MiR-144 overexpression as a promising therapeutic strategy to overcome glioblastoma cell invasiveness and resistance to chemotherapy

Abstract

Glioblastoma (GB) is the most aggressive and common form of primary brain tumor, characterized by fast proliferation, high invasion, and resistance to current standard treatment. The average survival rate post-diagnosis is only of 14.6 months, despite the aggressive standard post-surgery treatment approaches of radiotherapy concomitant with chemotherapy with temozolomide. Altered cell metabolism has been identified as an emerging cancer hallmark, including in GB, thus offering a new target for cancer therapies. On the other hand, abnormal expression levels of miRNAs, key regulators of multiple molecular pathways, have been correlated with pathological manifestations of cancer, such as chemoresistance, proliferation, and resistance to apoptosis. In this work, we hypothesized that gene therapy based on modulation of a miRNA with aberrant expression in GB and predicted to target crucial metabolic enzymes might impair tumor cell metabolism. We found that the increase of miR-144 levels, shown to be downregulated in U87 and DBTRG human GB cell lines, as well as in GB tumor samples, promoted the downregulation of mRNA of enzymes involved in bioenergetic pathways, with consequent alterations in cell metabolism, impairment of migratory capacity, and sensitization of DBTRG cells to a chemotherapeutic drug, the dichloroacetate (DCA). Taken together, our findings provide evidence that the miR-144 plus DCA combined therapy holds promise to overcome GB-acquired chemoresistance, therefore deserving to be explored toward its potential application as a complementary therapeutic approach to the current treatment options for this type of brain tumor.

Introduction

Glioblastoma (GB), a grade IV glioma, is the most malignant and aggressive type of brain cancer, for which no satisfactory therapy exists. Currently, the standard treatment strategy for GB consists of maximum surgical resection of tumor tissue followed by radiotherapy with concomitant or adjuvant chemotherapy with temozolomide (TMZ), along with symptomatic treatment (1). However, 50% of the patients with GB have a median survival time of only 12 to 15 months following diagnosis, and survival for the first year after diagnosis is 35%, decreasing to 13.7% in the second year and thereafter (2). New and effective therapeutic alternatives are, thus, needed to overcome GB malignancy and mortality rates.

GB cells are characterized by high levels of proliferation through sustained proliferative signaling, evasion from growth suppressors and high resistance to apoptosis. Malignant cells also acquire the ability to induce angiogenesis and promote the invasion/metastasis process. Cellular energy metabolism reprogramming was proposed as a new cancer hallmark (3). Signaling pathways altered in cancer cells converge to shift the cell metabolism, in order to respond to demanding energy requirements and promote cell proliferation by potentiating biosynthesis of macromolecules, rapidly generating ATP and conserving cell redox status (4). The metabolic switch to aerobic glycolysis in proliferating tumor cells results in the production of glycolytic intermediates, which fuel the biosynthesis of macromolecules, such as nucleotides, amino acids and lipids, required to generate new cells. The upregulation of glycolysis in malignant cells does not necessarily correspond to decreased or deficient mitochondrial function, as initially proposed by Warburg (5,6). In fact, tumor cells often have increased mitochondrial respiration (7).

The glycolytic flux is controlled at the level of the conversion of fructose-6-phosphate to fructose-1,6-bisphosphate by phosphofructokinase (PFK), the first irreversible reaction committed exclusively to glycolysis. The activity of PFK is stimulated by fructose-2,6-bisphosphate (thus promoting glycolysis) and inhibited by ATP, which switches off glycolysis. The glycolytic activator fructose-2,6-biphosphate is inhibited by dephosphorylation mediated by the TP53-induced glycolysis and apoptosis regulator (TIGAR), thus blocking the glycolytic flux (Fig. 1) (8,9). TIGAR, however, was described to promote or inhibit tumor growth, depending on the cellular context (3,10–13). In gliomas, TIGAR overexpression has a tumor-supporting role, by allowing increased mitochondrial respiration, hence maintaining energy levels (11).

Figure 1

Schematic representation of glycolysis and tricarboxylic acid (TCA) cycle, highlighting the enzymes whose mRNAs are targeted by miR-144, as well as the mechanism of action of the chemotherapeutic drugs dichloroacetate (DCA) and TMZ (see text for details).

