Prostate cancer is a urogenital cancer that accounts for 15% of all cancers in men globally (Group, UCSW, 2014; Van Gool and Pearson, 2014). PC is initially asymptomatic (Ratner, 2021), but painful urination, blood in the urine, or pelvic discomfort symptoms commonly arise as the disease progresses over the months and years (Screening and Board, 2002). The known risk factors for PC include increasing age, African-American ethnicity, fat rich diet, positive family history, and obesity (Bostwick et al., 2004). Early clinical testing and an accurate diagnosis of PC are critical in determining treatment options (Greenlee et al., 2001). The most common biomarker used to detect PC is prostate-specific antigen (PSA) test in the blood. However, it fails to distinguish indolent or aggressive cancer stages (Cancer, IAFRO, 2003). Besides this, increased age or an enlarged or inflamed prostate are the other factors that can also elevate serum PSA leading to misdiagnosis or unnecessary and costly treatment. This warrants the need to identify robust biomarkers that are not only specific and sensitive but also improve the overall accurate diagnosis and prognosis of prostate cancer.

Prostate cancer is estimated to have 58% heritability, highest among all major cancers (Lachance et al., 2018). Family-based linkage studies have identified a series of genes responsible for hereditary PC such as HPC1 (Carter et al., 1992; Cooney et al., 1997; Berry et al., 2000a), PCAP (Neuhausen et al., 1999; Berry et al., 2000a; Xu et al., 2001), HPCX (Schleutker et al., 2000), CAPB (Berry et al., 2000a; Xu et al., 2001), HPC20 (Berry et al., 2000b), HOXB13 (Breyer et al., 2012; Xu et al., 2013) etc. Over the last decades, Genome wide associated studies (GWAS) have identified about 160 common risk loci in PC, suggesting a polygenic model of PC (Farashi et al., 2019). Most of these risk loci genes are involved in the cell cycle or DNA repair (ATM, TERT, MYC, and MDM2), inflammatory response (IL8RB), and metabolism (JAZF1 and HNFB). However, under-representation of ethnic diversity necessitates us to look for universally applicable genetic susceptibility biomarkers for PC. Moreover, the molecular basis of PC could not be fully explained by candidate genetic variants alone, but through the global gene expression alterations. Some studies have confirmed the potential contribution of dysregulated gene expression changes in either blood (Wang et al., 1979; Papsidero et al., 1980; Kuriyama et al., 1981) or prostate biopsies (Ragde et al., 1988), and few were correlated with the survival rate of PC patients (Greenlee et al., 2001). These genes were mainly involved in diverse cellular functions like protein kinase binding, enzyme binding, cell activation and proliferation, Wnt regulation mechanisms, wound healing, apoptosis, etc. (Verras and Sun, 2006; Kypta and Waxman, 2012). However, the specific molecular regulators which control the gene expression changes have not been well characterized.

Gene expression is controlled at both post transcription and translation levels. MicroRNAs (miRNAs) are 18–22 bp long, and known to regulate more than 50% of protein-coding genes at post transcriptional level (Bartel, 2009). Few miRNAs have been identified as potential diagnostic biomarkers of tumour development and metastasis (Nikitina et al., 2012; Ilic et al., 2013). PC associated microRNAs in serum have the potential to successfully distinguish cancer and healthy controls. The aggressive PC is known to show over-expression of oncogenic miRNAs; miR-21, miR-125b, miR-221, and miR-222 (Sun et al., 2009). The miR-21 is over-expressed in PC and contribute to tumour growth (Nikitina et al., 2012) by knocking down PTEN and other tumour suppressor genes (Meng et al., 2007). The miR-200 family is regarded as key regulatory molecule in PC due to their down regulated expression tumor suppressive function and regulatory role in initiation and migration of prostate tumors (Feng et al., 2014). Furthermore, these 5 miRNAs when combined with routine PSA test were shown to improve the PC diagnosis (Siegel et al., 2020). However, influence of miRNAs on differentially expressed genes in prostate tissues is not explored in detail till now.

Owing to the sparse data, this study aimed to expand our current understanding of PC pathogenesis by exploring miRNA-mRNA interactions in prostate gland tissues. By involving a series of comprehensive bioinformatics approaches, this study has identified several gene-network clusters involved in cell communication, inflammation, proliferation, and differentiation processes which, are dysregulated in prostate cancers. Our findings provide a novel insight into understanding the mechanisms of PC, besides uncovering genetic markers with potential for disease diagnosis and therapeutic modulation.

The workflow of current work is presented in Figure 1.

