High-throughput screening uncovers miRNAs enhancing glioblastoma cell susceptibility to tyrosine kinase inhibitors

Abstract

Glioblastoma (GBM) is a deadly and therapy resistant malignant brain tumour, characterized by an aggressive and diffuse growth pattern, which prevents complete surgical resection. Despite advances in the identification of genomic and molecular alterations that fuel the tumour, average patient survival post-diagnosis remains very low (∼14.6-months). In addition to being highly heterogeneous, GBM tumour cells exhibit high adaptive capacity to targeted molecular therapies owing to an established network of signalling cascades with functional redundancy, which provides them with robust compensatory survival mechanisms. Here, we investigated whether a multimodal strategy combining multitargeted tyrosine kinase inhibitors (MTKIs) and microRNA (miRNA) modulation could overcome the signalling pathway redundancy in GBM and, hence, promote tumour cell death. By performing a high-throughput screening, we identified a myriad of miRNAs, including those belonging to the miR-302-3p/372-3p/373-3p/520-3p family, which coordinately act with the MTKI sunitinib to decrease GBM cell viability. Two members of this family, hsa-miRNA-302a-3p and hsa-miRNA-520 b, were found to modulate the expression of receptor tyrosine kinase mediators (including AKT1, PIK3CA and SOS1) in U87 and DBTRG human GBM cells. Importantly, administration of mimics of these miRNAs with sunitinib or axitinib resulted in decreased tumour cell proliferation and enhanced cell death, whereas no significant effect was observed when coupling miRNA modulation with temozolomide, the first-line drug for GBM therapy. Overall, our results provide evidence that combining the ‘horizontal’ inhibition of signalling pathways promoted by MTKIs with the ‘vertical’ inhibition of the downstream signalling cascade promoted by hsa-miR-302a-3p and hsa-miR-520 b constitutes a promising approach towards GBM treatment.

Introduction

Glioblastoma (GBM) is a fast growing, highly heterogeneous and therapy-resistant cancer, being the most common and malignant type of primary brain tumour. Despite being rare when compared to other tumours (only 2–3 cases per 100,000 habitants), GBM accounts as one of the deadliest human cancers with a 5-year survival rate of only 5%, due to the inefficacy of the available treatment options (1,2). Additionally, a striking and diffuse growth resulting in the invasion of the healthy brain tissue makes complete surgical resection impossible, which leads to tumour recurrence after partial response to treatment. Our current understanding of the complex biology of gliomas is largely derived from studies that characterized (epi)genetic and molecular abnormalities in GBM, the most prevalent being linked to key signalling networks such as the p53 and the receptor tyrosine kinase (RTK)/Ras/phosphoinositide 3-kinase (PI3K) pathways (3–7). In fact, accumulated evidence shows that GBM cells are equipped with functionally redundant signalling networks, providing them with a robust mechanism for adaptation to targeted molecular therapeutic approaches, ultimately preserving tumour cell survival. These observations help to explain the lack of success in the development of effective therapies for GBM, but also provide a new direction for the search of novel strategies aiming at blocking, simultaneously, multiple signalling cascade events (8). In this regard, a therapeutic intervention that includes multitargeted tyrosine kinase inhibitors (MTKIs) might be of great interest towards a successful clinical application. MTKIs are small molecule inhibitors of RTK signalling that act by displacing or blocking ATP binding at the catalytic site of the receptors, thus arresting RTK activity and blocking the downstream cascade of events. This family of molecules include drugs that have successfully progressed into the clinic and that are currently being used for anticancer treatments (e.g. erlotinib, sunitinib and axitinib), although to date none of them has been employed for GBM treatment (9). Importantly, several MTKIs have been evaluated in clinical trials for GBM. Sunitinib, a first generation MTKI and anti-angiogenic drug, was evaluated in phase I (10) and phase II (11) clinical trials but did not show significant antitumour activity. Nonetheless, a different treatment regimen of sunitinib is currently being tested in a phase II/III clinical trial for patients with recurrent GBM (12). On the other hand, the second generation MTKI axitinib was shown to perform better than the physician’s best alternative choice of therapy in patients with recurrent GBM (bevacizumab or lomustine), as assessed in a recent randomized phase II clinical trial enrolling 44 patients (13). However, it has been reported that the intrinsic or developed resistance of some cancer cells to RTK inhibition is related to functionally redundant pathways that are able to compensate for the blocked receptors (extensively reviewed in 14,15). Numerous studies provided evidence for the pivotal contribution of microRNAs (miRNAs) to the pathogenesis of most—if not all—human malignancies. MiRNA deregulation is known to underpin several GBM hallmarks (16), including the intrinsic resistance to chemotherapy, and restoration of the expression of key miRNAs that control GBM malignancy has proven to be detrimental to tumour cells across several studies (17–21). Our previous results have shown that silencing of miR-21, highly overexpressed in GBM, increases apoptosis and significantly enhances cytotoxicity of sunitinib in GBM cells, and importantly, when combined with drug treatment, leads to a remarkable antitumour effect in an orthotopic GBM mouse model (22,23). Taking advantage of their capacity to modulate key cellular processes by targeting multiple molecular targets, we aimed at using miRNAs to cooperatively act with MTKIs in order to overcome/bypass the functional signalling redundancy of GBM cells and, thus, to effectively tackle GBM pathogenesis.