On the other hand, pyruvate dehydrogenase kinases (PDKs), which are overexpressed in tumor cells under hypoxic conditions, inhibit pyruvate dehydrogenase complex (PDH), thus impairing the decarboxylation of pyruvate, one of the main contributors to the tricarboxylic acid (TCA) cycle, in the form of acetyl-coA. Thus, PDKs decrease the oxidation of pyruvate in the mitochondria and contribute to the increased reduction of pyruvate into lactate in the cytosol (14) (Fig. 1).

Isocitrate dehydrogenases (IDHs), such as IDH2, which was found to be overexpressed in several types of cancer, including GB, participate in the reductive metabolism of glutamine originating citrate that participates in the synthesis of fatty acids and sterols, contributing to tumor growth (15). Both IDH2, which exists in the mitochondrial matrix, and its isozyme IDH1, which locates to cytoplasm and peroxisomes, support redox homeostasis through the maintenance of NADPH levels in tumor cells (16).

MiRNAs have been implicated in diverse types of human malignancies, including GB pathogenesis. Dysregulation of miRNA expression levels has been observed in virtually all tumors, being responsible for tumor initiation, progression and metastasis formation (17–19). MiRNAs can, thus, constitute key molecular targets for cancer gene therapy. In the present study, miRNA-144 (miR-144), which predictably targets PDK1, TIGAR, IDH1 and IDH2 (Fig. 1; Supplementary Material, Table S1 and Fig. 1), was found to be under-regulated in two human GB cell lines, U87, derived from a primary astrocytoma, and DBTRG, obtained from a relapsed GB patient, as well as in human GB tumor samples. Therapeutic modulation of miR-144 activity in U87 and DBTRG cells resulted in decreased expression of its targets, impaired energetic metabolism and cell migration ability. In addition, exposure of miR-144 overexpressing DBTRG cells to dichloroacetate (DCA), an inhibitor of PDK, and, consequently, an activator of PDH, which was found to induce apoptosis and impair GB tumor growth (20), while sparing the surrounding oxidative healthy organs (21), rendered the cells highly susceptible to this drug. This effect was not observed with TMZ, the first-line chemotherapeutic agent used to treat GB, which acts as a DNA alkylating agent promoting tumor cell death (Fig. 1).

Overall, our results indicate that this bimodal gene and chemotherapeutic approach holds promise to circumvent tumor-acquired chemoresistance, tumor fast growth rate and invasion ability, features that confer extremely high malignancy to GB.

Results

Analysis of miRNA/mRNA expression in human GB cells and tissue

The levels of six miRNAs, previously shown to be involved in GB malignant phenotype (22–26) and predicted, from bioinformatic databases (miRWalk, MicroT4, miRanda, RNA22 and RNAhybrid), to target mRNAs of proteins implicated in energy metabolism processes, were quantified in U87 and DBTRG human GB cells and compared with those in normal human astrocytes (HA). Fig. 2A shows that miR-19a, miR-183, miR-155, miR-200c and miR-144 are downregulated in both cell lines, as compared to normal HA. Expression of miR-144 was not detected in DBTRG cells and was extremely low in U87 cells. On the other hand, miR-23a was overexpressed in both U87 and DBTRG cell lines. Due to its extremely low expression in the two GB cell lines tested, miR-144 was selected for subsequent functional studies and evaluation of its potential as a molecular target for therapeutic intervention. The expression profile of genes coding for proteins involved in energy metabolism processes and whose mRNAs are predicted targets of miR-144 were determined in U87 and DBTRG cells, as compared to normal HA. In this regard, the mRNA levels of IDH1 and IDH2, proteins associated with mitochondrial respiration, and TIGAR and PDK1, proteins related to glycolysis and TCA cycle, were evaluated. Figure 2B shows that IDH1 mRNA levels are unchanged in both cell lines, whereas IDH2 is downregulated. PDK1 is downregulated in U87 cells but is upregulated in DBTRG cells, and TIGAR levels are increased in U87 cells and unchanged in DBTRG cells. To ascertain the physiological significance of the results observed for the two GB cellular models in the human disease, the levels of miR-144 and of its putative target mRNAs were evaluated in 24 (frozen) human GB samples obtained from the Tumor Bank of the Centro Hospitalar e Universitário de Coimbra. As observed, the levels of miR-144 were decreased in 19/24 tumor samples, as compared to normal HA (Fig. 2C). The miR-144 target genes IDH1, IDH2 and PDK1 were found to be overexpressed in 24/24 tumor samples, while TIGAR was overexpressed in 17/24 tumor samples, downregulated in 2/24 samples (Fig. 2D) and undetectable in the remaining 5/24 samples, with respect to normal HA.

