circSMARCA5 Is an Upstream Regulator of the Expression of miR-126-3p, miR-515-5p, and Their mRNA Targets, Insulin-like Growth Factor Binding Protein 2 (IGFBP2) and NRAS Proto-Oncogene, GTPase (NRAS) in Glioblastoma

The involvement of non-coding RNAs (ncRNAs) in glioblastoma multiforme (GBM) pathogenesis and progression has been ascertained but their cross-talk within GBM cells remains elusive. We previously demonstrated the role of circSMARCA5 as a tumor suppressor (TS) in GBM. In this paper, we explore the involvement of circSMARCA5 in the control of microRNA (miRNA) expression in GBM. By using TaqMan® low-density arrays, the expression of 748 miRNAs was assayed in U87MG overexpressing circSMARCA5. Differentially expressed (DE) miRNAs were validated through single TaqMan® assays in: (i) U87MG overexpressing circSMARCA5; (ii) four additional GBM cell lines (A172; CAS-1; SNB-19; U251MG); (iii) thirty-eight GBM biopsies; (iv) twenty biopsies of unaffected brain parenchyma (UC). Validated targets of DE miRNAs were selected from the databases TarBase and miRTarbase, and the literature; their expression was inferred from the GBM TCGA dataset. Expression was assayed in U87MG overexpressing circSMARCA5, GBM cell lines, and biopsies through real-time PCR. TS miRNAs 126-3p and 515-5p were upregulated following circSMARCA5 overexpression in U87MG and their expression was positively correlated with that of circSMARCA5 (r-values = 0.49 and 0.50, p-values = 9 × 10−5 and 7 × 10−5, respectively) in GBM biopsies. Among targets, IGFBP2 (target of miR-126-3p) and NRAS (target of miR-515-5p) mRNAs were positively correlated (r-value = 0.46, p-value = 0.00027), while their expression was negatively correlated with that of circSMARCA5 (r-values = −0.58 and −0.30, p-values = 0 and 0.019, respectively), miR-126-3p (r-value = −0.36, p-value = 0.0066), and miR-515-5p (r-value = −0.34, p-value = 0.010), respectively. Our data identified a new GBM subnetwork controlled by circSMARCA5, which regulates downstream miRNAs 126-3p and 515-5p, and their mRNA targets IGFBP2 and NRAS.


Introduction
Glioblastoma multiforme (GBM) is the most common and aggressive malignant tumor among those affecting the central nervous system (CNS), with an average annual

The miRNAome Expression Profile Is Dysregulated upon circSMARCA5 Overexpression in U87MG
To check for circSMARCA5-mediated regulation of the microRNAome in GBM cells, the expression of 748 miRNAs was assayed in U87MG transfected for 24 h with the plasmid vector expressing circSMARCA5 or with an empty pcDNA3.1 vector (NC) through realtime PCR, by using two different sets of TaqMan ® Array MicroRNA Cards (A and B) (see Section 4). As shown in Figures 1A and S1, a total of 11 and 17 miRNAs were differentially expressed (DE) between U87MG overexpressing circSMARCA5 and NCs, in Cards A and B, respectively. Among DE miRNAs, 15 were upregulated and 13 downregulated in U87MG overexpressing circSMARCA5 as compared to NCs. Analysis performed with DIANA miRPath showed an involvement of DE miRNAs in biological processes (BPs) and pathways related to glioma (Figure 1B,C; Table S1). ascertained to play a critical role in the pathogenesis and progression of GBM [38][39][40][41][42][43][44].
To expand our knowledge of the pathways regulated by circSMARCA5, in this study we investigated its potential role as an upstream regulator of the microRNAome (miRNAome) in GBM cells.

