In our study, for each of the five subtypes of glioma, specific gene and microRNA expression profiles were described (Table 2). As a result, patterns of specifically expressed markers were determined: 211 unique DEGs and 21 DE miRNAs were identified in astrocytoma G2, 63 and 65 were identified in astrocytoma G3, 119 and 10 were identified in oligodendroglioma G2 and 147 and 51 were identified in oligodendroglioma G3.
The identified expression patterns were generally related to signalling pathways associated with neuronal activity, proliferation and cancer progression. In glial tumour samples at earlier stages (G2 and G3 samples), a similarity in the set of signalling pathways with aberrant activity was demonstrated (Table 3). At the same time, signalling pathways that were not shared by the astrocytomas or oligodendrogliomas were identified for the GBM group.
Based on the initial screening data, a panel of genes was developed, the differential expression of which not only was unique for each subtype but also correlated with the overall survival of patients (Online Resource 20). The panel includes 26 DEGs unique to astrocytoma G2, 15 DEGs unique to astrocytoma G3, 17 DEGs unique to oligodendroglioma G2, 7 DEGs unique to oligodendroglioma G3 and 151 DEGs unique to GBM.
Currently, the role of individual miRNAs in glial tumours is not well understood. MiRNAs are small non-coding regulatory RNAs that decrease the stability and/or inhibit the translation of target mRNAs with which they have full or partial complementarity. These regulatory molecules, playing an important role in the pathogenesis, development and progression of cancer and the response to therapy, are especially promising for diagnostic strategies, including minimally invasive methods (measuring the level in plasma, serum or cerebrospinal fluid).
Primary data analysis via TCGA allowed us to differentiate diffuse astrocytomas and oligodendrogliomas from other subtypes of gliomas by the expression of 21 and 10 miRNAs, respectively; anaplastic astrocytoma and oligodendroglioma could be differentiated from other gliomas based on the expression of 65 and 51 miRNAs, respectively. For the GBM group, 13 unique DE miRNAs were identified. After the Pearson correlation analysis and the validation of the correlation pairs in the databases, unique miRNAs that were found to target the DEGs were identified for some subtypes, and these miRNAs were associated with survival. Below, we briefly consider their characteristics.
In diffuse astrocytomas, a decrease in the levels of hsa-let-7c-5p and hsa-miR-125b-5p was detected. In parallel, it was noted that the expression of hsa-let-7c-5p was associated with the level of transcriptional activity of the BEGAIN gene; the link between hsa-miR-125b-5p and the mRNA level of the DUSP3 and RAB3D genes was similarly found. Earlier, in an experimental study on cells of acute erythroleukaemia, the antiproliferative effect of let-7c-5p was demonstrated to be associated with the targeting of the PBX2 oncogene [24]. In colorectal cancer cells after exposure to 5-FU or starvation, Hou N et al. identified a miRNA pool including hsa-let-7c-5p [25]. The identified correlation between the expression of hsa-let-7c-5p and BEGAIN is consistent with the presence of one binding site (target score—90) within the 3′-UTR of the mRNA of this gene.
The BEGAIN protein is involved in the regulation of the activity of postsynaptic neurotransmitter receptors (GO: 0098962); the contribution of this locus to oncogenesis has not been confirmed to date. The antitumour effect of miR-125b has been described in several studies of lung cancer, hepatocellular carcinoma and breast cancer [26,27,28,29]. For example, the study of Yuan M et al. demonstrated the antiproliferative effect of increased expression of hsa-miR-125b-5p in cell lines of diffuse astrocytoma in children [30]. In contrast, a number of studies of endometrial cancer, prostate cancer and GBM indicate the oncogenic effects of hsa-miR-125b-5p, which targets the known tumour suppressors TP53 and GJA1 [31,32,33]. The target of hsa-miR-125b-5p in diffuse astrocytomas is probably the mRNA of the DUSP3 gene, a member of a subfamily of protein phosphatases that negatively regulate the members of the mitogen-activated protein kinase superfamily (MAPK/ERK, SAPK/JNK and p38) and in such a way control proliferation and differentiation. The RAB3D protein (which has three 3′-UTR binding sites for miR-125b-5p) is described as an oncogenic small GTPase that promotes both cell migration in glial tumours and metastasis in breast cancer via the intracellular AKT/GSK3β signalling pathway [34, 35]. In a single tumour, hsa-miR-125b-5p exhibits both oncosuppressive and oncogenic effects.