Figure 2

Expression levels of miRNAs (A and C) and mRNA (B and D) in U87 (black bars) and DBTRG (gray bars) cell lines (A and B) and in GB tumor samples (C and D), with respect to normal HA. RNA extracted from normal HA, U87 and DBTRG cells, 24 h after plating and from 24 frozen tumor samples was transcribed using Exiqon (miRNAs) or NZYTech (mRNA) cDNA synthesis kits, and qRT-PCR was performed with adequate primers from Exiqon (miRNAs) or Invitrogen (mRNA). The Pfaffl method was used to evaluate miRNA and mRNA expression levels, which were normalized to the reference U6 (miRNAs) or HPRT1 (mRNAs) and are presented as relative expression values with respect to normal HA. Data from cells are presented as mean ± SD of at least three independent experiments. Pairwise comparisons were performed between the levels of each miRNA or mRNA determined in U87 or DBTRG cells and those presented by normal HA (^*P < 0.05, ^**P < 0.01, ^***P < 0.001). Data from frozen tumor tissue are presented as mean ± SD of 24 samples. Pairwise comparisons were performed between the levels of each miRNA or mRNA determined in the tumor samples and those presented by normal HA (^*P < 0.05).

Effect of miR-144 modulation on the levels of its target mRNA

To determine the functional consequences of miR-144 overexpression, U87 and DBTRG cells were transfected with miRCURYLNA miR-144 mimics using delivery liposomal system (DLS) as delivery vector. Due to the absence of basal miR-144 expression in DBTRG cells, miRNA levels were determined with respect to those of SNORD44, a miRNA that showed high and consistent expression across all samples, thus being a suitable control for miRNA level determination. Transfection resulted in increased miR-144 expression levels in both cell lines, without unspecific effects (transfection with control ON did not result in any increase of miR-144 levels) (Fig. 3A and E). The levels of the mRNAs of IDH1, IDH2, TIGAR and PDK1 (miR-144 predicted targets) were unchanged in U87 cells transfected with miR-144 mimics (Fig. 3B) but were remarkably downregulated in transfected DBTRG cells, without showing unspecific effects (Fig. 3F). Regarding the protein levels of miR-144 targets, PDK1 was significantly downregulated both in U87 and DBTRG cells as a consequence of transfection with miR-144 mimics, whereas TIGAR levels remained unaffected in both cell lines (Fig. 3C, D and G, H).

Figure 3

Expression levels of miR-144 (A and E), of miR-144 target IDH1/2, TIGAR and PDK1 mRNA (B and F), and of miR-144 target PDK1 and TIGAR proteins (C, D, G and H) in U87 (A–D) and DBTRG cells (E–H) after transfection with miR-144 mimics. Cells were plated onto 12-well plates at a density of 6 × 10⁴ cells/well. Twenty-four hours after plating, cells were transfected during 4 h using DLS as a delivery vehicle for miR-144 mimics (gray bars in A, B and E, F) or the control ON (black bars in B and F). Forty-eight hours after transfection total RNA and protein were extracted for miRNA/mRNA and protein quantification, respectively. Fold changes in miR-144 levels were estimated with respect to Snord44 expression levels by qRT-PCR (A and E) in untransfected cells (NTC) and in cells transfected with the control ON (Control ON) and with miR-144 mimics (miR-144). The Pfaffl method was used to determine the expression levels of miRNA target genes, as compared to untransfected cells (control cells, NTC), normalized to the reference gene HPRT1 (B and F). Protein levels were determined in protein extracts obtained from cells transfected with miR-144 mimics (black bars in C and G) or the control ON (gray bars in C and G) using lysis buffer containing protease inhibitors, and analyzed by western blot, with normalization to β-actin levels. Protein levels are presented as a percentage of untransfected control cells (NTC). Representative western blot images are depicted in D and H. Data are presented as mean ± SD of three independent experiments.