Here, we have employed a high-throughput screening analysis using a genome-wide library of miRNA mimics, in order to discover additional miRNAs capable of rendering GBM cells highly susceptible to sunitinib. Studies performed to address the effects of modulating key miRNAs, followed by MTKI administration, on GBM cell proliferation, cell DNA content and cell death provided evidence that RTK inhibition in combination with the downregulation of downstream signalling molecules holds promise as a prospective strategy towards GBM treatment.

Results

The miR-302/520 family potentiates the toxic effects of sunitinib, a multitargeted tyrosine kinase inhibitor, on human GBM cells

In order to identify miRNAs that render GBM cells susceptible to the action of MTKIs and thus enhance their therapeutic outcome, a high-throughput screening analysis was performed, using a genome-wide library of microRNA mimics (988 miRNAs, 875 unique sequences, miRBase 13.0 http://mirbase.org, Fig. 1 and Supplementary Material, Fig. S1). Human U87 GBM cells, widely employed as a cellular model for GBM studies, were transfected with the library of miRNA mimics at 50 nM (selected on the basis of the resulting biological effect achieved with minimal toxicity) in combination, or not, with sunitinib, a first generation MTKI currently being tested in phase II/III clinical trials for GBM treatment (12) (Fig. 1A). For these studies, sunitinib was used at 15 µM, which induces only a modest decrease (ca. 10%) of cell viability. The effect of the addition of miRNA mimics on cell viability, either per se or in combination with 15 µM sunitinib, was evaluated for each miRNA and compared to that of cells transfected with a control miRNA mimic (cel-miR-67). As shown in Figure 1B, the screening identified a myriad of miRNAs that render GBM cells susceptible to sunitinib (195 miRNAs increasing susceptibility by at least 2-fold). Interestingly, among the miRNAs exhibiting a stronger effect (Supplementary Material, Fig. S1, top 50 miRNAs), we could identify miR-128, reported to target RTK transducers (17) and to be highly detrimental toward GBM cells (24), and several miRNAs belonging to miR-302-3p/372-3p/373-3p/520-3p family, hereafter referred to as miR-302/520 family (8 in the top 20, highlighted in bold in Supplementary Material, Fig. S1). Of note, the 12 family members that share a common 7-mer seed sequence (Fig. 1C) exhibit a similar phenotype (Fig. 1D). The screening (Fig. 1B) also revealed miRNAs that, per se, resulted in decreased cell viability. Although these miRNAs could be considered as potential therapeutic targets in a strategy against GBM, we focused on a combined approach encompassing miRNA modulation and MTKI treatment, which we reasoned would sustain a robust blockade of the established GBM functional redundancy. Interestingly, the use of DIANA-miRPath (25), a computational tool able to correlate miRNAs with KEGG pathways (based on the predicted targets), indicated that the glioma de novo pathway, as well as the ErbB, mitogen-activated protein kinase (MAPK), Ras and mechanistic target of rapamycin (mTOR) signalling cascades are among the top KEGG pathways with predicted targets of the miR-302/520 family (Supplementary Material, Table S1). This in silico analysis suggests that, indeed, these miRNAs might affect some of the pathways initiated by receptor tyrosine kinases, corroborating the rationale of the screening. Altogether, based on the fact that several miR-302/520 family members are among those whose modulation promoted the most pronounced sensitization of GBM cells to sunitinib and that this family of miRNAs is predicted to alter key signalling pathways, we selected two representative members of the miR-302/520 family, hsa-miRNA-302a-3p and hsa-miRNA-520 b, to be modulated in combination with MTKIs, for carrying out our study.

Figure 1.

MiR-302/520 miRNA family increases susceptibility of GBM cells to sunitinib treatment. (A) Overview of the high-throughput screening procedure. U87 GBM cells were transfected with a genome-wide library of microRNA mimics arrayed on 384-well plates and 48 h later were treated with sunitinib (15 μM), or with vehicle (DMSO). (B) Viability of cells transfected with the genome-wide library of microRNA mimics, followed by treatment with vehicle (DMSO) or sunitinib. (C) Mature sequences of 12 miR-302/520 family members. The common 7-mer seed sequence is highlighted in grey. (D) Viability of the cells transfected with miR-302/520 family members and treated with vehicle (DMSO) or sunitinib. Cell viability was evaluated by measuring ATP content (ATPlite). Results were obtained from two independent screening replicates (n = 2); non-transfected cells (no miRNA) and cel-miR-67 (control miRNA) were used as negative controls.

MiR-302a and miR-520 b downregulate the expression of tyrosine kinase signal transducers

To understand the molecular cues underpinning the cumulative toxicity of miR-302/520 family and sunitinib on U87 GBM cells, we searched for mRNAs that could simultaneously be predicted as potential targets of this miRNA family by the TargetScanHuman (v7.1) and MicroT-CDS (v7.0) web-based computational tools (26–29). Focusing on those mRNAs related to the same signalling pathways which are affected by MTKIs (i.e. cell signalling by receptor tyrosine kinases), we were able to identify AKT serine/threonine kinase 1 (AKT1), SOS Ras/Rac guanine nucleotide exchange factor 1 (SOS1) and phosphatidylinositol-4, 5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA) as potential targets of miR-302/520 miRNA family, with a cumulative TargetScan score of −0.29, −0.17 and −0.10, respectively. In this regard, AKT1 has been previously identified as a direct target of the miR-302 cluster (composed of miR-302a-d and miR-367) in cervical cancer cells (30) and PIK3CA is among the subset of genes that are significantly mutated in GBM tumours, which results in the amplification of its expression (6).

