Identification of key genes and multiple molecular pathways of metastatic process in prostate cancer

Introduction

Prostate cancer (PCa) is the most frequently diagnosed cancer and the fifth leading cause of cancer death in men, with 1.3 million new cases and 359,000 associated deaths worldwide in 2018 (Bray et al., 2018). The 5-year survival rate for metastaic PCa rapidly decreases to 30% despite the rate for localized PCa is almost 100% (Litwin & Tan, 2017). Therefore, cancer metastasis is the last thing that patients and doctors want to see. Usually, localized PCa spread to lymph nodes at first and then to the bones, liver or lungs (Wang et al., 2018). Mechanisms on how PCa develop and metastasize have been extensively investigated, which mainly include the process of PCa metastasis (Weidle et al., 2019), the biomolecules such as miRNAs and enzymes involved in metastasis (Pan et al., 2019; Wu et al., 2018) or aberrant glucose metabolism of metastaic PCa (Pichla et al., 2019). However, the mechanisms still remains incomplete for limitation by the failure to holistically reproduce each individual element of the metastatic cascade in PCa metastasis (Berish et al., 2018). Moreover, there is lack of easily accessible biomarkers which can accurately predict the metastatic potential of localized PCa. Therefore, it is necessary for us to figure out the differences about the biology between localized and metastatic PCa for developing new prognostic markers and therapeutic targets.

Microarray technology and bioinformatic analysis have been also widely applied, in order to explain the differences of gene expression which can help us better understand mechanisms about prostate tumorigenesis and progression (Huang et al., 2017). However, previous studies mainly focused on analysis of localized PCa and normal tissue samples. In the present study, we downloaded two gene expression profiles (GSE32269 and GSE6919) from the Gene Expression Omnibus database, which contain localized and metastaic PCa tissue samples. Then gene expression profiles were analyzed, and the differentially expressed genes (DEGs) were identified between localized PCa and metastaic PCa. Subsequently, gene ontology (GO) analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis and protein–protein interaction (PPI) analysis of DEGs were performed. Finally, a total of 90 DEGs and 10 hub genes were identified, which may be potential prognostic markers and therapeutic targets for PCa.

Materials and Methods

Results

Microarray data normalization and identification of integrated DEGs

The PCa chip expression datasets GSE32269 and GSE6919 were normalized by R language, then the DEGs were screened using the limma R package (adjusted P < 0.05 and |log₂fold change (FC)| > 1). The GSE32269 dataset contained 869 differential genes, including 706 upregulated genes and 163 downregulated genes. The GSE6919 dataset contained 427 differential genes, including 319 upregulated genes and 108 downregulated genes. Then an overlap of 90 DEGs was identified from the two profile data sets, which was displayed by Venn diagram (Fig. 1A), including 47 upregulated genes and 43 downregulated genes in the metastaic PCa compared to localized PCa (Table S1).

Hub gene selection and analysis

As the criteria described above, a total of 10 genes were identified as hub genes in the PPI network. Nine of these hub genes were upregulated including UBE2C, Cyclin B2 (CCNB2), CKS2, HMMR, androgen receptor (AR), CDKN3, TPX2, AURKA, SPP1 and only one gene, FOS, was downregulated. A network of the hub genes and their co-expression genes was constructed using cBioPortal online platform (Fig. 2A) and the BP analysis of the hub genes is shown in Fig. 2B.

Validation and prognostic significance analysis of the hub genes

UALCAN was used to analyze the hub gene transcript expression in the 476 samples which are derived from the TCGA project. The samples included 52 normal samples, 345 localized PCa samples and 79 metastaic PCa samples, so expression values in normal samples, localized and metastatic samples were compared. As shown in Fig. 3, the expression of the majority of hub genes in metastaic tumor samples were significantly elevated when compared with localized tumor samples or normal samples. These findings suggested our results of the identified candidate hub genes are reliable.

To further investigate whether these hub genes contributed to the prognostic in patients, GEPIA, an online tool with data sourced from TCGA and GTEx, was used to analyze the disease-free survival of these hub genes in PCa. As shown in Fig. 4, high expression of UBE2C, CKS2, CCNB2, HMMR, CDKN3, TPX2 and AURKA showed worse disease-free survival in PCa patients, while the disease-free survival of the other three genes were not significant, which revealed that UBE2C, CCNB2, CKS2, HMMR, CDKN3, TPX2 and AURKA were associated with prostate tumor progression and might be used as tumor progression predictors for prostate tumor patients.

Discussion

Over the last decades, many researches including microarray bioinformatics were focused on the molecular mechanisms of PCa generation, development and metastasis, however, the biological programs underlying this complex process still remain unclearly. Different from former studies, we firstly adopted an integrated bioinformatics approach to directly compare the the differences in gene expression between localized PCa and metastaic PCa. An overlap of 90 DEGs was identified, including 43 downregulated genes and 47 upregulated genes. Then, GO enrichment and KEGG pathway analyses were performed to show us how alterations in DEGs expression affect the biological pathways involved in the development of PCa. The BP of upregulated genes were mainly enriched in regulation of catabolic process, regulation of cell cycle and regulation of protein metabolic process, while the downregulated genes were mainly enriched in response to steroid hormone and response to glucocor ticoid. Previous studies have reported that cancer cells possess different metabolic pathways when compared with healthy cells, in order to meet a fast growing need (Sikander et al., 2019). In addition, recent studies showed that metastasizing cancer cells require metabolic changes depend on the environment in which they reside (Fendt, 2019). Moreover, steroid hormones have far-ranging biological impacts involved in PCa initiation, development, metastasis and treatment (Agbo & Lambert, 2019; Snaterse et al., 2017). So, all these studies support our results. The most significant module of PPI network were mainly enriched in mitotic cell cycle, cell cycle and cell division, which also involved in tumorigenesis or progression (Sawada et al., 2019).