Effect of miR-144 overexpression on GB cell metabolism

To evaluate the effects of miR-144 overexpression on GB cell metabolism, the cellular oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured in U87 and DBTRG cells, 48 h after miRNA modulation, using a XF24 Extracellular Flux Analyzer. As observed in Figure 4A, no alterations in the non-mitochondrial and mitochondrial basal respiration were detected in U87 cells after miRNA modulation, as deduced by considering the OCR levels after rotenone injection (for non-mitochondrial respiration) and the difference between OCR of resting cells (cultured in a medium containing glucose) and OCR after rotenone injection (for mitochondrial basal respiration). Furthermore, miR-144 overexpression did not promote alterations in the ATP-coupled respiration, deduced from the difference between OCR of resting cells and OCR after oligomycin injection, which inhibits ATP synthesis at complex V. Apparently, the permeability of the inner mitochondrial membrane to protons (proton leak) was not affected by miR-144 overexpression, since the difference between OCR after oligomycin injection and OCR after rotenone injection was not changed. The glycolytic flux, determined from the difference between ECAR of resting cells and ECAR after injection of the glycolysis inhibitor 2-DG, showed to be significantly reduced in U87 cells upon transfection with miR-144 mimics (Fig. 4B), as compared both with untransfected U87 cells and transfected with control ON. Similarly, the glycolytic capacity, measured as the ECAR after oligomycin-mediated blockade of mitochondrial ATP production, was significantly reduced after miR-144 overexpression. No alterations, however, were observed in the glycolytic reserve and non-glycolytic acidification. ATP production in U87 cells overexpressing miR-144 was similar to that of untransfected control cells, but, when U87 cells were challenged with oligomycin or with 2-DG, miR-144 induced an increase in ATP production taking ATP levels to values close to those of untreated cells (Fig. 4C). Therefore, relying exclusively on glycolysis (when oligomycin is present in the medium) or on OXPHOS (when 2-DG is present in the medium), U87 cells transfected with miR-144 mimics respond to these insults with increased energy production.

Figure 4

Effect of U87 (A–C) and DBTRG (D–F) cell transfection with miR-144 mimics (gray bars) or control ON (black bars), as compared to untransfected control cells (white bars, NTC), on OCR (A and D), ECAR (B and E) and ATP production (C and F). Cells were plated at a density of 6 × 10³ in an XF24 plate. Twenty-four hours after plating, cells were transfected for 4 h with miR-144 mimics or control ON, at a final concentration of 50 nm per well, using DLS as a delivery vehicle. Cells were allowed to grow for 48 h, and 1 h before the assay; the media were replaced with XF24 Assay Media. Oligomycin, FCCP, 2-DG and rotenone were diluted in XF24 media and loaded into the cartridge to achieve final concentrations of 1, 0.3, 600 and 1 μm, respectively. OCR and ECAR were monitored using a Seahorse Bioscience XF24 Extracellular Flux Analyzer, and data were normalized to total protein content of each well. ATP production was determined from cell lysates in a VICTOR Multilabel Plate Reader (Perkin Elmer) using a luciferase/luciferin luminescence assay and an ATP standard curve. ATP production was normalized to total protein content of each well. Data are presented as mean ± SD from three independent experiments. Pairwise data comparisons were performed for each parameter between cells transfected with miR-144 mimics and the respective untransfected control (^*P < 0.05, ^**P < 0.01, ^***P < 0.001) or control ON (^#P < 0.05, ^##P < 0.01).

In DBTRG cells, miR-144 overexpression resulted in decreased mitochondrial basal respiration in comparison with control cells, accompanied by a decrease of ATP-coupled OCR (Fig. 4D). Proton leak and non-mitochondrial respiration were unaffected by miR-144 overexpression (Fig. 4D). ECAR measurements in DBTRG cells showed no alteration in glycolytic-related parameters after miR-144 overexpression (Fig. 4E). DBTRG cells transfected with miR-144 mimics presented similar ATP production as control cells, which was stimulated by oligomycin treatment, although to a lesser extent than that observed for control cells, but incubation of transfected cells with 2-DG resulted in a significant decrease of ATP production, indicating a high dependence on glycolysis (Fig. 4F).