The miRNAome Expression Profile Is Dysregulated upon circSMARCA5 Overexpression in U87MG
To check for circSMARCA5-mediated regulation of the microRNAome in GBM cells, the expression of 748 miRNAs was assayed in U87MG transfected for 24 h with the plasmid vector expressing circSMARCA5 or with an empty pcDNA3.1 vector (NC) through real-time PCR, by using two different sets of TaqMan ® Array MicroRNA Cards (A and B) (see Section 4). As shown in Figures 1A and S1, a total of 11 and 17 miRNAs were differentially expressed (DE) between U87MG overexpressing circSMARCA5 and NCs, in Cards A and B, respectively. Among DE miRNAs, 15 were upregulated and 13 downregulated in U87MG overexpressing circSMARCA5 as compared to NCs. Analysis performed with DIANA miRPath showed an involvement of DE miRNAs in biological processes (BPs) and pathways related to glioma (Figure 1B,C; Table S1).

miRNAs 126-3p and 515-5p Are Downregulated in GBM Biopsies and Their Expression Positively Correlates with That of circSMARCA5
To further analyse the correlation between circSMARCA5 and miRNAs 126-3p, 515-5p, and 1257, their expression was initially assayed in a training set cohort made up of five GBM and five UC samples. As shown in Figure S2, miRNAs 126-3p and 515-5p were significantly downregulated, while miR-1257 was upregulated in GBM vs. unaffected  To further analyse the correlation between circSMARCA5 and miRNAs 126-3p, 515-5p, and 1257, their expression was initially assayed in a training set cohort made up of five GBM and five UC samples. As shown in Figure S2, miRNAs 126-3p and 515-5p were significantly downregulated, while miR-1257 was upregulated in GBM vs. unaffected brain parenchyma (UC). The expression of miR-126-3p, miR-515-5p, and circSMARCA5 was then assayed in a validation set cohort made up of 38 GBM and 21 UC samples: circSMARCA5, miR-126-3p, and miR-515-5p were confirmed as downregulated in GBM biopsies, as compared to UCs (FC = −2.32, −1.59, and −2.82, p-values = 9.34 × 10 −9 , 7.9 × 10 −4 and 2.97 × 10 −6 , Student's t-test, respectively) ( Figure 3A). MiR-126-3p and miR-515-5p were also positively correlated with circSMARCA5 (r-values = 0.49 and 0.50, p-values = 9 × 10 −5 and 7 × 10 −5 , Spearman's correlation test, respectively) ( Figure 3B) and downregulated in all the GBM cell lines used in this study, as compared to brain cells from healthy donors ( Figure S3).  Figure 3A). MiR-126-3p and miR-515-5p were also positively correlated with circSMARCA5 (r-values = 0.49 and 0.50, p-values = 9 × 10 −5 and 7 × 10 −5 , Spearman's correlation test, respectively) ( Figure 3B) and downregulated in all the GBM cell lines used in this study, as compared to brain cells from healthy donors ( Figure S3). Correlogram showing correlations among circSMARCA5, miR-126-3p, and miR-515-5p. The colour of the circle is linked to the type of correlation (colours from light blue to dark blue and from light red to dark red are representative of positive and negative correlation, respectively); the size of the circle is inversely proportional to the p-value (the bigger is the circle, the less is the p-value). The r-values have been calculated by applying Spearman's correlation test (*** p-value < 0.001, **** p-value < 0.0001).

IGFBP2 (Target of miR-126-3p), NRAS and ROCK1 (Targets of miR-515-5p) mRNAs Are Downregulated upon circSMARCA5 Overexpression in U87MG
To widen our knowledge of the identified circSMARCA5/miR-126-3p/miR-515-5p axis, we searched for the targets of the two candidate miRNAs. A manual literature search allowed us to initially select 17 validated targets (9 of miR-126-3p and 8 of miR-515-5p) in GBM or other neoplasias (Table S2). This first selection was followed by the analysis of the Tumor Cancer Gene Atlas (TCGA), whose data were retrieved from the University of Alabama Cancer Database (UALCAN) to search for upregulated targets in GBM as compared to UCs. Based on TCGA data, we focused on eight mRNA targets (IGFBP2, NRAS, PLXNB2, ROCK1, SOD2, TCF12, TRIP13, and VCAM1) ( Table 2), whose expression was first assayed in U87MG overexpressing circSMARCA5. Among the assayed targets, IGFBP2, NRAS, and ROCK1 mRNAs were significantly downregulated (FC IGFBP2, NRAS, ROCK1 = −1.97, −1.33, and −1.31, respectively) ( Figure 4). Correlogram showing correlations among circSMARCA5, miR-126-3p, and miR-515-5p. The colour of the circle is linked to the type of correlation (colours from light blue to dark blue and from light red to dark red are representative of positive and negative correlation, respectively); the size of the circle is inversely proportional to the p-value (the bigger is the circle, the less is the p-value). The r-values have been calculated by applying Spearman's correlation test (*** p-value < 0.001, **** p-value < 0.0001).