In order to assess the biological relevance of the aforementioned bioinformatic predictions, we transfected human GBM cells with mimics of hsa-miRNA-302a-3p and hsa-miRNA-520 b (hereafter referred to as miR-302a and miR-520 b), two miRNAs from the miR-302/520 family, and evaluated the mRNA expression levels of AKT1, SOS1 and PIK3CA by qPCR (Fig. 2). MiR-302a and miR-520 b were chosen based on previous evidence showing that these miRNAs hold promise in anticancer approaches (30–34), namely towards GBM (35–37). Transfection of U87 and DBTRG cells with miR-302a and miR-520 b mimics resulted in a striking increase in the levels of miR-302a and miR-520 b, which were barely detected by qPCR before transfection (data not shown). Following miRNA modulation, a moderate, though not statistically significant, decrease in the mRNA expression levels of AKT1, PIK3CA and SOS1 was observed in both cell lines as compared to cells transfected with the control miRNA (Fig. 2A and B). Having been validated as a direct target of miR-302a (38) and miR-520 b (39), cyclin dependent kinase inhibitor 1 A (CDKN1A) was additionally assessed in terms of the mRNA expression levels, which were found to be significantly decreased in U87 and DBTRG cells after miR-302a and miR-520 b modulation (data not shown). The Western blot analysis showed that PIK3CA protein levels are significantly reduced after miR-302a and miR-520 b modulation in both U87 (Fig. 2C and I) and DBTRG (Fig. 2F and L) cells, which is consistent with the decrease observed for the corresponding mRNA levels (Fig. 2A and B). On the other hand, the phospho-AKT (pAKT) to AKT ratio was reduced in DBTRG cells after miRNA modulation, but not in U87 cells (Fig. 2D and G). The expression levels and activation (phosphorylation) of MAPK3/MAPK1 (ERK1/2), a signalling cascade mediator downstream of SOS1, were not significantly altered upon miRNA modulation.

Figure 2.

Treatment of GBM cells with mimics of miR-302a and miR-520 b reduces the expression of their potential targets. U87 and DBTRG cells were collected for RNA isolation and subsequent qPCR (A, B), and Western blot analysis (C–N) after transfection for 48 h with miR-302a, miR-520 b or control miRNA (cel-miR-239 b) mimics. mRNA (A, B) and protein (C–H) levels were normalized to the reference gene hypoxanthine phosphoribosyltransferase 1 (HPRT1) and to housekeeping control α-tubulin (tubulin), respectively, and presented as relative expression values to control cells. Phospho-AKT (p-AKT Ser473) and phospho-ERK1/2 (p-ERK1/2 Thr202/Tyr204) protein levels are presented as a ratio to the corresponding total AKT (D, G) and ERK1/2 (E, H) protein levels. Representative results from Western blot analysis are shown for PIK3CA (I, L), p-AKT and AKT (J, M), p-ERK1/2 and ERK1/2 (K, N). *P < 0.05, **P < 0.01 as compared to the control miRNA mimic transfection.

MiR-302a and miR-520 b increase the susceptibility of U87 and DBTRG cells to sunitinib and axitinib

After identifying miRNAs whose modulation enhances the sensitivity of U87 cells to sunitinib, we investigated whether those miRNAs would play a similar role in different human GBM cells, not only following treatment with sunitinib, but also upon exposure to other chemotherapeutic drugs with relevance to GBM. For that, the efficacy of treatment with sunitinib, axitinib or with temozolomide (TMZ; the standard first-line agent for GBM) (1), per se or in combination with miRNA modulation, was evaluated in U87 and DBTRG cells, by measuring cell viability using the Alamar blue assay (Fig. 3). Chemotherapeutic concentrations of 15 µM for sunitinib, 5 µM for axitinib and 200 µM for TMZ were selected as being those that promoted a modest viability decrease (ca. 10%) on the basis of dose-response curves obtained over a range of drug concentrations from 1 to 600 µM (data not shown). Of note, both sunitinib and axitinib exerted the same cytotoxic effect as TMZ at significantly lower concentrations (10- to 40-fold lower doses). As shown in Figure 3, transfection of U87 and DBTRG cells with miR-302a or miR-520 b mimics for 48 h resulted in a 10–20% decrease in cell viability with respect to that observed for cells transfected with the control miRNA mimic, although only in DBTRG cells this reduction was statistically significant (Fig. 3B). More importantly, a dramatic increase in susceptibility to MTKIs (sunitinib and axitinib) was achieved upon miRNA (miR-302a or miR-520 b) modulation, an effect that was evident for both cell lines. No relevant effect was observed when temozolomide was included in the combined therapy, except when coupled with miR-302a mimic in DBTRG cells. The viability of U87 and DBTRG cells was also assessed by measuring the intracellular levels of ATP using the ATPlite assay (Supplementary Material, Fig. S2). As shown, a similar trend to that found by the Alamar blue assay was observed in terms of the increase in MTKI susceptibility of cells transfected with both miRNAs, although the detected effects have been less pronounced, particularly regarding DBTRG cells.

Figure 3.

Combination of miRNA-302a/520 b expression with multitargeted tyrosine kinase inhibitors decreases GBM cell viability. U87 and DBTRG cells were transfected with 50 nM of miRNA-302a, miR-520 b or control miRNA (cel-miR-239 b) mimics for 48 h. Cells were subsequently exposed to sunitinib (Sun, 15 μM), axitinib (Axi, 5 μM) or temozolomide (TMZ, 200 μM) for 24 h, after which the cell viability was evaluated by the Alamar Blue assay in U87 (A) or DBTRG (B) cells. Results are normalized to non-transfected and non-treated cells. *P < 0.05, **P < 0.01, ***P < 0.001 compared to the respective control miR mimic; ^#P < 0.05, ^##P < 0.01 relative to transfection per se.