Effect of miRNA modulation on GB cell migration and invasion

The effect of miR-144 overexpression on the migratory ability of U87 and DBTRG cells was determined in terms of velocity (Fig. 5A and C) and accumulated travelled distance (Fig. 5B and D) over a 14 h period. U87 cells transfected with miR-144 mimics presented reduced velocity with respect to untransfected control cells, but part of this effect could be attributed to unspecific cytotoxicity due to the transfection process, since U87 cells transfected with the control ON also showed a less pronounced but still significant decrease of velocity (Fig. 5A). The distance travelled by U87 cells transfected with miR-144 mimics was significantly smaller than that travelled by untransfected cells, although a similar tendency was observed in cells transfected with the control ON (Fig. 5B). As observed, transfection of DBTRG cells with the control ON did not alter their velocity. However, miR-144 overexpression resulted in a significant decrease of DBTRG cell velocity, as compared to untransfected control cells and to cells transfected with the control ON (Fig. 5C). The distance travelled by DBTRG cells was significantly decreased by miR-144 overexpression, but transfection with the control ON showed the same tendency (Fig. 5D). The migration data indicate a mild impairment of the mobility of DBTRG cells induced by miR-144 overexpression.

Figure 5

Effect of transfection with miR-144 mimics or control ON on U87 (A and B) and DBTRG (C and D) cell migration. Twenty-four hours after plating in 12-well plates, DBTRG cells were transfected either with miR-144 mimics or the ON control. Forty-eight hours after transfection, cells were transferred into a uSlide Chemotaxis (Ibidi) for migratory capacity assessment and photographs were taken every 5 min during 14 h, using a Carl Zeiss Axio Observer Z1 microscope with a Plan-Apocromat 20×/0.8 air objective and CCD digital camera (AxiocamHRm), equipped with an incubator at 37°C and 5% CO₂. Cell velocity (A and C) and the accumulated distance travelled by the cells (B and D) were analyzed using Image J and the Chemotaxis and Migration tool from Ibidi. Data are presented as mean ± SD of at least 20 cells per experiment and from three independent experiments. Pairwise data comparisons were performed between cells transfected with miR-144 mimics and untransfected control cells (^*P < 0.05, ^**P < 0.01), between cells transfected with miR-144 mimics and cells transfected with the control ON (^#P < 0.05) and between untransfected control cells (NTC) and cells transfected with the control ON (^&P < 0.05).

To further estimate the effect of miR-144 overexpression on GB cell capacity to invade adjacent tissues, U87 cells were cultured into three-dimensional spheroid models, which were transfected with miR-144 mimics, control ON or left untreated. Forty-eight hours after transfection, spheroids were embedded into a collagen matrix and pictures were taken every day for 4 days. U87 cells formed spherical and homogeneous three-dimensional supramolecular structures, which provided a suitable platform for invasion analysis. Figure 6 depicts the invasion ability of U87 cells untreated or transfected with a control ON or miR-144 mimics. Untreated control spheroids and spheroids transfected with the control ON (transfection control) invaded the collagen matrix in a time-dependent manner and with a similar rate of invasion, as indicated by the similar area of invasion. On the opposite, spheroids transfected with miR-144 mimics presented an extremely reduced ability to invade the polymeric matrix, as assessed by the area occupied by cells detaching from the spheroid (Fig. 6A). Figure 6B shows representative pictures of U87 spheroids at Days 0 and 3 after collagen embedding, where it is apparent the reduced number of cells detached from the spheroid transfected with miR-144 mimics (at Day 3), as compared with the control conditions at the same timepoint. Furthermore, in miR-144-transfected conditions, a larger number of the cells detaching from the spheroids present a round shape as compared to the control conditions. This may indicate that miR-144 overexpression decreases cell viability or induces cell death (Fig. 6B inserts).

Figure 6

Effect of transfection with miR-144 mimics or control ON on U87 cell invasion. Three days after plating the cells into Gravity Plus Hanging Drop 96-well plates (InSphero), spheroids were collected into microcentrifuge tubes and transfected either with miR-144 mimics or the ON control. Forty-eight hours after transfection, spheroids were embedded into a collagen matrix for assessment of invasion capacity and photographs were taken every day for 4 days, using a Leica DM IL LED Fluo inverted light microscope (Leica DFC450C camera) and a 4×/0.10 objective. The area occupied by the invading cells was analyzed in untransfected cells (NTC, white bars), cells transfected with control ON (dotted bars) and cells transfected with miR-144 mimics (gray bars) using Image J software (A). Representative pictures taken immediately after spheroid embedding and 3 days after are shown in (B). Data are presented as mean ± SD from six independent experiments. Pairwise data comparisons were performed between cells transfected with miR-144 mimics and untransfected control cells, between cells transfected with miR-144 mimics and cells transfected with the control ON (^**P < 0.01, ^***P < 0.001) and between untransfected control cells and cells transfected with the control ON (ns, not significant).

