as mean ± standard error of the mean (SEM). Graphics and statistical analysis were performed using GraphPad Prism version 8.0.2 (GraphPad Software, San Diego, CA, USA). Independent variables were analysed by the Mann-Whitney test, whereas Kruskal-Wallis or one-way ANOVA was used for multiple comparisons. Statistical significance was set at the level of p < 0.05. For the Kaplan-Meier analysis of OS patient samples, statistical differences in survival curves were calculated by log-rank (Mantel-Cox) test. Figure illustrations were created with adobe illustrator or BioRender.com.