Hub genes, namely UBE2C, CDKN3, TPX2, CCNB2, AURKA, CKS2, HMMR, AR, SPP1 and FOS were identified in analyzing the PPI network of DEGs, indicating these genes may be vital in metastaic PCa. UBE2C belongs to ubiquitin-conjugating enzyme family for the destruction of mitotic cyclins and cell cycle progression. Dastsooz et al. (2019) reported that UBE2C is over expressed in all 27 cancers including PCa and significantly higher in late-stage tumors, and also participates in tumor progression or invasion by regulating the cell cycle, apoptosis and metastasis. UBE2C has also been found over expressed in androgen-independent PCa (Wang et al., 2009) and castration-resistant PCa (Chen et al., 2011). Similarly, CDKN3 (also called CDI1 or KAP) is a member of the dual specificity protein phosphatase family, and plays a critical role in PCa via regulating cell cycle and DNA replication signaling (Yu et al., 2017). CCNB2 is a member of the B-type cyclin family, which contains CCNB1 and CCNB2, serves a key role in progression of G2/M transition. Previous studies have shown that CCNB2 is highly expressed in a variety of tumors, such as colorectal adenocarcinoma, breast cancer and even serves as a poor prognostic biomarker in non-small-cell lung cancer (Park et al., 2007; Shubbar et al., 2013; Qian et al., 2015).

The androgen receptor gene is located on the X chromosome at location Xq11-12. It is well known that AR represents a vital driving force in the development and progression of prostate (Beretta & Zaffaroni, 2019) and the signaling pathway of AR plays important roles in prostate carcinogenesis and metastatic or androgen-independent progression of the disease (Tien & Sadar, 2019). In addition, Mills (2014) reported that AR involves in PCa development by affecting genomic stability and DNA repair. However, we found that the expression of AR in metastaic PCa was not significantly elevated compared with localized tumor samples (Fig. 3). In fact, there is no significant expression difference between and among metastaic PCa, localized PCa and normal tissue samples. Moreover, we assessed the expression of AR in relation to disease-free survival of PCa, and there is no significant difference too. The same results were found when we used another online cancer database tool (the data didn’t show). A possible reason for this is that AR expression can be heterogeneous in PCa patients (Koochekpour, 2010). It was proposed that AR has several genetic alteration forms involved in metastatic or progression of PCa including genomic amplification of AR, hypersensitive AR resulting from point mutations, promiscuous mutant AR protein activated by nonandrogenic ligands and AR-polymorphisms changing the response to androgen (Heinlein & Chang, 2004). Structural and functional AR abnormalities might lead to an alteration of the AR activity, which result some patients have high AR activity while the expression is normal, and further research is needed to confirm the exact molecular mechanisms.

Another interesting finding of hub genes was the expressions of FOS and SPP1. There is no significant difference between localized PCa and normal tissue samples, but when compared between the metastaic PCa and normal samples, the expression of FOS was decreased significantly and SPP1 was increase significantly, which indicating these genes might be potential predictors for the metastatic ability of PCa. SPP1, also known as osteopontin, an ECM proteinis involved in tumorigenesis and cancer progression and has been shown to be a key role of PCa metastasize to the bone (Desai, Rogers & Chellaiah, 2007). Pang et al. (2019) reported that SPP1 is significantly associated with survival in several forms of cancer, and plasma SPP1 levels may serve as markers for stage, grade and early tumor progression in PCa. FOS is one of the first studied proto-oncogenes and has been implicated as regulators of cell growth, differentiation and apoptosis (Piechaczyk & Blanchard, 1994). In the present study, FOS was observed to be downregulated, indicating its underlying tumor metastasis suppressor effect, which was consistent with Zhang et al. (2007) finding that FOS has a proapoptotic function by repressing the antiapoptotic molecule to induce apoptosis in PCa cells.

Differing from a previous study (Liu et al., 2018), our study included a larger sample size and presented a more comprehensive and systematic study. But there are still some certain limitations to the present study. PCa is extremely heterogeneous and multiple molecular subtypes of primary and metastatic PCa have been characterized (Harris et al., 2009; Cancer Genome Atlas Research Network, 2015). However, the subclass of PCa datasets in this study is not known, which may lead a number of important genes ignored and restrict the ability to draw a valid result. The absence of experiments verifying the expression of hub genes was also a limitation in the present study.

Conclusions

In conclusion, our study identified some significant genes and biological pathways using an integrated bioinformatics analysis, which could improve our understanding of the metastaic process of PCa. These genes might be potential predictors and therapeutic targets for the metastaic PCa. However, further studies are required to validate these findings.

Supplemental Information