IGFBP2 and NRAS mRNAs Are Upregulated in GBM Biopsies and Cell Lines with Respect to UCs and Their Expression Negatively Correlates with That of circSMARCA5
The expression of IGFBP2, NRAS, and ROCK1 mRNAs was then assayed in the same validation set cohort (made of 38 GBM and 21 UC samples) used to verify the differential expression of candidate DE miRNAs.

IGFBP2 and NRAS mRNAs Are Upregulated in GBM Biopsies and Cell Lines with Respect to UCs and Their Expression Negatively Correlates with That of circSMARCA5
The expression of IGFBP2, NRAS, and ROCK1 mRNAs was then assayed in the same validation set cohort (made of 38 GBM and 21 UC samples) used to verify the differential expression of candidate DE miRNAs. IGFBP2 and NRAS were upregulated in GBM biopsies and cell lines with respect to UCs (Figures 5A and S4) and their expression negatively correlated with that of circSMARCA5 (r-values = −0.58 and −0.30, p-values = 0 and 0.019, Spearman's correlation test, respectively), and with that of their negative regulators, miR-126-3p and miR-515-5p, respectively (r-values = −0.36 and −0.34, p-values = 0.0066 and 0.010, Spearman's correlation test, respectively) ( Figure 5B). IGFBP2 mRNA and protein were significantly upregulated in classical (C) and mesenchymal (M) GBM subtypes, when compared to the proneural (P) and neural (N) ones, based on TCGA data (Figures S5 and S6); NRAS mRNA was, instead, significantly upregulated in the N GBM subtype when compared to the other subtypes ( Figure S7). These data are in agreement with the worst prognosis of M-subtype patients showing a higher IGFBP2 mRNA expression (p-value = 0.023, Kaplan-Meier survival curve comparison) ( Figure 5C).
IGFBP2 and NRAS mRNAs were positively correlated based on our dataset (r-value = 0.46, p-value = 0.00027) ( Figure 5B), and GBM TCGA and normal brain cortex GTEx gene expression data (r-value = 0.76, p-value = 4.5 × 10 −52 ) ( Figure S8). To extend the pathway downstream to IGFBP2, we assayed the expression of the vascular endothelial growth factor A (VEGFA) mRNA (whose transcription is known to be activated by IGFBP2 protein that acts as an enhancer on its promoter-see Section 3) in (i) U87MG overexpressing circSMARCA5 and (ii) in the same validation set cohort used to verify the differential expression of IGBP2 mRNA. VEGFA mRNA was significantly downregulated in U87MG upon circSMARCA5 overexpression and its expression positively correlated with that of IGFBP2 in the validation set cohort used in this study ( Figure S9). mRNAs were positively correlated based on our dataset (r-value = 0.46, p-value = 0.00027) ( Figure 5B), and GBM TCGA and normal brain cortex GTEx gene expression data (r-value = 0.76, p-value = 4.5 × 10 −52 ) ( Figure S8). To extend the pathway downstream to IGFBP2, we assayed the expression of the vascular endothelial growth factor A (VEGFA) mRNA (whose transcription is known to be activated by IGFBP2 protein that acts as an enhancer on its promoter-see Section 3) in (i) U87MG overexpressing circSMARCA5 and (ii) in the same validation set cohort used to verify the differential expression of IGBP2 mRNA. VEGFA mRNA was significantly downregulated in U87MG upon circSMARCA5 overexpression and its expression positively correlated with that of IGFBP2 in the validation set cohort used in this study ( Figure S9).

