Screening of diagnostic markers for osteosarcoma

Introduction

Osteosarcoma is a common clinical malignant bone tumor, representing ~35% of cases and constituting ~0.07% of all the neoplasms (1), and it mainly targets the adolescent age group (2). There has been great progress in the treatment of osteosarcoma, with almost 80% of patients being treated with limb-salvage and the 5-year survival rate for patients with osteosarcoma has increased from 20% to about 80%; however, more than half of all patients succumbed to metastasis and recurrence of osteosarcoma (3). With increased knowledge of tumor molecular biology, the concept of gene therapy for tumors is proposed and the experimental results show potential for clinical application. Several cytogenetic and molecular studies have recently been undertaken in osteosarcoma, and a number of genes (tumor suppressors, oncogenes and genes coding for growth factors) have been identified to be abnormally expressed. These genes may be used as diagnostic cytogenetic or molecular markers for osteosarcoma. For example, the retinoblastoma (RB) gene, the association of which has been well recognized between osteosarcoma and RB. It is reported that patients with hereditary RB have up to 1,000 times the incidence of osteosarcoma compared with the general population (4,5), and sporadic osteosarcoma revealed alterations of the RB gene in ~70% of cases (6). p53, a tumor suppressor gene, is also thought to be significant in the development of osteosarcoma. The p53 gene product can induce the transcription of numerous genes that are involved in the cell cycle control and apoptosis (7), and numerous reports have identified abnormalities of the p53 gene in osteosarcoma in up to 50% of cases. Additionally, the MDM2 gene has also been identified to be overexpressed in osteosarcoma, which may provide an alternative mechanism for the disruption of the normal p53 pathway (6,8). In addition, MDM2 is reported to be amplified in 36% of osteosarcomas and 100% of parosteal osteosarcomas (9).

Several alterations of genes and gene pathways associated with osteosarcoma have been identified in the past few years, which enhance the knowledge of the pathological mechanisms of osteosarcomas. However, there remains no specific diagnostic biomarker and further strides are urgently required to identify abnormalities that have prognostic and therapeutic implications, which may be useful in guiding the diagnosis and therapy for osteosarcomas.

In the present study, the expression profiling data between samples from human osteosarcoma and normal individuals were compared in order to identify significant tissue-specific genes. In addition, their function was investigated with the aim of identifying noticeable diagnostic markers for osteosarcoma.

Materials and methods

Osteosarcoma related gene expression profiles

The gene expression data GSE16088 were obtained from the Gene Expression Omnibus database (10). A total of 20 samples were examined in the present study, among which 14 patients were diagnosed with osteosarcoma while the 6 normal samples served as negative controls.

Preprocessing of expression data

At first, the raw data downloaded were transformed into the expression data format by Affy package in R language (11), and then the missing data was imputed (12). At last, the expression profiling data were standardized using the median standardization method.

Analysis of differential expression

Three statistical tests in the multtest package in R language (13), including the t-test and Wilcox and Fisher’s exact tests were used to conduct the differential-expression analysis between the osteosarcoma and normal groups. The P-value obtained from each test was corrected by the multiple testing Benjamini and Hochberg method (14), and the log FC (fold change) of every gene expression value was also inspected. The genes that could be screened by a variety of test methods simultaneously were regarded as differentially-expressed genes with high degree of confidence. These differentially-expressed genes were used to identify the abnormal genes between the osteosarcoma and normal control samples according to the threshold of P<0.05 and |logFC|>1.

Construction of interaction network

After the differentially-expressed genes were selected with high confidence through rigorous testing methods, Search Tool for the Retrieval of Interacting Genes (STRING) (15) was utilized to search for the interaction objects of differentially-expressed gene products and a protein interaction network was constructed.

Analysis of network module

The module of the whole network obtained previously was decomposed and the function modules with modular nature, including the genes identified through analysis, were found. Gene Otology annotations were performed using Cytoscape (16), Mcode (17) and Bingo (18) software based on the hypergeometric distribution and function enrichment threshold of adjp<0.05.

