Identication of TIMP2 and RAX as key regulators in chemosensitive osteosarcoma

Background This study is trying to investigate the regulation mechanism of chemosensitive osteosarcoma. mRNA and miRNA proles were explored between chemosensitive and chemoresistant osteosarcoma patients based on GSE39058 dataset. IPA was used for pathway and function enrichment. Another dataset GSE21257 with clinical characteristics was used for survival analysis, and 68 new enrolled osteosarcoma patients with chemotherapy responses were used for RT-PCR validation. Finally, lentiviruses infected U-2 OS cells were subjected to cell counting and real-time cell analyzer. Results A total of 567 differentially expressed genes and 34 differentially expressed miRNAs were screened out. These genes mainly involved in the biology functions of cell survival and cell proliferation, and enriched in the pathways of VEGF signaling, Paxillin signaling, and Inhibition of matrix matalloproteases. In 53 validation osteosarcoma patients from GSE21257, high-expression of TIMP2 and BAX, two common genes among the enriched biology functions, have favorable prognosis with p-values 0.052 and 0.027 respectively. In the 68 new enrolled patients, both TIMP2 and BAX were RT-PCR validated to be high-expressed in chemosensitive group with p-values 0.04 and 0.003 respectively. In addition, cell count and cell index either in TIMP2 or BAX infected U-2 OS cells were smaller compared with control cells within four consecutive days. still

inhibition, enhanced DNA repair, detoxi cation, and osteosarcoma stem cells have been rasied to explain the widely existed chemoresistance, [10,11]. Apoptosis, the key step in the process of chemotherapy inducing cell death, can be inhibited by Bcl-2, p53 and miRNAs [12,13]. Clinical trials showed that high expression of Bcl-2 protein reduced patients' survival rate [12]. Mutation of p53 (R273H) was suggested to have the potential to increase resistance to drug toxicity in human Saos-2 cell line [14,15], and a type of p53 deletion resulted in increasing chemotherapy sensitive compared with p53 wild type [16]. In addition, multiple miRNAs such as miR-140 and miR-215 were involved in the chemoresistance by inhibiting cell proliferation through G1 or G2 phases [17,18]. Besides, miR-33a affected chemotherapy by regulating the expression of TWIST [19].
Chemotherapy resistance has signi cantly reduced the survival of osteosarcoma patients. In this study, we are trying to reveal the molecular mechanism of chemosensitive osteosarcoma and indentify potential targets to reverse chemoresistant osteosarcoma.

Pathway and Biology Function Enrichment
To explore the biology functions of the DEGs, canonical pathway and biology function enrichment were carried out using IPA. The DEGs can signi cantly enriched into 7 pathways with absoluate z-score > 1 ( Fig. 1A), in which 4 pathways were predicted to be activated and 3 pathways were predicted to be inhibited. Several of these enriched pathways had ability to regulate cell survival and cell proliferation like VEGF signaling (p-value = 6.4E-6), Paxillin signaling (p-value = 5.7E-3) and Inhibition of matrix metalloproteases pathways (p-value = 1.3E-2) (Fig. 1B).
Downstream disease and function analysis showed that 159 genes of the 567 DEGs mainly involved in the biology functions of cell survival, cell viability and invasion of broblast cell lines with z-score less than − 2 (Table 1). Besides, 218 genes can activate the biology functions of mortality and organismal death with z-score more than 2 (Table 1). In addition, 10 common genes among these enriched biology functions can signi cantly inhibite cell survival based on IPA knowledge base ( Fig. 2A). Within the 10 genes, BAX has been documented to be involved in chemotherapy resistance of metastatic cancer cell line [20], and MAPK8 and TIMP2 can regulate osteosarcoma via PXN, CRK, BCAR1 and PHKA2 based on IPA knowledge base (Fig. 2B). The miRNA-mRNA pairings were constructed using microRNA target lter module in IPA. Among the 10 common genes, TIMP2, DCT and CBR1 can be regulated by ve differentially expressed miRNAs like miR-3612, miR-4300, miR-34c-3p, miR-454-5p and miR-548b-3p (Fig. 3). In addition, 40 DEGs were experimently validated or predict to be regulated by 20 differentially expressed miRNAs.

Survival Analysis and RT-PCR Validation
Acoording to the expression of TIMP2 and BAX, the samples from GSE21257 were catagorized into upregulation and down-regulation groups. Kaplan-Meier survival analysis based on 5-year overall survival showed that TIMP2 down-regulation led to a worse overall survival (log-rank test p-value = 0.052, Fig. 4A). And BAX up-regulation was correlated with a better overall survival (log-rank test p-value = 0.027, Fig. 4A).
Further, the expression level of TIMP2 and BAX were validated in 68 chemotherapy grouped osteosarcoma samples by RT-PCR. Results showed that both TIMP2 and BAX were high-expressed in chemosensitive group with p-values 0.04 and 0.003 (Fig. 4B). The results were concordant with microarray analysis.