Discussion
The pathway leading to the expression of miRNAs is very complex and tightly regulated by several factors at different steps [64]. Altered expression of the microRNAome in GBM has been extensively studied, although the molecular mechanisms steering specific miRNA dysregulation have been elucidated only in a few cases [65][66][67]. Numerous RBPs play a crucial role in the post-transcriptional processing of pri-miRNAs and pre-miRNAs [68]: among them, the splicing factor SRSF1 has been demonstrated to regulate the expression of several mature miRNAs by a cross-talk with the enzymes involved in the processing of their precursors, through mechanisms that have been only partially explained to date [69]. The interplay between circRNAs and miRNAs has been mainly described as a ceRNA network, in which circRNAs, including circSMARCA5, act as sponges for miR-NAs [24][25][26][27][70][71][72][73][74]; however, circRNAs have not been yet reported as upstream regulators of miRNA expression, to the best of our knowledge. Because of their role as decoys for several RBPs, here we hypothesize that circRNAs may act as upstream epigenetic regulators of the miRNAome inside cells. In this work, we specifically investigated the circSMARCA5mediated regulation of the miRNAome in GBM cells. We previously characterized circS-MARCA5 as a TS circRNA in GBM and we demonstrated that it performs its function by sponging the RBP SRSF1 [22,23,75]. Our data ascertained that circSMARCA5 plays a role in the control of the miRNA expression inside GBM cells and that several dysregulated miRNAs upon circSMARCA5 overexpression are involved in glioma pathways. Further investigation led us to focus on miRNAs 126-3p and 515-5p: (i) both were upregulated in U87MG upon circSMARCA5 overexpression; (ii) their expression was positively correlated with that of circSMARCA5; (iii) they have been characterized as TS in GBM and additional cancers by other scholars [45][46][47][48][49][54][55][56][76][77][78][79][80][81][82][83][84][85][86][87]. CircSMARCA5-mediated upstream control of miRNAs 126-3p and 515-5p is also supported by: (i) the observed downregulation of their two selected mRNA targets, IGFBP2 and NRAS, upon circSMARCA5 overexpression; (ii) positive correlation between the expression of IGFBP2 and NRAS mRNAs; (iii) negative correlation between the expression of IGFBP2 and NRAS mRNAs and circSMARCA5. Circ-SMARCA5 is also functionally linked to miRNAs 126-3p, 515-5p and their targets; indeed, similar to circSMARCA5 [21][22][23], miR-126-3p and its target IGFBP2 are involved in GBM progression, by regulating cell migration, invasion [45,[88][89][90], and angiogenesis [91,92]. To deepen the knowledge of the latter molecular axis, we also investigated the expression of VEGFA mRNA, both in U87MG overexpressing circSMARCA5 and in the validation cohort of GBM and UC biopsies. We previously showed that circSMARCA5 affects the ratio between pro-and anti-angiogenic isoforms of VEGFA mRNA in GBM by regulating alternative splicing of VEGFA pre-mRNA, tethering the splicing factor SRSF1 [22]. Here we showed that the amount of pan-VEGFA mRNA decreased in U87MG upon circSMARCA5 overexpression and that VEGFA and IGFBP2 mRNAs were positively correlated. Unless here we focused on an axis involving (non-coding and coding) RNA molecules, data obtained on the expression of VEGFA mRNA suggest that IGFBP2 may be upstream regulated by circSMARCA5 also at the protein level: indeed, IGFBP2 was described as an enhancer for the transcription of VEGFA in neuroblastoma cells [93] and IGFBP2 and VEGFA were shown to be positively correlated at the protein level in GBM tissues [94]. MiR-126-3p can be also carried in biological fluids through EVs [95,96]: it would be interesting to investigate if and how the delivery of this molecule to cells at different sites from the bulk tumor play a role in the cancer progression and resistance to the current therapies. MiR-515-5p and its target NRAS are also known to be involved in GBM progression by regulating cell migration, growth [53][54][55][56]97], and angiogenesis [98,99]. In an attempt to find a link between the upstream regulator circSMARCA5 and the downstream-regulated intronic miRNAs 126-3p and 515-5p, we also searched for RBPs that may