MCAM is associated with metastasis and poor prognosis in osteosarcoma by modulating tumor cell migration

Abstract Background Although there are standard treatment options for osteosarcoma (OS), the prognoses of patients with OS remain varied. Therefore, it is important to profile OS patients at a high risk of mortality to develop focused interventions. Although tumor biomarkers are closely associated with clinical outcomes, data on prognostic biomarkers for OS remain scarce. Methods We collected RNA expression profiles and clinical data of 90 OS patients from the GEO database (dataset GSE21257 and GSE39055) and 96 patients in the TARGET program. The data were analyzed using univariate Kaplan‐Meier survival analysis to screen candidate gene sets that might be associated with OS survival. Results Our analysis demonstrated that melanoma cell adhesion molecule (MCAM) was associated with overall survival of patients with OS in the three cohorts. The data showed that MCAM was upregulated in OS patients who had metastases within 5 years compared to those without metastases. GO analysis revealed that genes correlated with MCAM were mainly involved in cell migration and wound healing processes. In addition, wound healing assays and gene set enrichment analysis results from RNA sequencing data of small interfering (si)‐MCAM‐transfected OS cells demonstrated that MCAM modulated tumor cell migration. Conclusions Our data demonstrate that MCAM may be a novel prognostic biomarker for OS. MCAM is associated with increased cell migration ability and risk of metastasis, thus leading to poor prognoses in OS patients.

interventions and improve clinical outcomes. Previous data have shown that assigned male sex, tumor at the axial site, large tumor size, poor response to initial treatments, elevated alkaline phosphatase levels, metastases, pathological fractures, and <90% tumor necrosis after neoadjuvant chemotherapy are associated with poor OS outcomes. [5][6][7] Advanced studies have highlighted that genetic tumor biomarkers may be closely associated with clinical outcomes. For instance, in prostate tumors, a three-gene panel of FGFR1, PMP22, and CDKN1A can accurately predict the risk of tumor recurrence. 8 In OS, increased expression of APE1 and MDR1 was shown to be associated with poorer OS prognosis. 9,10 In addition, a risk signature of three survival-associated genes, MYC, CPE, and LY86, could discriminate between low-and high-mortality risk in OS patients. 11 Unfortunately, data on prognostic biomarkers for OS remain unsatisfactory for any clinical use.
Using global gene expression profiles in multiple OS cohorts, The GEO (https://www.ncbi.nlm.nih.gov/geo/) is a gene expression database established by the National Center for Biotechnology Information (USA) containing array-and sequence-based data 12 The subsets of OS cohorts in the GEO database were searched using the key words 'osteosarcoma,' 'RNA,' and 'survival.' The GSE21257 dataset contained clinical information and gene expression data of 53 OS patients. 13 The patient data included survival status, overall survival time, presence or absence of metastasis, and expression of 24,998 genes in biopsy samples. The GSE39055 dataset included 37 OS patients and data on survival status, overall survival time, and expression of 20,819 genes in the biopsy tissue samples. 14 In both the GSE21257 and GSE39055 datasets, the OS tissue samples were obtained via biopsy prior to chemotherapy.
In addition, GSE16088 with 14 human OS tissue samples and 6 normal tissue samples (2 kidney samples, 2 liver samples, and 2 lymph node samples) 15 ; GSE14359 with 18 human OS tissue samples (8 men and 10 women, age 31 ± 19.9) and 2 primary non-neoplastic osteoblast cell samples 16 ; and GSE52063 with 4 mesenchymal stem cell samples, 4 osteosarcoma stem cell samples, and 4 adherent osteosarcoma cell samples were also identified. 17 These datasets included gene expression data but not clinical and survival data. Therefore, the datasets were used to profile the expression of screened candidate genes in the OS and normal samples.
The TARGET program (https://ocg.cancer.gov/progr ams/target) incorporates multiple tumor projects, such as those for acute lymphoblastic leukemia, acute myeloid leukemia, kidney tumors, neuroblastoma, and OS. 18 We obtained clinical data and tissue samples in the OS project of the TARGET program from patients who were recruited in OS biopsy studies or clinical trials. 18

| Screening of survival-related candidate genes
Here, the overall survival time was defined as the time between the establishment of a clinical diagnosis of OS and death from all causes.
Using Kaplan-Meier (KM) survival analysis, univariate survival analysis was performed for each gene in the GEO GSE21257, TARGET, and GEO GSE39055 datasets. Genes with a p-value of <0.01 in the KM survival analyses were selected as candidate genes. By determining overlapping significant candidate genes among the three datasets, gene sets potentially associated with the survival of OS patients were identified. The KM survival analysis was performed using the "survival" package in R software (version 3.6.2,). Based on the median expression of each candidate gene, we classified patients in high expression (higher than the median) or low expression (lower than the median) groups.