Results

Results of data preprocessing

There were various factors, including background and probe design that could cause a difference in the original chip data. Therefore, standardization of the data was required prior to the analysis. The difference between data prior to and following standardization was significant (Fig. 1). It was observed that fluctuation of the data following standardization was significantly less compared with that without standardization.

Figure 1 Chip data prior to and following normalization. The blue box represents the normal samples, while yellow represents the disease samples. The black line in the box is approaching on a horizontal line following standardization, which indicates that the standardization degree of data is higher compared with the original data.

Analysis of differential expression

Three statistical testing methods were used to analyze the gene expression data and the data were further corrected by multiple methods. Genes with P<0.05 and with |log FC|>1 obtained in three different manners were selected (Table I). Three differentially-expressed genes met the strict requirements of high degree confidence, and these genes were analyzed further.

Table I List of differentially-expressed genes.

Table I

List of differentially-expressed genes.

ID-REF	Gene symbol	T test-adj	Wilcox-adj	Fisher-adj	logFC
201667-at	GJA1	1.11E-11	0.00054674	0.02724097	1.18443996
202404-s-at	COL1A2	6.05E-13	0.00137961	0.02724097	1.20747741
221729-at	COL5A2	1.23E-12	0.00137961	0.02724097	1.24898200

[i] FC, fold change.

A volcanic plot provides a simultaneous representation of P-values and log2 fold change for the gene expression data. The smaller the P-value, the higher the corresponding fold change (Fig. 2)

Figure 2 Association between the P-value and log FC. Since the smaller the P-value, the greater the logFC will be, the graph presents a volcano-like shape with 0 as the center, and it is termed a volcano plot. FC, fold change.

Results of interaction network construction

The three genes screened as the core were selected and combined with their possible interaction protein predicted from the STRING database. Next, the interaction network was built (Fig. 3).

Figure 3 Interaction network diagram. Protein information from multiple databases was integrated by Search Tool for the Retrieval of Interacting Genes and the interaction objects for the target genes were predicted through the characteristics and spatial structure of the protein sequence itself.

Analysis results of network module

Cytoscape software was used to perform module analysis of the network constructed (Fig. 4). Next, Mcode was applied to identify the common module in the two genes, and the annotation of module function was undertaken by Bingo (Table II).

Figure 4 Sub-module diagram excavated by Cytoscape and Mcode. The yellow circle represents the interacting genes, while the red diamond represents the differentially-expressed genes screened.

Table II List of GO functional annotation.

Table II

List of GO functional annotation.

GO-ID	Function	Corr P-value
7229	Integrin-mediated signaling	1.18E-12
48513	Organ development	1.18E-12
48731	System development	3.69E-10
22610	Biological adhesion	3.69E-10
7155	Cell adhesion	3.69E-10
48856	Anatomical structure development	3.24E-09
9888	Tissue development	5.69E-08
32501	Multicellular organismal process	7.84E-08
7275	Multicellular organismal development	7.84E-08
7166	Cell surface receptor linked signal transduction	1.35E-07
7167	Enzyme-linked receptor protein signaling pathway	1.68E-07
7178	Transmembrane receptor protein serine/threonine kinase signaling pathway	1.94E-07

[i] GO, Gene Otology.

Discussion

In the present study, through comparing the expression data from normal and osteosarcoma tissues, three differentially-expressed genes were identified. Among which, the COL1A2 and COL5A2 genes were selected as the core in order to construct a network and to analyze the interaction between genes correlated with osteosarcoma in order to understand the role of the collagen gene (COL) family and proteins in osteosarcoma.