commonly bind and, potentially, regulate pre-miR-126 and pre-miR-515 processing. Our prediction allowed us to identify the splicing factor SRSF3 as an RBP that potentially binds both pre-miRNAs. We previously found that SRSF3 splicing is regulated by SRSF1 and, indirectly, by circSMARCA5 in GBM cells, where the pro-oncogenic full-length functional SRSF3 mRNA isoform is overexpressed when compared to the truncated non-functional one [21]. SRSF3 has been described as a direct or indirect positive regulator of the processing of several pri-miRNAs such as pri-miRNAs 30a, 142, and miR-132/212, by interacting with a CNNC motif, recruiting DROSHA to the cleavage site, and enhancing the Microprocessor activity [100,101]. Based on our prediction, SRSF3 would interact with different motifs other than CNNC on pre-miRNA 126 and 515 sequences, paving the way to alternative mechanisms of SRSF3-mediated pri-and, eventually, pre-miRNA processing. Most specifically, based on our data, we speculate that in a GBM cell context and in particular for pre-miRNAs 126 and 515-1 processing, SRSF3 may function as a negative regulator ( Figure 6). As previously reported for the RBP RNA binding fox-1 homolog 3 (RBFOX3), the same RBP can stimulate or block the processing of individual pri-or pre-miRNAs depending on the cell context and the specific miRNA precursor structure [102]. and that VEGFA and IGFBP2 mRNAs were positively correlated. Unless here we focused on an axis involving (non-coding and coding) RNA molecules, data obtained on the expression of VEGFA mRNA suggest that IGFBP2 may be upstream regulated by circSMARCA5 also at the protein level: indeed, IGFBP2 was described as an enhancer for the transcription of VEGFA in neuroblastoma cells [93] and IGFBP2 and VEGFA were shown to be positively correlated at the protein level in GBM tissues [94]. MiR-126-3p can be also carried in biological fluids through EVs [95,96]: it would be interesting to investigate if and how the delivery of this molecule to cells at different sites from the bulk tumor play a role in the cancer progression and resistance to the current therapies. MiR-515-5p and its target NRAS are also known to be involved in GBM progression by regulating cell migration, growth [53][54][55][56]97], and angiogenesis [98,99]. In an attempt to find a link between the upstream regulator circSMARCA5 and the downstream-regulated intronic miRNAs 126-3p and 515-5p, we also searched for RBPs that may commonly bind and, potentially, regulate pre-miR-126 and pre-miR-515 processing. Our prediction allowed us to identify the splicing factor SRSF3 as an RBP that potentially binds both pre-miRNAs. We previously found that SRSF3 splicing is regulated by SRSF1 and, indirectly, by circSMARCA5 in GBM cells, where the pro-oncogenic full-length functional SRSF3 mRNA isoform is overexpressed when compared to the truncated non-functional one [21]. SRSF3 has been described as a direct or indirect positive regulator of the processing of several pri-miRNAs such as pri-miRNAs 30a, 142, and miR-132/212, by interacting with a CNNC motif, recruiting DROSHA to the cleavage site, and enhancing the Microprocessor activity [100,101]. Based on our prediction, SRSF3 would interact with different motifs other than CNNC on pre-miRNA 126 and 515 sequences, paving the way to alternative mechanisms of SRSF3-mediated pri-and, eventually, pre-miRNA processing. Most specifically, based on our data, we speculate that in a GBM cell context and in particular for pre-miRNAs 126 and 515-1 processing, SRSF3 may function as a negative regulator (Figure 6). As previously reported for the RBP RNA binding fox-1 homolog 3 (RBFOX3), the same RBP can stimulate or block the processing of individual pri-or pre-miRNAs depending on the cell context and the specific miRNA precursor structure [102]. Collectively, our data suggest circSMARCA5 as an upstream regulator of the expression of TS miRNAs 126-3p and 515-5p and their downstream targets IGFBP2 and NRAS mRNA in GBM cells, extending our knowledge on the disrupted tumor suppressive pathways mediated by circSMARCA5 in GBM cells. Prospectively, these pathways may be considered for targeted molecular therapeutic approaches, especially by using recently discovered genomic editing techniques [103].