Anaplastic astrocytomas were characterized by a change in the level of three unique miRNAs: hsa-miR-190b-3p, hsa-miR-196a-3p and hsa-miR-218-2-3p. Correlation analysis allowed us to associate the level of hsa-miR-190b-3p with the expression of the EFNB1 gene, which is involved in cell adhesion and the formation of neural synapses. In this case, the expression of the hsa-miR-190b-3p target was increased, which can be explained by the presence of additional mechanisms regulating the transcription of EFBN1. However, hsa-miR-190b-3p may mitigate the increased expression of EFBN1. An increase in the level of hsa-miR-190b-3p has also been shown in other types of malignant neoplasms. For example, validation of an analysis of 1083 breast cancer samples (from TCGA) confirmed an increased level of expression of hsa-miR-190b-3p in both surgical biopsies and breast cancer cell lines, and this overexpression significantly contributed to cell proliferation and migration [36]. The expression of hsa-miR-190b-3p correlates with the expression of the C8orf34 gene, which is involved in the regulation of gene expression and the cell cycle. Mutations at this genetic locus can mediate severe toxicity during chemotherapy [37].
The increased activity of hsa-miR-196a-3p in colorectal cancer was demonstrated by the study of Wang YN [38]. Similarly, Chen Y. et al. showed that a decrease in hsa-miR-196a-3p expression was associated with a high risk of metastasis in breast cancer. The authors found that transforming growth factor β1 negatively regulates the expression of hsa-miR-196a-3p and activates neuropilin-2, which causes the migration and invasion of tumour cells [39].
Our study demonstrated a decrease in hsa-miR-218-2-3p expression. As a result of the correlation analysis and subsequent database validation, the correlation of hsa-miR-218-2-3p and LMO7 expression was revealed. In contrast, earlier in the Feng Z study, overexpression of hsa-miR-218-2-3p was detected both in cell lines and in surgical biopsy specimens of glial tumours of various grades. In the study, an increase in the activity of has-miR-218-2-3p was associated with the invasive and proliferative potential of glial tumours [40]. As a possible target of hsa-218-2-3p, LMO7 is described as a transcription factor and stabilizer of adhesion compounds, a change in the expression of which is associated with some oncological nosologies [41, 42]. Decreased LMO7 expression in mouse models leads to the development of lung adenocarcinoma [43], and low expression in human lung adenocarcinoma was associated with a negative prognosis [42]. However, contrasting results have also been shown, in which high expression of LMO7 was identified as a negative prognostic factor in non-small-cell lung cancer associated with LRIG1 expression [44].
Despite the numerous miRNA-mRNA correlation pairs (n = 75) identified in diffuse oligodendrogliomas, there were no subtype-specific miRNAs among them.
The analysis made it possible to distinguish three DE miRNAs unique to anaplastic oligodendrogliomas: hsa-miR-20b-5p, hsa-miR-466 and hsa-let-7a-5p.
Earlier, in a study of colorectal cancer (CRC), the association of hsa-miR-20b-5p with survival, as well as a negative correlation of this miRNA with one of the E2F5 cell cycle regulators, was noted [45]. In addition, increased levels of this microRNA in the blood were associated with a better prognosis in patients with metastatic CRC receiving chemotherapy based on bevacizumab [46]. Our analysis demonstrated a negative correlation between hsa-miR-20b-5p expression and PRKACB. This genetic locus has been identified as a significant oncogene involved in the progression of endocrine cancer by modulating the signalling activity of cAMP [47, 48]. In turn, non-small-cell lung cancer shows a lower expression level of PRKACB than normal tissues, and an increase in PRKACB transcriptional activity reduces the number of proliferative, colony-forming and invasive cells and increases the incidence of apoptosis [49].