The multimodal strategy decreases the proliferation of U87 and DBTRG GBM cells

After demonstrating the reduction in GBM cell viability through the combination of miR-302a or miR-520 b modulation with sunitinib or axitinib treatment, we evaluated the ability of this multimodal strategy in reducing the proliferation rate of GBM cells. After miRNA transfection for 48 h, U87 and DBTRG cells were treated with 15 µM sunitinib or with 5 µM axitinib over 72 h. Cell proliferation was assessed in terms of cellular protein content, using the SRB assay, at 48, 72, 96 and 120 h after transfection, the results being expressed as percentage of protein content relative to control cells at the time point of 48 h (Fig. 4). The treatment of U87 cells with sunitinib combined with miR-520 b modulation significantly inhibited cell proliferation, 36% and 70% growth inhibition being achieved at the 96 and 120 h time points, respectively, as compared to the control miRNA transfection (Fig. 4A). On the other hand, the combination of miR-302a modulation with sunitinib treatment decreased, although not significantly, cell proliferation at 120 h after transfection in U87 GBM cells (Fig. 4A). However, no effects were detected in the proliferation of DBTRG cells treated with sunitinib following miRNA modulation (Fig. 4B). As shown in Figure 4C, the combination of axitinib treatment with miR-302a or miR-520 b modulation resulted in a significant growth inhibition (ca. 43%) of U87 GBM cells at 120 h after transfection, as compared to the control miRNA transfection. In DBTRG cells, the expression of both miRNAs in combination with axitinib was detrimental to cell proliferation at 120 h, a significant decrease upon transfection being promoted by miR-302a as compared to control miRNA mimic (146.87 ± 26.53 vs 188.80 ± 4.32, Fig. 4D). Figure 4E and F displays phase contrast images of cells submitted to the combined therapies at the 120 h time point, showing that both cell types were significantly reduced in number as compared to those treated with MTKIs and control miRNA mimic.

Figure 4.

Combination of miR-302a and miR-520 b expression with multityrosine kinase inhibitors decreases GBM cell proliferation. After transfection with miR-302a, miR-520 b or control miRNA (cel-miR-239 b) mimics for 48 h, U87 (A, C) and DBTRG (B, D) cells were exposed to 15 μM sunitinib (A, B) or 5 μM axitinib (C, D) over 72 h. Cell proliferation was assessed by sulforhodamine B (SRB) assay to measure protein content at 48, 72, 96 and 120 h after transfection.*P < 0.05, **P < 0.01, ***P < 0.001 compared to the control miRNA transfection. Representative phase contrast images of U87 (E) and DBTRG (F) cultures were taken 120 h after transfection.

MiR-302a and MTKIs impact on U87 and DBTRG GBM cell DNA content

To better understand the antiproliferative effect of the multimodal strategy, we monitored the cell cycle progression of U87 and DBTRG GBM cells through flow cytometry, by determining the DNA content of the cells following miRNA modulation and MTKI treatment (Fig. 5). A sustained, although not significant, increase in DNA content was induced by cell transfection with miR-302a mimic, which translated into an increase of the percentage of cells in the S (2 < N < 4) and G2/M (4 N) phases and of aneuploid (>4 N) cells. This finding is consistent with previous reports showing that miR-302a promotes an increase in the percentage of cells in S-phase with a decrease in the G₀/G₁ population (40). However, as shown in Figure 5, the increase of DNA content corresponding to S-phase did not translate into increased cell proliferation. Importantly, a dramatic increase in DNA content corresponding to G₂/M and/or aneuploidy (N ≥ 4) was observed with both sunitinib and axitinib in U87 cells (Fig. 5C), and with axitinib in DBTRG cells (Fig. 5D). This was further exacerbated with miR-302a modulation when combined with sunitinib in U87 cells (65.07 ± 10.99) and with axitinib in DBTRG cells (85.42 ± 7.06), as compared to the transfection with the control miRNA mimic (57.57 ± 12.12 and 70.21 ± 11.70, respectively). Of note, the same pattern was obtained from parallel experiments performed with miR-520 b (data not shown). Interestingly, sunitinib promoted completely different alterations in the DNA content of DBTRG and U87 GBM cells. While DBTRG cells were apparently arrested in G₀/G₁, U87 cells showed a dramatic increase in DNA content (N ≥ 4), corresponding either to a G₂/M arrest and/or to an increased percentage of aneuploidic cells.

Figure 5.

Cellular DNA content increases upon miR-302a transfection and treatment with multitargeted tyrosine kinase inhibitors. Cells were incubated with miR-302a or control miRNA (cel-miR-239 b) mimics for 48 h, and subsequently treated with sunitinib (15 μM) or axitinib (5 μM) for 24 h and then fixed/permeabilized with 70% ethanol before being labelled with PI/RNAse solution. Representative DNA histograms of U87 (A) and DBTRG (B) cells and the percentages of different cellular DNA contents for U87 (C) and DBTRG cells (D) are shown. DNA content can be correlated to different phases of the cell cycle: 2 N: diploid G₀/G₁; 2 < N < 4: diploid S phase; 4 N: diploid G₂/M and/or tetraploid G₀/G₁; >4 N: tetraploid S phase and G2/M.