Cell Lines and Biopsies
GBM cell lines A172, CAS-1, SNB-19, U251MG, and U87MG were cultured as described in the supplementary materials and methods. All cell lines were purchased from the Interlab Cell Line Collection (ICLC), located at the IRCCS Ospedale Policlinico San Collectively, our data suggest circSMARCA5 as an upstream regulator of the expression of TS miRNAs 126-3p and 515-5p and their downstream targets IGFBP2 and NRAS mRNA in GBM cells, extending our knowledge on the disrupted tumor suppressive pathways mediated by circSMARCA5 in GBM cells. Prospectively, these pathways may be considered for targeted molecular therapeutic approaches, especially by using recently discovered genomic editing techniques [103].

Cell Lines and Biopsies
GBM cell lines A172, CAS-1, SNB-19, U251MG, and U87MG were cultured as described in the supplementary materials and methods. All cell lines were purchased from the Interlab Cell Line Collection (ICLC), located at the IRCCS Ospedale Policlinico San Martino, Genova, Italy. Cell lines were used between the 5th and 10th passage and their viability was assessed through the Trypan Blue Exclusion Test (ThermoFisher Scientific, Waltham, MA, USA) before each experiment, according to the protocol described by W Strober [104]. In total, 38 GBM and 21 UC biopsies were obtained, characterized by pathologists, and stored until their processing, as previously described [22]. Informed consent was signed by the patients before surgery. Demographic data of the patients enrolled in this study are summarized in Table S3. The entire study was performed according to the Declaration of Helsinki and approved by the local ethical Committee of the Azienda Ospedaliero-Universitaria "Policlinico-Vittorio Emanuele", Catania, Italy (project identification code: 166/2015/PO, 17 December 2015).

Cell Transfection
U87MG cells were transfected by using lipofectamine 2000 (Thermofisher Scientific, Waltham, MA, USA), as previously described [21]. Briefly, 5 × 10 4 cells were seeded in a 24-well plate, cultured for 24 h, and transfected with 500 ng of NC or the vector expressing circSMARCA5 (pcDNA3.1_circSMARCA5) for 24 h, according to the manufacturer's instructions. Three replicates for each experimental condition were carried out and analysed accordingly. Data on circSMARCA5 overexpression upon transfection of U87MG are reported in Figure S11.

RNA Extraction
Total RNA was isolated through Trizol™ (Thermofisher Scientific, Waltham, MA, USA) in accordance with the manufacturer's instructions and quantified by spectrophotometry, as previously described [105]. FirstChoice ® Human Brain Reference RNA (Ambion, Austin, TX, USA) was used as an additional UC.

microRNA TaqMan ® Arrays
MicroRNA TaqMan ® Arrays (Thermofisher Scientific, Waltham, MA, USA) were performed as previously described [106]. Briefly, 300 ng of total RNA isolated from each of the three biological replicates of U87MG, transfected for 24 h with the vector pcDNA3.1_circSMARCA5 or with NC, were reverse-transcribed into specific cDNAs of 748 microRNAs using the TaqMan™