Cell counting and growth
To measure the in uence of TIMP2 and BAX on U-2 OS cells growth, we built a model of TIMP2 or BAX stable over-expression U-2 OS cells. As shown in Fig. 5, the cells with TIMP2 or BAX over-expression inhibit cell growth compared with control cells within four consecutive days. RTCA also con rmed that the cell index was slowed down in TIMP2 and BAX over-expression cells.

Discussion
In this study, the transcriptomic pro les of chemosensitive and chemoresistant osteosarcoma were explored. A total of 567 DEGs and 34 differentially expressed miRNAs were identi ed. Downstream pathway and biology function enrichment showed that the DEGs enriched pathways like VEGF signaling, Paxillin signaling, and Inhibition of matrix matalloproteases mainly involved in the biology functions of cell survival, cell proliferation, cell movement and development. VEGF as one of the angiogenic factors can promote tumor vasculature development by providing necessary oxygen and nutrients [21]. Different kinds of regulators like PARK2 [22], P2 × 7 [23] and α-CaMKII [24] can regulate the growth and metastasis of osteosarcoma via VEGF signaling pathway as well. In this study, VEGF signaling pathway was activated in chemosensitive osteosarcoma probably due to the effect of the therapy on vascular growth. But we still need further investigation to uncover the underlying mechanism. The inhibition of Paxillin signaling pathway and activation of Inhibition of matrix metalloproteases pathway can inhibite tumor cell migration, cell survial and proliferation. The fundamental ability of cancer cell is to penetrate surrounding extracellular matrices for invasion or metastasis, which is facilitated by paxillin transducing signals from integrin receptors to the actin cytoskeleton [25]. It was previously demonstrated that overexpression of paxillin can contribute to the high risk of osteosarcoma metastasistic [26].
Metalloproteinases (MMP) play a key role in the reponse of cells to microenvironment which in turn in uence cell migration and survival. Several metalloproteinases like MMP16 and MMP9 have been reported to be regulated by miR-328-3p and Wogonin respecitively, which can suppress osteosarcoma malignant progression [27,28].
Disease and function enrichment also showed that the DEGs can inhibite the cell viability, invasion of broblast cell lines, cell survival, and activate mortality and death processes. Two of the common genes among these enriched biology functions, TIMP2 and RAX, were related to the clinical characteristic of overall survival. The following RT-PCR, infection U-2 OS cell couting and RTCA validation also con rmed that both TIMP2 and RAX can inhibit osteosarcoma malignant progression. These evidences suggested thatthe two genes could be important regulators in chemosensitive osteosarcoma patients. TIMP2 as one of the natural inhibitor of the matrix metalloproteinases can also activate the pathway of inhibition of matrix metalloproteases. In osteosarcoma and lung cancer, TIMP2 has been demonstrated to be regulated by different kinds of miRNAs such as miR-552, miR-93 and miR-550a-3p to promote migration, proliferation and tumorigenesis. Our study found that the down-regulated miR-3612 and miR-4300 can also target TIMP2. BAX as a pro-apoptotic regulator is involved in a wide variety of cellular activities and has been shown to be involved in p53-medicated apoptosis [29]. The high levelexpression of BAX in chemosensitive osteosarcoma have the ability to inhibit tumor progression. Previous study have demonstrated that down-expression of BCL-2 and up-expression of BAX sensitized osteosarcoma to doxorubicin [30]. Furthermore, down-expression of BAX in breast adenocarcinoma was associated with poor response to combination chemotherapy [31]. In addition, studies have shown that activation of BAX increased osteosarcoma cell sensitivity to apoptosis [32], and osteosarcoma patients with BAX inhibition had unfavorable prognosis [33].

Conclusion
In summary, the biology functions of cell survival and proliferation were identi ed to be signi cantly inhibited in chemosensitive osteosarcoma patients. Two up-regulated genes, TIMP2 and BAX, were validated to be important regulators of cell proliferation and apoptosis in chemosensitive groups. Further experimental evidences are still needed to consolidate this results.

Data collection
Lots of publicly available gene expression datasets have been submitted to GEO database (Gene Expression Omnibus Database). Here we carefully searched GEO database with the key words of "osteosarcoma AND mRNA AND microRNA". Six datasets were screened out. Among the 6 datasets, three datasets (GSE89370, GSE89074 and GSE86109) only contain mRNA expression pro les no miRNA data, Analysis of Differentially Expressed Genes mRNA and miRNA expression pro les were divided into chemoresistant group and chemosensitive group. Then the expression pro les were processed using internal R scripts and available annotation databases. Brie y, background correction, quantile normalization and log 2 transformation were rstly applied using GeneChip Robust Multi-array Analysis algorithm (GC-RMA) [35]. Then, useless probes were ltered out and mean expression values were calculated for each gene. Finally, differentially expressed genes (DEGs) and miRNAs were obtained by Limma package (Linear Models for Microarray Analysis) within Bioconductor [36]. Adjust p-value and absolute log2 fold change were set to less than or equal to 0.05 and greater or equal to 2 respectively.