| Expression, co-expression, and functional enrichment analyses
Gene expression profiles in OS cell lines were obtained from the Cancer Cell Line Encyclopedia (CCLE) (www.broad insti tute.org/ ccle). Biological networks of the top 10 genes that were correlated with the candidate genes were constructed using STRING (version 11.0, https://strin g-db.org) and CytoHubba (Cytoscape, version 3.7.2, https://cytos cape.org/). Similarly, the top 100 co-expression genes were extracted from the GEPIA2 database (http://gepia2.cance rpku.cn), which contains data on genes and their functions in various human cancers.
Cluster analysis was performed using the "ConsensusClusterPlus" package in R software. Gene Ontology (GO) functional enrichment analysis of the gene sets was performed using Metascape (http://metas cape.org) while Gene Set Enrichment Analysis (GSEA) (https://www.gsea-msigdb.org/gsea/index.jsp) for GO sets was performed using the "clusterProfiler" and "enrichplot" packages in R software.

| Cell culture and transfection
The human OS cell line 143B was purchased from ZhongQiaoXinZhou Biotechnology (NO.ZQ0455). The cells were routinely cultured in complete culture medium (ZhongQiaoXinZhou, NO.ZQ-303) at 37°C in a 5% CO 2 atmosphere.
Transfection was performed at a cell density of 60% in a six-well Thereafter, the effect of siRNA transfection was examined using western blot analysis, and the cells were used for subsequent experiments.

| Western blot analysis
Total proteins were extracted from the cells using radio immunopre-

| Wound healing assay
The cells transfected with si-MCAM or si-con were cultured in a 6well plate (2.5 × 10 6 cells/well). Using a 100 µl plastic pipette tip, each well was scratched. The cells were washed three times with PBS and cultured for 24 h. Using an inverted microscope, we captured images of the wounded area immediately after the scratch and after 24 h (magnification × 10). The gap area of each culture well was measured using the Image-Pro Plus program (version 7.0, Media Cybernetics), and the mean values from the wells were obtained for analysis.

| RNA-Seq analysis
The cells were lysed using TRIzol (Invitrogen, No. 15596018) and prepared for RNA-Seq. RNA-Seq procedures including quality inspection, database construction, sequencing, mapping, and preliminary analysis were commissioned from BGI Company. We performed differential expression analysis using the "limma" package in R software with an adjusted p-value of <0.05 and |log 2 FC|≥1.

| Statistical analysis
KM univariate survival analysis was used to screen survival-related candidate genes in OS. Quantitative data are presented as the mean ± SD or mean ± SE. Unpaired two-tailed t tests were used for the statistical analysis.

| Gene expression signature in OS biopsies revealed potential prognostic values
To identify possible subgroups with diverse gene expression signatures, we analyzed the gene expression profiles of 53 OS samples from GSE21257. Unsupervised clustering analysis using "ConsensusClusterPlus" revealed two distinct subgroups in GSE21257 GSE39055, n = 37) and the TARGET program (n = 96) ( Table 1).

| Higher MCAM expression associated with poor OS prognosis
Among the genes that were expressed in the tissue samples from the GSE21257 OS cohort, 359 were significantly associated with

| MCAM knockdown impaired OS cell migration
The expression levels of MCAM in OS cell lines were determined using the Broad Institute CCLE ( Figure 3A). siRNA was used to knockdown MCAM in the human OS cell line 143B. A concentration of 50 nM MCAM siRNA, with a relatively high knockdown efficiency, was used in experiments to evaluate the migration ability and facilitate gene expression analysis in OS cells ( Figure 3B). A wound healing assay was performed to assess the migration ability of cells in vitro. As presented in Figure 3C and 3D, MCAM knockdown significantly impaired OS cell migration.

| MCAM modulates multiple biological processes in OS
To understand the possible mechanisms of MCAM in OS prognosis, In addition, in the si-MCAM cells, 106 and 43 genes were significantly up-and downregulated, respectively (|log 2 FC|≥1, adjusted p-value <0.05, Figure 4D). GO analysis with Metascape of the differentially expressed genes revealed that genes responsible for wound healing, positive cell death regulation, apoptotic signaling pathway function, cell proliferation, and cytokine-mediated signaling pathway function were significantly downregulated in si-MCAM cells, whereas no GO terms were enriched for upregulated genes ( Figure 4E).

| Roles of MCAM in other cancers
The possible roles of MCAM in other cancers were further inves-  Figure S1).
In addition, survival analyses demonstrated that higher MCAM expression was significantly associated with worse prognosis in patients with brain low-grade glioma (HR = 2.2, adjusted p-value    as for ovarian, cervical, and liver cancers. 32,33 Similarly, MCAM has been shown to be highly expressed in vari-

| CON CLUS ION
Our findings demonstrate that MCAM, with multiple biological roles in OS pathogenesis, is a novel prognostic biomarker for OS patients.
MCAM was associated with increased cell migration ability and a greater risk of metastasis and thus could lead to relatively poor prognoses in OS patients.

ACK N OWLED G M ENTS
We thank the investigators of TARGET OS Project and GEO pro-

CO N FLI C T S O F I NTE R E S T
All authors declare that they have no conflict of interest. The data that support the findings of this study are available from the corresponding author upon reasonable request.