Array Data Analysis
EDS files generated by the run of microRNA TaqMan ® Arrays were imported into the dashboard of a ThermoFisher cloud (https://apps.thermofisher.com/apps/spa/#/ dashboard, accessed on November 2021) and then analysed through relative quantification application. Cycle thresholds (Ct s ) were then calculated by the software and exported in a CSV file. Data from Cards A and B were analysed independently. Briefly, miRNAs that showed Ct s > 35 in all the experimental conditions were considered too late and filtered out from data analysis. Correlations between the mean or median Ct value of each card and the Ct value of each miRNA were calculated to select candidate housekeeping (HK) miRNAs. A selection of 15 and 3 candidate HK miRNAs, among those showing the strongest correlation with the mean and median Ct values, were given as input to RefFinder (http://blooge.cn/RefFinder/?type=reference, accessed on November 2021) to select the best reference miRNAs within Cards A and B, respectively (Table S4). Reference miRNAs (miR-192-5p and miR-106a-5p for Card A; miR-452-3p and miR-19a-5p for Card B) were used to obtain DCt s , (Ct of the transcript of interest-Ct of the reference transcript). DCt s of 190 and 58 miRNAs were given as input to the MeV (Multiple Experiment Viewer) tool v. 4.7.1 to retrieve significant DE miRNAs within Cards A and B, respectively. A graphical representation of microRNA TaqMan ® Arrays' data analysis is reported in Figure S12.

Real-Time PCR
DE miRNAs were validated through single TaqMan™ microRNA assays. Briefly, 30 ng of total RNA were reverse transcribed through the TaqMan™ microRNA Reverse Transcription Kit (ThermoFisher Scientific, Waltham, MA, USA) by using miRNA-specific primers and then amplified through the TaqMan™ Universal Master Mix II (ThermoFisher Scientific, Waltham, MA, USA) by using specific TaqMan™ assays, according to the manufacturer's instructions. Messenger RNAs of candidate miRNA targets were amplified by using the Power SYBR™ Green RNA-to-Ct™ 1-Step Kit (ThermoFisher Scientific, Waltham, MA, USA). PCRs were run in a QuantStudio™ 5 Real-Time PCR System (ThermoFisher Scientific, Waltham, MA, USA). Real-time PCR data were represented as −1*DCt s , FC, or log FC within the text (see supplementary methods for further explanation). The list of TaqMan™ assays and primers used in this study is shown in Table S5.

In Silico Analyses
BPs and pathways regulated by DE miRNAs were retrieved through DIANA miRPath 3.0 [107]: validated targets stored in TarBase v. 7.0 were selected to calculate BP and pathway enrichment. The expression of miRNA targets from the GBM TCGA dataset was retrieved through the UALCAN, GBM-BioDP, and Gene Expression Profiling Interactive Analysis (GEPIA) databases [108][109][110]. Multiple Em for Motif Elicitation (MEME) suite v. 5.4.1 [111] was used to retrieve RNA motifs within pre-miRNA sequences, using default parameters. ATtRACT database v. 0.99β [112] identified RBPs validated to bind specific RNA motifs. The RNA Structure tool [113] was used to calculate and visualize the secondary structures common to pre-miRNA sequences.

Statistical Analyses
Pearson's and Spearman's correlation tests were used to calculate correlations between the mean or median Ct value of each card and the Ct value of each miRNA, in order to select candidate housekeeping (HK) miRNAs. Only miRNAs showing correlation coefficients (r-values) ≥ 0.8 and p-value < 0.05 were considered as candidate DE miRNAs to be given as input to RefFinder. DE miRNAs were calculated through the Significance Analysis of Microarray (SAM) method within MeV tool v. 4.7.1 [114]. Only miRNAs reporting q-values = 0 were considered DE. Correlation tests and statistical significance were performed and calculated through GraphPad Prism v. 8.0.2. Student's t-test was used to identify DE miRNAs and targets after single real-time PCR assays; for this, p-values < 0.05 were considered significant. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Written informed consent was obtained from the patients to publish this paper.

Data Availability Statement:
The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.