Collagens are the main fibrous proteins of connective tissue and are the most abundant proteins of the extracellular matrix (ECM). Thus, mutations in different COL genes may disrupt the same collagen fibril. In humans, mutations in 13 different COL genes have been associated with disease phenotypes (19,20). Mutations in the same COL gene can also give rise to different human diseases. For example, mutations in the COL1A2 gene result in at least five different forms of chondrodysplasia and cartilage degeneration (21). There are 19 to 20 types of collagen that have been identified to date, and type I COL, the major component of ECM in skin, bone and ligaments is composed of glycine- and proline rich two-α1 (I) and one-α2 (I) chains (22). In the present study, COL was found to be significantly upregulated and the established network module with COL1A2 and COL5A2 as core revealed that numerous other genes participated in this network and interacted with COL1A2 and COL5A2, such as TGFβ, MMP, SMAD and RUNX2. These genes have all been demonstrated to to be important in the pathogenesis of osteosarcoma. A number of studies have been performed on the correlation between these genes and COL1A2 and COL5A2 and their interaction on the pathogenesis of osteosarcoma. For example, TGFβ2 is a member of the transforming growth factor (TGF)-subfamily, TGFβ induces the synthesis of numerous ECM proteins, such as COL, fibronectin, laminin and tenascin, and inhibits the matrix degrading enzymes. Therefore, TGFβ2 is involved in wound healing, fibrosis, embryogenesis and tumorigenesis (23,24). A previous study provides evidence that TGFβ stimulates human COL1A2 promoter activity through Smad signaling molecules (25). The Smad family contains TGF-β receptor-dependent R-Smads (such as Smad2/3), Co-Smads (such as Smad4) and anti-Smads (such as Smad6/7). It is reported that transient expression of Smad3 or Smad4 in human skin fibroblasts leads to stimulation of COL1A2 promoter activity. In addition, overexpression of anti-Smad and Smad7 represses basal as well as TGFβ-stimulated COL1A2 promoter activity, indicating its antagonistic effect on TGFβ signaling in human dermal fibroblasts (25).

MMP, which is a hallmark of invasive cancers, including osteosarcoma (26), is capable of cleaving type I COL triple helices. Additionally, membrane-bound MMP-14 may be particularly important, as MMP-14 knockout mice exhibit severe COL turnover deficiencies (27).

Collagen fibers, as well as fibers produced by cancer-associated fibroblasts, may form an invasion barrier (28). However, MMPs are essential for maintaining the invasive phenotype and supporting migration and proliferation of cancer cells (29,30). Moreover, the critical function of collagenolytic MMPs in cancer is clearing an invasion path through the barrier of type I collagen fibers in the stroma and blood vessel walls (31,32). Therefore, massive production of MMPs that degrade collagen may solve this problem of invasion barriers. As a consequence, the COL1A2 gene may be used as a diagnostic marker with a higher expression level in cancer cells (33) and restoring type I collagen may be used as a therapeutic target for cancer.

Additionally, the Runt-related transcription factor Runx2, which also interacts with COL1A2, has a well-defined role in mediating the final stages of osteoblast maturation and is required for normal osteogenesis. Runx2 deficiency or mutations affecting the function of Runx2 protein result in severe bone abnormalities in mice and humans. Runx2 is implicated in early osteoblast differentiation, and is involved in the later stages of chondrocyte differentiation, maturation and possibly, endochondral ossification (34). Runx2 expression is essential for the commitment of preosteoblasts to the osteoblast lineage. The cooperative interaction between Runx2 and the RB tumor suppressor protein leads to progressive growth arrest and increases expression of the mature osteoblast phenotype (35), which is reconciled with the molecular pathogenesis of osteosarcoma (36). As the genes mentioned above are all osteosarcoma related and co-interact with COL1A2 and COL5A2, a small mutation in the COL1A2 and COL5A2 genes may affect these other genes, resulting in pathological changes in osteoblasts which lead to osteosarcoma development.

Osteosarcoma, has been a key focus of research as it is a disease that effects adolescents. Jeon et al (37) investigated 25 cases of patients with primary malignant curettage, the local recurrence rate was 18% and the overall survival rate was 65%. The average 5- and 10-year survival rates of these patients were reduced due to local recurrence (38). Bramer et al (39) observed 89 cases of adult osteosarcoma following chemotherapy and observed that their alkaline phosphatase (AP) levels correlated with response to chemotherapy and survival. The results revealed that the prognosis was poor when the AP level was elevated 2-fold compared with the normal level.

In conclusion, in the present study, it was observed that the selected COL gene was significantly upregulated, and this gene was involved in the interleukin conducted signaling pathway. Thus, the upregulation of the COL gene may be considered as a diagnostic marker for osteosarcoma. However, further studies are required to understand the role of collagen in osteosarcoma development in detail.

References