Prostate cancer has a high clinical heterogeneity, ranging from fairly indolent to a fatally aggressive form (Litwin and Tan, 2017). This necessitates the need to search for novel prostate tumour biomarkers that can improve the molecular diagnosis and clinical outcomes. Several investigations have identified unique molecular subtypes of primary prostate tumors as well as important genetic changes that contribute to metastasis (Chen et al., 2021) (Zhao et al., 2021) (De Schaetzen Van Brienen et al., 2021). But these studies differ from each other in sample type, study design, statistical measures, and systems biology methods used. Moreover, the data on the role of miRNAs in regulating gene expression changes in prostate cancer tissues is sparce (Doghish et al., 2021) (Gordanpour et al., 2012) (Watahiki et al., 2011). Therefore, this new study design combined the publicly available expression datasets from different prostate tissues and across different microarray platforms to precisely map the interactions of key genes and miRNA involved in prostate tumorigenesis.

By normalization of expression data and comparison to normal prostate tissues, we detected 250 differentially expressed genes (125 up- and 125 downregulated genes) in the prostate cancer tissues. Pathway enrichment of these key upregulated genes highlight their involvement in Hedgehog signalling and ErbB signalling pathways. The Hedgehog (Hh) signalling network plays an important role in metastatic human prostate tumours that expressed Hh ligands and Hh-target genes at higher levels (Karhadkar et al., 2004). Moreover, chemical inhibitors of Hh signalling have been demonstrated to suppress the growth of PC cell lines (Sanchez et al., 2005). ErbB-2 oncogene contributes to PC progression by regulating numerous signalling pathways, including AKT, ERK1/2 and STATs, by positively influencing cell survival, migration, and proliferation of cancer cells (Miller et al., 2019).

The downregulated genes were enriched in key prostate cancer pathway (KEGG:05,215), which regulates growth factor receptor mediated cytokine-cytokine interactions (Lee and Rhee, 2017). These growth factor receptors indirectly activate the RAS-MAPK signalling that regulates gene expression, cellular growth, and survival of prostate cancer cells (Rodríguez-Berriguete et al., 2012). Increased cell proliferation and resistance to apoptosis may be caused by abnormal MAPK signaling pathway (Dasgupta et al., 2020). cAMP serves as a intermediary messenger molecule in several signaling pathways such as cell growth and differentiation (Kim et al., 2005). In the context of PC, cAMP levels modulate the activity of androgen receptor (AR) signalling, which could negatively impact normal prostate development and function (Yan et al., 2016).

We also identified miRNA alterations targeting genes which are regulating the transition of normal prostate to aggressive cancerous cells. The hub gene is a gene with the highest degree of connectivity (at top 10%) in the key module and identifying the hub genes is shown to improve the molecular dissection of human diseases like obesity (Sabir et al., 2019), autoimmune diseases (Banaganapalli et al., 2020), myocardial infarctions (Mujalli et al., 2020), and cancers (Sahly et al., 2021). So, by constructing the miRNA-mRNA functional network of 250 target genes and 53 miRNAs, we mapped 32 hub target genes and six hub miRNAs showing high degree of network centrality parameter (>150).

The inverse correlation analysis of expression statuses of these target gene-miRNA pairs has narrowed down the total hub genes to 9 (NFIB, DICER1, GSK3B, DCAF7, FGFR1OP, ABHD2, NACC2, NR3C1, and FGF2) and hub miRNAs to 6 (hsa-miR-455-3p, hsa-miR-548c-3p, hsa-miR-582-5p, hsa-miR-671-5p, hsa-miR-125a-3p and hsa-miR-338-3p). Among the hub miRNAs exclusively deregulated in PC, only 3 miRNAs (hsa-miR-455-3p, hsa-miR-548c-3p and hsa-miR-582-5p) were found to target 9 hub genes. The pathway analysis of these hub genes and miRNAs showed their involvement in different pathways connected to cell division or differentiation in response to extracellular signals, and a variety of cellular processes including gene expression, regulation of growth proteins, cell proliferation, oncogenic transformation, cell migration, and membrane trafficking in cancers (Kim et al., 2005). Thus, our findings broadly agree with previous studies, which highlighted differential regulation of key genes and miRNAs involved in the transformation of normal prostate tissue to prostate cancers (Gordanpour et al., 2012) (Chen et al., 2021) (Zhao et al., 2021) (De Schaetzen Van Brienen et al., 2021).

Both miRNAs and TFs regulate the gene expression at post-transcriptional and post-translation levels, respectively. Interestingly, miRNAs and TFs can regulate each other and also co-regulate a common target gene by forming a feed-forward loop (FFL) unit, which further forms gene regulatory networks (Qin et al., 2020). In this study, three miRNAs were predicted to co-regulate the network of TF-target gene network. For example, hsa-mir-548c-3p is shown to coregulate the interaction of NACC2 and NR3C1 with TF-ZNF524; hsa-mir-455-3p co-regulates the interaction of NFIB with TF-ZNF524, and hsa-mir-582-5p coregulate the interaction of ABHD2 and DICER1 with TF-ZNF524 and TF-CRX.