Function Analysis
The differentially expressed genes and miRNAs along with fold change and p-value were uploaded to Ingenuity Pathway Analysis (IPA) (QIAGEN, Redwood City, CA, USA). Then core analysis was performed with parameters setting as direct and indirect relationships included, causal networks included, selected all node types, and include experimentally observed results. The rest parameters were set as default. The core analysis consists of canonical pathway enrichment, upstream regulator analysis, downstream functions and disease, regulator effect analysis, network analysis and so on. In the results, IPA will calculate a statistical z-score indicating effect direction, and usually z-score large than 0 means activation.

miRNA and mRNA Pairings
We have built mRNA-miRNA pairings by microRNA Target Filter module in IPA to construct the regulation network. The analysis is based on experimentally validated mRNA and miRNA interaction from TarBase [37], miRecords [38] and peer-reviewed literatures as well as predicted mRNA-miRNA interaction from TargetScan [39].

Survival analysis
To evaluate the prognostic value of identi ed genes, we downloaded the dataset of GSE21257 with 53 samples including documended clinical characteristics. The dataset was divided into two groups according to the quantile expression of the identi ed genes [40]. Survival analysis was carried out using the Kaplan-Meier method with the log-rank test to compare overall survival and gene expression groups.

Osteosarcoma tissue analysis
Comparing the pain degree, tumor size, tumor boundaries, and calci cation before and after the chemotherapy was used to evaluate chemotherapy e cacy, which was classi ed into three categories: RT-PCR experiments were preformed on the 68 samples to validate the mRNA expression of the identi ed differentiately expressed genes (TIMP2 and BAX). Total RNA was extracted using TRIzol reagent with manufacturer's instruction (ThermoFisher Inc., USA). Then cDNA was synthesized using M-MLV Reverse Transcriptase from Promega (Promega, Madison, WI, USA), and mRNA expression was detected using 7500 Real-Time PCR system (ThermoFisher, USA). The internal GAPDH mRNA expression was used for normalization, and relative quanti cation was calculated using 2 −ΔCt method [42].

Cell lines and Lentiviruses infection
The human osteosarcoma cell line U-2 OS cells were purchased from Shanghai YiYan Biotech Inc. (Shanghai, China), and cultured in HyClone-Dulbecco's modi ed eagle medium (DMEM) which contained 10% fetal bovine serum at 37℃ in a humidi ed incubator with 5% CO 2 . To build TIMP2 and BAX stable over-expression cells, we designed and purchased Lenti-TIMP2-EGFP and Lenti-BAX-EGFP from GeneChem Company (Shanghai, China). Based on the standard procedure, the U-2 OS cells were seeded in six-well plates supplied with DMEM and 10% fetal bovine serum. The cells were infected with Lenti-TIMP2-EGFP or Lenti-BAX-EGFP and Lenti-vector control, and stable cell clones were selected in the presence of puromycin for 1 to 3 weeks. The gene and protein expressions of TIMP2 and BAX in the cell line were evaluated using real-time PCR, agarose gel electrophoresis and Western blot by GeneChem Company (Shanghai, China).
Cell counting and growth curve assay To evaluate the colony formation, the TIMP2 and BAX infected U-2 OS cells were rstly cultured to the logarithmic growth stage. Then cells were planted into four plates at a density of 50/100/150/200cells per plate and culture for 2 weeks. Finally, the cells with crystal violet stain were counted using microscope and quanti ed using ImageJ (NIH, USA).
The cells growth curve was evaluated by using real-time cell analyzer (RTCA) xCELLigence system (ACEA Bioscience, San Diego, CA). According to the manufacturer's instructions, the cells were transferred to RTCA E-plates with a density of 1 × 10 4 cells per well and incubated for 1 hour. Then the baseline was recorded, and the electrical impedance was measured for four days in each well. The cell index was calculated based on the electrical impedance which can re ect cell viability. The DEGs enriched canonical pathways with absolute z-score > 1 (A) and the top one enriched VEGF signaling pathway (B).  Three out of the ten common genes can be regulated by 5 identi ed miRNAs Page 15/16

Figure 4
Overall survival analysis of TIMP2 and BAX using 53 osteosarcoma samples from GSE21257 (A). RT-PCR validation of TIMP2 and BAX expression in 68 collected chemotherapy grouped osteosarcoma patients (B).