Expression of miR-302a and miR-520 b enhances human GBM cell death and caspase activity when combined with sunitinib or axitinib

Since the multimodal strategy consistently showed to decrease GBM cell viability and proliferation, we investigated whether apoptosis could be implicated in those events. As shown in Figure 6A and C, a dramatic increase in the percentage of early apoptotic and dead U87 cells was observed upon incubation with miR-302a and miR-520 b mimics for 48 h and subsequent exposure, for 24 h, to axitinib, but not to temozolomide, as compared to that found for the combination of the control miRNA mimic with axitinib or for the miRNA modulation per se (∼2.4-fold increase, P < 0.001). On the other hand, results from parallel experiments performed in DBTRG cells showed a moderate increase in cell death when cells were exposed to axitinib (∼1.2-fold increase, p > 0.05), while, similarly to U87 cells, no effect was observed when temozolomide was included in the combined therapy (Fig. 6B and D). As sunitinib is a fluorescently active molecule that greatly interferes with cytometric analysis without cell permeabilization such as the detection of FITC-labelled annexin V and PI, U87 and DBTRG cells exposed to sunitinib per se or in combination with miRNA modulation were assayed for caspase activation. As shown in Figure 6E and F, incubation of U87 and DBTRG cells with miR-302a or miR-520 b mimics, followed by exposure to sunitinib resulted in increased caspase specific activity as compared to cells transfected with the control miRNA mimic. As shown (Fig. 6E), a 3.1-fold increase in caspase activity was observed in U87 cells when miR-520 b modulation was included in the combined therapy, while that obtained upon miR-302a modulation prior to drug treatment was approximately 1.8-fold. This same increase (1.8-fold) was observed in DBTRG cells after sequential treatment with miR-302a or miR-520 b mimics and sunitinib (Fig. 6F).

Figure 6.

Combination of miR-302a/520 b transfection with axitinib leads to GBM cell death. Following transfection with miR-302a, miR-520 or control miRNA (cel-miR-239 b) mimics for 48 h, cells were incubated with sunitinib (15 μM), axitinib (5 μM) or temozolomide (200 μM) for 24 h. Cells were then stained with propidium iodide (PI) and FITC-annexin V for cell death analysis or assayed for caspase activity. Representative plots of PI/FITC-annexin V staining are shown in (A) and (B) for U87 and DBTRG, respectively. Percentages corresponding to FITC- annexin V+/PI- cells (early apoptotic) and to FITC-annexin V+/PI+ cells (late apoptotic/dead) are presented as means ± standard deviation for U87 (C, n = 3) and DBTRG (D, n = 2). **P < 0.01, ***P < 0.001, compared to the respective control miRNA mimic transfection. ^#P < 0.05, ^##P < 0.01, relative to transfection per se. Caspase specific activity in U87 (E) and DBTRG (F) cells.

Discussion

It has become increasingly clear that GBM cells possess an intricate array of signalling pathways, whose functional redundancy and rewiring underpin the lack of success of GBM-addressed molecular targeted therapies (3–8). In addition to that, and similarly to other cancer cells, GBM cells are physiologically equipped with mechanisms able to circumvent the blockade of cellular signalling promoted by targeted strategies. A classic example of intrinsic resistance in GBM involves the epidermal growth factor receptor (EGFR) and platelet derived growth factor receptor beta (PDGFRB). Upon targeted inhibition of EGFR, PDGFRB is transcriptionally unrepressed and thus provides alternative RTK signalling, leading to drug resistance (41). Another report showed that the inhibition of mTOR (a signalling mediator downstream of PI3K/AKT) resulted in exacerbated MAPK signalling (42). These evidences point to the need of developing strategies able to suppress signalling redundancy, in order to strongly deprive cells from their growth and survival stimuli. In fact, such strategies have been proven successful as reviewed by Sun and Bernards (15). In this context, new generation drugs, such as MTKIs, have been applied with success in antitumoural therapies due to their capacity to simultaneously target several RTKs (9). However, this horizontal blockade of signal transduction has failed to significantly increase patient survival in clinical trials for GBM (10,11,43), except for one clinical trial using axitinib in patients with recurrent GBM (13). Importantly, despite minimal benefits for GBM patients were obtained with alterations in the standard adjuvant TMZ regimen (44,45), it has been proposed that MTKI dosage and administration mode can have high impact on the efficiency of cancer treatment (46,47). We hypothesize that the network rewiring capacity is impaired in cells that are abruptly subjected to intense signalling blockade, which could explain the efficiency of high-dose-based strategies. Of note, despite failing in several studies for GBM (10,11), sunitinib is now being tested in phase II/III clinical trials in a treatment regimen that takes this concern into consideration (12).

MiRNAs, implicated in virtually all physiological and pathological processes, including human malignancies, are attractive therapeutic molecular targets due to their capacity to regulate multiple molecules, and, thus, affect a myriad of signalling transducers (26). In this regard, the present study aimed to increase the expression of specific miRNAs to promote a vertical inhibition of signalling cascade intermediates, which in combination with MTKIs could abrogate pathway redundancy and result in substantial tumour cell death. A high-throughput screening analysis was performed, which allowed to identify a set of miRNAs whose modulation sensitized U87 GBM cells to sunitinib (Fig. 1 and Supplementary Material, Fig. S1). Because 8 of the top 20 miRNAs belong to the broadly conserved miR-302/520 family and share the same seed sequence, we selected two representative miRNAs of this family (hsa-miR-302a-3p and hsa-miR-520 b) as potential targets of a GBM therapy combining miRNA modulation with MTKI chemotherapy. MiR-302a belongs to the miR-302-367 cluster, which is specifically expressed in human embryonic stem cells and has been shown to convert human somatic cells into induced pluripotent stem cells (48). However, in GBM, the upregulation of this cluster was shown to induce the differentiation of GBM stem-like cells (49). Specifically, miR-302a impacted on the proliferation and migration of U87 and U251 GBM cells (35). Although much less studied, miR-520 b was shown to be detrimental towards GBM cell lines in a recent report (37).