In a study devoted to the development of predictors of GBM treatment, high expression of hsa-let-7a-5p was interpreted as a favourable prognostic factor associated with survival [50]. In breast cancer, hsa-let-7a-5p mediates sensitivity to the proteasome inhibitor bortezomib, regardless of the molecular subtype [51]. The potential significance of let-7a-5p was determined by Zhou X and colleagues in a study of non-small-cell lung cancer [52].
Using layer-by-layer data filtering, the hsa-let-7a-5p and CYP46A1 correlation pair was identified. The CYP46A1 genetic locus encodes cholesterol 24-hydroxylase, an enzyme responsible for eliminating cholesterol in 24S-hydroxycholesterol (24S-OHC, oxysterol) that is specific to the brain [53]. A decrease in CYP46A1 expression was demonstrated in a study by Han M et al., in which in silico analysis of DEGs and subsequent in vitro verification were performed for GBM samples. Single-cell transcriptome sequencing data demonstrated prevailing CYP46A1 expression in neurons, astrocytes and oligodendrocyte progenitor cells compared to tumour cells [54]. Overexpressed CYP46A1 modulated a decrease in colony formation, proliferation and generation of tumour spheroids in vitro due to reduced cholesterol accumulation [54].
The level of hsa-miR-466 was found to be decreased in osteosarcoma and correlated with IRS1 expression. Increased expression of has-miR-466 inhibited the in vitro proliferation and invasion of osteosarcoma cells [55]. Similarly, a decrease in hsa-miR-466 expression has been demonstrated in hepatocellular carcinoma cells, and an increase in hsa-miR-466 expression inhibits proliferation, induces apoptosis and reduces the metastatic potential of the tumour by targeting MTDH [56]. In prostate cancer, a decrease in hsa-miR-466 expression was demonstrated due to the activity of the long non-coding RNA TUC338. The authors also revealed the antiproliferative potential of hsa-miR-466 [57]. However, an increase in hsa-miR-466 transcriptional activity was observed in malignant tumours of the cervix uteri [58].
In our study, increased expression of hsa-miR-466 negatively correlated with the mRNA level of the CKS2 oncogene. CKS2 encodes a regulatory subunit of 2 cyclin-dependent kinases. Increased expression of CKS2 is associated with tumour progression in bladder cancer, hepatocellular carcinoma, breast cancer and GBM [59,60,61,62].
We did not find microRNA-mRNA correlation pairs in the GBM group. However, thirteen expressed miRNAs were specific for GBM. These miRNAs are also very promising for screening panels. It is worth noting that in the TCGA GBM project, there is an extreme lack of miRNA sequencing data. Future studies using next-generation sequencing technology will help uncover new links and mechanisms in GBM oncogenesis.
Previously, studies have been carried out aimed at finding genetic markers for the molecular classification of glial tumours. In the study of Shai R et al. [63], cluster analysis of 170 DEGs made it possible to differentiate the subtypes of astrocytomas, oligodendrogliomas and glioblastomas. However, the number of samples included in the study was relatively small—35 samples of gliomas of various grades of malignancy. In addition, the percentage of errors in predicting survival was relatively high (22%) [63]. Later, a transcriptomic study of 225 samples of glial tumours was carried out. The authors were able to find a relationship between a favourable prognosis and a mutation in IDH1/2, which was most common in G1 tumours. Nonetheless, no subtype-specific genetic signature was found [64]. There are also a number of other research papers devoted to the study of the molecular characteristics of one or two types of glial tumours [65,66,67]. Our study is characterized by a large number of samples and tumour subtypes (939 samples of 5 different subtypes of gliomas were analysed) and an integrative approach, which consists in the analysis of signalling pathways, survival and the search for subtype-specific markers among DEGs and DE miRNAs. The results obtained open up broad prospects in the early diagnosis of certain subtypes of gliomas and in predicting the overall survival of patients using miRNA and mRNA panels. We hope that further experimental studies of expressions subtype-specific miRNAs and mRNAs will confirm the effectiveness of appropriate genetic tests, which will allow clinicians to timely stratify patients depending on the tumour subtype and choose the optimal treatment regimen.