The comprehensive systems biology validation of 9 hub genes has revealed that downregulated miRNA like hsa-miR-582-5p is correlated with the elevated the expression of ABHD2 and GSK3B genes in prostate cancer tissues, and upregulated miRNA like hsa-miR-548c-3p correlates with the lower expression of NR3C1 (glucocorticoid receptor) gene. The hsa-miR-582-5p inhibits metastasis of prostate cancer by repressing multiple components of TGF-β signalling, resulting in the inactivation of TGF-β signalling (Gordanpour et al., 2012) (Huang et al., 2019). Of note, miR-548c-3p acts as regulator in many cancers by significantly affecting both ErbB and Hippo signalling pathways (Feng et al., 2019). Hence, both miRNAs can be potential biomarkers for metastatic prostate cancers. Open target analysis findings have predicted that, Proteolysis Targeting Chimeras (PROTACs) can target all ABHD2, GSK3B, and NR3C1 genes. Additionally, ABHD2 is targeted by antibody molecules, and GSK3B is targeted by lithium carbonate inhibitor which has currently completed phase 1 trial in prostate cancer. NR3C1 agonists like prednisone and dexamethasone are already under usage in lymphoid cancers and can be explored as potential repurposed drug for prostate cancers.

The GSK3B, a central component of PI3K/Akt survival pathway, determines the cancer cell survival through its effects on apoptosis. Expression status of GSK3B is positively correlated with tumor progression in multiple human malignancies (Li et al., 2015). Immunohistochemistry analysis of 499 PC surgical specimens showed higher levels of cytoplasmic (not nuclear) of GSK3B protein, in addition to its correlation with aggressive clinicopathological parameters such as late clinical stage, lymph node metastasis, extracapsular extension, and a high Gleason score, as well as a 2-fold reduction in recurrence-free survival with up to 12-years follow-up (Li et al., 2009). Since it is an androgen-regulated gene that is associated with the development of prostate cancer and its ability to resist chemotherapy, ABHD2 is a promising new target for prostate cancer screening and treatment. Gleason score, pathological node stage, low cancer-specific survival rates, and resistance to docetaxel-based chemotherapy have been linked to ABHD2 levels in prostate cancer specimens by immunohistochemical analysis. (Obinata et al., 2016). The NR3C1encodes a nuclear hormone receptor, acting as a transcription factor and modulate the gene expression across different tissues (Gu et al., 2017). Lower NR3C1 expression levels were found in cancer cells compared to normal tissues of breast (Mamoor, 2021). For lymphoid cancers, acute lymphoblastic leukemia, chronic lymphocytic leukemia and multiple myeloma, NR3C1 activation has been proven to be an effective treatment strategy. There are, however, several lines of evidence suggesting that GR activation has a strong effect on cancer cell behaviours such as invasion, apoptosis resistance and growth (Prekovic et al., 2021). Thus, it is imperative to use therapeutic interventions, whether agonists or antagonists to control the prostate carcinogenesis by regulating the expression of concerned candidate genes.

This study would like to sincerely clarify some limitations. First, we studied the role of miRNA and target expressions in metastatic prostate cancers. Thus, we do not generalise our findings neither blood samples nor primary tumors of prostate cancer patients. Second, we took microarray data of prostate tissue biopsies. Thus, mapping the key miRNAs or genes in body fluids like blood, urine and prostatic fluid is important to develop non-invasive biomarkers for prostate carcinogenesis. Third, the key hub genes and hub miRNAs identified in this study, need to be further validated by targeted approaches like real-time PCR, functional biology assays and on larger number of patient samples.

In conclusion, this study identified the inversely correlated dysregulation of 9 target genes (NFIB, DICER1, GSK3B, DCAF7, FGFR1OP, ABHD2, NACC2, NR3C1, and FGF2) and three key miRNAs (hsa-miR-455-3p, hsa-miR-548c-3p and hsa-miR-582-5p). This study supports the robustness of computational concepts, like protein interactome mapping, functional enrichment of biological pathways and construction of miRNA-mRNA and transcription factor gene networks in identifying prostate cancer biomarkers from large-scale gene expression data. Tissue based prostate cancer markers can be correlated with cancer development, which might offer early clinical interventions and therapy. This study also provides the initial evidence for the future knowledge driven functional studies of these miRNAs and their gene targets in prostate cancer.