In this work, miR-302a and miR-520 b modulation in combination with sunitinib or axitinib decreased U87 and DBTRG GBM cell viability (Fig. 3), while such treatments induced apoptosis only in U87 GBM cells (Fig. 6). However, we also obtained evidence showing that, in comparison to the transfection with a control miRNA, miR-302a mimic transfection significantly impacts on DBTRG cell motility, as assessed by the wound healing assay (Supplementary Material, Fig. S3), while the alterations in U87 cell motility were minimal (data not shown). The distinct responses of the two cell lines to the combined strategy might be explained by phenotypical differences, namely regarding the growth rate, which is 2-fold higher in U87 as compared to the DBTRG growth rate (data not shown). Importantly, the combination of miRNA modulation with the first-line treatment (temozolomide) for GBM did not result in significant cytotoxicity, which highlights both the lack of efficacy of TMZ and the need for horizontal blockade of signalling pathways, which can be achieved with the MTKIs.

We also found that both miRNAs increased the percentage of cells in the S-phase (Fig. 5), which is consistent with previously reported data (40), and with our own observations (data not shown), revealing that these miRNAs downregulate CDKN1A (a negative regulator of the G1/S checkpoint). In combination with axitinib, the miRNA modulation further promoted an increase in cell DNA content corresponding to G₂/M arrest and/or aneuploidy (N ≥ 4). However, the increase in S-phase cell DNA content was not accompanied by a higher proliferation rate but rather sustained by the multimodal strategy. Therefore, genomic instability promoted by miR-302a and miR-520 b and further exacerbated by the MTKIs might be suggested. Being primarily associated with an oncogenic event, higher levels of genomic instability can be in fact detrimental to cancer cells (50).

Additionally, the ectopic expression of both miRNAs decreased mRNA levels of the signalling cascade molecules AKT1, PIK3CA and SOS1 (Fig. 2). While PIK3CA protein levels were decreased upon miR-302a and miR-520 b modulation in U87 and DBTRG cells, the AKT activation was only decreased in DBTRG cells, as assessed through the pAKT to AKT ratio. This might underpin the increased cytotoxicity observed upon miRNA modulation per se in DBTRG cells, which was not statistically relevant for U87 cells (Fig. 3), although, in general, for both cell types, the maximal cytotoxicity has been achieved upon the combined treatment (miRNA modulation plus MTKIs). As previously described (51), miR-302/367 cluster represses the PI3K/AKT signal transduction, which was shown to sensitize lung and breast cancer cells to MTKIs (reviewed in 52). Since the resistance mechanisms to PI3K/AKT inhibition involve an exacerbated/rewired RTK signalling (53), the RTK inhibition promoted by MTKIs would abrogate pathway redundancy. The observed downregulation of the signal transducers AKT1, PIK3CA and SOS1, induced by miR-302a and miR-520 b modulation, might have decreased signalling plasticity in U87 and DBTRG GBM cells, thus rendering them more susceptible to sunitinib and axitinib.

Overall, the results presented in this study unravel a series of miRNAs to be modulated in therapy combination with MTKIs to overcome GBM cell chemoresistance. This approach holds promise and deserves to be further explored towards successful treatment of highly heterogeneous brain tumours as GBM.

Materials and Methods

Materials

Temozolomide and the multitargeted tyrosine kinase inhibitors sunitinib malate and axitinib were acquired from Selleckchem (Houston, USA) and stored at − 20 °C in DMSO. The miRIDIAN microRNA mimics of hsa-miR-302a-3p, hsa-miR-520 b and cel-miR-239 b (microRNA mimic negative control #2) were obtained from Dharmacon (Lafayette, CO, USA). Primary antibodies for Western blot analysis were acquired from Cell Signaling Technology (Boston, USA). Corning multiwell plates and cell culture flasks were obtained from Sigma (St. Louis, USA).

Cell lines and culturing conditions

U87 (human glioblastoma) and DBTRG-05MG (DBTRG, human recurrent glioblastoma) cells were kindly donated by Dr. Peter Canoll (Columbia University, New York, USA) and Dr. Massimiliano Salerno (Siena Biotech, Italy), respectively. U87 cells were maintained in DMEM (Sigma) supplemented with 10% heat-inactivated FBS (Gibco, Paisley, Scotland), 100 U/ml penicillin (Sigma), 100 µg/ml streptomycin (Sigma), 12 mmol/l sodium bicarbonate and 10 mmol/l HEPES. DBTRG cells were maintained in RPMI 1640 (Sigma) supplemented with 10% heat-inactivated FBS, 100 U/ml penicillin, 100 µg/ml streptomycin, and 12 mmol/l sodium bicarbonate. Cells were cultured in adherent conditions at 37 °C under a humidified atmosphere containing 5% CO₂. For all experiments, U87 and DBTRG cells were seeded in multiwell plates at the densities of 1.8x10⁴ cells/ml and 1.6x10⁴ cells/ml, respectively.

Cell transfection

Before use, miRNA mimics (resuspended to 20 μM in sterile nuclease free water) were freshly diluted to 500 nM with a buffer solution (6 mM HEPES pH 7.5, 60 mM KCl, 0.2 mM MgCl₂). U87 and DBTRG cells were transfected with miRNA mimics using a standard reverse transfection protocol, at a final miRNA concentration of 50 nM. Briefly, the Lipofectamine RNAiMAX transfection reagent (Invitrogen, Waltham, USA) was diluted in OptiMEM (Gibco) and added to the miRNA mimics on the multiwell plates in order to obtain the final Lipofectamine RNAimax: miRNA ratio of 0.04 μl/pmol; 30 min after, cells were added to the plates. The miRNA high-throughput screening was performed with the library of miRNA mimics (miRIDIAN miRNA mimics, Dharmacon) corresponding to all the human mature miRNAs from miRBase 13.0 (988 miRNAs, 875 unique sequences). MiRNA mimics were transferred robotically from stock library plates to 384-well plates (Culturplate-384 white plates, PerkinElmer).

Quantification of miRNA and mRNA expression by qPCR

RNA was extracted using miRCURY RNA isolation Kit (Exiqon, Skelstedet, Denmark). MicroRNA and mRNA expression levels were quantified by two step real-time PCR (RT-PCR). The reverse transcription was performed with NZY First-Strand cDNA Synthesis Kit (NZYtech, Lisbon, Portugal) and Universal cDNA Synthesis Kit (Exiqon) for mRNA and miRNA, respectively. Quantitative RT-PCR (qPCR) was performed in the StepOne Plus thermocycler (Applied Biosystems). For mRNA quantification, the NZY qPCR Green Master Mix (2x), ROX plus (NZYtech) was used. Primers for HPRT and AKT1 were designed with Primer-BLAST from NCBI and those of PIK3CA and SOS1 were retrieved from the PrimerBank (54) (primer sequences are listed in Supplementary Material, Table S2). MiRNA quantification was performed with SYBR Green Master Mix (Exiqon) using the miRCURY LNA primers from Exiqon. HPRT and hsa-miRNA-23a were used as internal controls to determine the relative expression of mRNA and miRNA, respectively. Relative mRNA and miRNA levels were determined following the Pfaffl method (55) in the presence of target and reference genes with different amplification efficiencies obtained from a standard curve. All protocols from purchased kits followed the manufacturer’s instructions.

Western blot analysis

Total protein extracts were prepared from pellets of cultured U87 and DBTRG cells and homogenized at 4 °C in radio immunoprecipitation assay lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 50 mM EDTA, 0.5% sodium deoxycholate, 1% Triton X-100) containing a protease inhibitor cocktail (Sigma), 2 mM dithiothreitol and 0.1 mM phenylmethylsulfonyl fluoride. The concentration of protein lysates was determined using the Bio-Rad Dc protein assay (Bio-Rad, Hercules, USA), and 10–20 μg total protein were resuspended in loading buffer (20% glycerol, 10% SDS, 0.1% bromophenol blue), incubated for 5 min at 95 °C and loaded onto a 7.5% polyacrylamide gel for electrophoretic separation. Proteins were blotted onto a polyvinylidene fluoride membrane. After blocking in 5% non-fat milk for 1 h at room temperature the membrane was incubated, overnight at 4 °C, with anti-PIK3CA (#4249, 1: 1000), anti-AKT (#9272, 1: 1000), anti-pAKT (Ser473) (#9271, 1: 2000), anti-ERK1/2 (#9102, 1: 1000), anti-pERK1/2 (Thr202/Tyr204) (#9101, 1: 2000), antibodies and with the appropriate alkaline phosphatase labelled-secondary antibodies (1: 10000) (Amersham Biosciences, Uppsala, Sweden) for 2 h at room temperature. Equal protein loading was verified by reprobing the membrane with an anti-α-tubulin antibody (1: 10000) (Sigma) and with the secondary antibody. Following anti-pAKT and anti-pERK probing, the amount of total AKT and ERK was assessed after a stripping protocol encompassing a 10 min incubation of the membranes with 100% methanol, which was after washed with TBS-T (Sigma) and further incubated under agitation at 50 °C with a solution containing 100 mM β-mercaptoethanol, 7 mM Tris (pH 6.8) and 2% SDS. Following the striping protocol, the blots were washed with TBS-T and blocked in 5% non-fat milk for 1 h at room temperature. After antibody incubation, the membranes were washed several times with TBS-T, incubated with the enzyme substrate ECF (Amersham Biosciences) for 5 min at room temperature and, subsequently, subjected to fluorescence detection using a ChemiDoc Touch System (Bio-Rad). The analysis of band intensity was performed using the Quantity One software (Bio-Rad).

Cell viability

Cell viability was evaluated by a modified Alamar blue assay (56). Briefly, 48 h after miRNA mimic transfection, cells were incubated in the appropriate culture media with 15 μM sunitinib, 5 μM axitinib or 200 μM temozolomide for 24 h, after which 10% (v/v) resazurin was added to culture media. After 1 h of incubation at 37 °C, the absorbance of the medium was measured at 570 nm (reduced form) and 600 nm (oxidized form) in a microplate reader (SpectraMaxPlus 384, Molecular Devices, Sunnyvale, USA). Cell viability was calculated as percentage of control cells using the equation: [(A_570nm–A_600nm) _{treated cells}/(A_570nm–A_600nm) _{control cells}] x 100.

Cell death and caspase 3/7 activity

The detection of cell death was performed by flow cytometry using the probes FITC-Annexin V (Immunostep, Salamanca, Spain) and propidium iodide (PI, Sigma). Briefly, 72 h after cell transfection with the miRNA mimics, drugs were added to the cells and following 24 h incubation, cell media and detached cells were harvested, washed with ice-cold PBS and resuspended in 100 μl of binding buffer containing 10 mM HEPES (pH 7.4), 140 mM NaCl, 2.5 mM CaCl₂, to which 2 µl of FITC annexin V (0.05 mg/mL) and 1 µl of PI (0.05 mg/ml) were added. Samples were analysed in a FACSCalibur flow cytometer after 15 min of incubation in the dark at room temperature. FITC fluorescence was evaluated in the FL-2 channel, propidium iodide was evaluated in the FL-3 and a minimum of 10,000 events were collected. The data were analyzed with FlowJo software (FlowJo LCC, Ashland, USA). Caspase 3/7 activity was assessed using the Ac-DEVD-AFC substrate (Sigma). Briefly, at the same time point used for cell death assessment, cell media and detached cells were harvested, washed with ice cold PBS and stored at − 80 °C. Cells were lysed in 30–60 µl of lysis buffer containing 10 mM Tris (pH 7.5), 100 mM NaCl, 1 mM EDTA, 0.01% Triton X-100, followed by a freeze/thaw cycle in liquid nitrogen. After 30 min on ice, cell lysates were centrifuged at 10, 000 g for 10 min at 4 °C. The supernatant was recovered and protein concentration was determined using the DC Protein Assay (Bio-Rad). In an opaque 96-well plate, 7.5–15 µg of protein lysate were mixed with 50 µl reaction buffer containing 50 mM HEPES (pH 7.5), 0.2% CHAPS, 20% (m/v) sucrose and 20 mM DTT (freshly added), to which 20 µl of 10 mM caspase substrate per millilitre were added. The final reaction volume was adjusted to 100 µl with lysis buffer. The production of the AFC fluorophore, released as a result of caspase activity, was measured for a period of 4 h in 5 min intervals in a microplate reader (SpectraMax Plus 384, Molecular Devices) at excitation/emission of 390/495 nm. Results were presented as specific caspase activity (relative fluorescence units per minute normalized for the protein amount in each well).

DNA content analysis

Forty-eight hours after transfection and 24 h following drug treatment, cells were collected and fixed in 200 µl of 70% ethanol for 2–3 h at 4 °C. After being washed with 2 mL of 2% BSA in ice-cold PBS, cells were resuspended in 250 µl PI/RNase solution (Immunostep) and analyzed in a FACSCalibur flow cytometer following a 15 min incubation in the dark at room temperature. Data were analyzed with FCS Express 6 Plus Research Edition (De Novo Software, Glendale, USA) and the different cellular DNA contents (2 N: diploid G₀/G₁; 2 < N < 4: diploid S phase; 4 N: diploid G₂/M and/or tetraploid G₀/G₁; >4 N: tetraploid S phase and G₂/M) were presented as percentage of the total DNA content.

Cell proliferation assay

Cell proliferation was measured in terms of the cellular protein content using the sulforhodamine B (SRB) assay, as previously described (57). Briefly, cells plated onto 96-well plates were fixed by adding 100 µl cold 10% (m/v) trichloroacetic acid to each well, followed by incubation at 4 °C. The solution was then aspirated and the wells were dried before the addition of 150 µl 0.05% (w/v) SRB solution containing 1% acetic acid. After 1 h incubation at 37 °C, the unbound dye was removed by washing with a solution of 1% acetic acid in water. Finally, after drying the wells, 200 µl of 10 mM Tris base (pH 10) were added to each well followed by 30 min incubation at room temperature under gentle agitation to solubilize the bound protein stain. The optical density was read at 540 nm in a microplate reader (SpectraMaxPlus 384, Molecular Devices). Protein content was calculated as percentage relative to that of control cells (non-transfected cells) at the first time point (48 h after transfection), taken as 100%.

Statistical analysis

All data are presented as means ± standard deviation of at least three independent experiments, unless stated otherwise. One-way analysis of variance combined with the Tukey posthoc test was used for multiple comparisons in cell culture experiments (unless stated otherwise) and considered significant when P < 0.05. For experiments involving the use of more than one drug, statistical analysis was performed in order to compare cells without drug with cells in the presence of each drug. Statistical differences are presented at probability levels of P < 0.05, P < 0.01 and P < 0.001. Calculations were performed with standard statistical software (Prism 6, GraphPad, San Diego, USA).

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

The authors would like to thank Dr. Isabel Nunes and Dr. Cândida Mendes (Center for Neuroscience and Cell Biology), and Dr. Alexandre Salvador (Enzifarma) for the assistance with the flow cytometry. We also thank Prof. Luis Almeida, Prof. Ramiro Almeida and Prof. Carlos Duarte (Center for Neuroscience and Cell Biology) for supplying some of the antibodies used for Western blot detection.

Conflict of Interest statement. None declared.

Funding

This work was supported by the European Regional Development Fund (ERDF), through the Centro 2020 Regional Operational Programme under the project CENTRO-01-0145-FEDER-000008: BrainHealth 2020, and through the COMPETE 2020 - Operational Programme for Competitiveness and Internationalisation and Portuguese national funds via FCT – Fundação para a Ciência e a Tecnologia, I.P., under projects POCI-01-0145-FEDER-016390: CANCEL STEM and POCI-01-0145-FEDER-007440. This work was also supported by the Italian Ministry of Education and the FCT Investigator Programme [IF/00694/2013 to M.M]. A.M.C. and A.L.C. are recipients of fellowships from the FCT with references SFRH/BPD/99613/2014 and SFRH/BPD/108312/2015, respectively.