Effect of miR-144 overexpression combined with chemotherapy on GB cell density

The sensitization of U87 and DBTRG cells to chemotherapeutic agents was evaluated using the sulforhodamine B (SRB) assay after overexpression of miR-144 followed by drug incubation (Fig. 7). The concentrations of 20 and 400 μm for the chemotherapeutic drugs DCA and TMZ, respectively, were found to be optimal on the basis of dose-response curves obtained with SRB assay over a range of drug concentrations after 48 h of incubation with the cells (data not shown). At such concentrations, the drugs exerted a modest cytotoxicity (ca. 20% reduction of cell viability), which we anticipated would be enhanced by previously overexpressing miR-144, and they were, therefore, selected for these studies. As observed in Fig. 7A, overexpression of miR-144 did not induce any decrease of U87 cell density, either per se or in combination with DCA or TMZ, when comparing with cells transfected with the control ON and non-transfected cells. However, in DBTRG cells, a significant decrease in cell density was achieved upon overexpression of miR-144 (27% viability reduction), in contrast to what was observed in cells transfected with the control ON. Importantly, cell transfection with miR-144 combined with DCA treatment resulted in a significant decrease of cell density (to 40%), as compared to DCA treatment or miRNA modulation per se (Fig. 7B). Such effect was not observed with TMZ, the first-line chemotherapeutic drug currently used for GB treatment (Fig. 7B).

Figure 7

Effect of DCA and TMZ, either per se or in combination with transfection with miR-144 mimics or control ON, on U87 (A) and DBTRG (B) cell density. Cells were seeded onto 96-well plates at a density of 6 × 10³ cells/well. Twenty-four hours after plating, cells were transfected for 4 h using DLS as a delivery vehicle for miR-144 mimics and control ON. Twenty-hours after transfection, cells were exposed to TMZ or DCA for 48 h. Subsequently, cells were fixed with 1% acetic acid in methanol and stored at −80°C, after which the SRB assay was performed. Pairwise data comparisons were performed between cells transfected with miR-144 mimics and the respective untransfected control (^*P < 0.05, ^***P < 0.001), between cells transfected with miR-144 mimics and the control ON (^#p < 0.05) and between the combined treatment and the miRNA modulation per se (^&P < 0.05).

Discussion

In the present work, the efficiency of a combined therapeutic strategy involving modulation of the expression levels of miR-144, predicted to target mRNAs of enzymes of the bioenergetic pathways (IDH1/2, TIGAR and PDK1), followed by incubation with chemotherapeutic drugs, was evaluated both in a primary human GB cell line (U87) and a human GB cell line obtained from a recurrent GB (DBTRG).

MiR-144 was found to be downregulated in both GB cell lines, and, importantly, also in the majority of the human GB tumor samples obtained from the Tumor Bank of the Centro Hospitalar e Universitário de Coimbra (Fig. 2A and C), which shows the universal character of miRNA downregulation in the context of GB pathogenesis. Application of bioinformatic tools that use sequence complementarity matching algorithms to predict the binding probability of miRNAs to mRNAs suggested that miR-144 targets IDH1/2, TIGAR and PDK1. The increased expression of PDK1 in DBTRG cells as compared to normal HA is most likely a consequence not only of miR-144 downregulation but also of other redundant regulatory mechanisms that are at place in these cells but absent in U87 cells, in which PDK1 was strongly downregulated. On the other hand, in U87 cells, TIGAR was found to be overexpressed, whereas in DBTRG cells its levels were unchanged with respect to normal GA (Fig. 2B). These apparently contradictory results may be due to the presence of other miRNAs expressed differently in U87 and DBTRG cells that target specifically PDK1 or TIGAR, keeping their levels under control despite the absence of miR-144. Such different miRNA expression pattern in U87 and DBTRG cell lines may also reflect tumor heterogeneity, or even the existence of different GB subtypes. This possibility is supported by the results obtained for the levels of miR-144 target mRNAs in tumor samples, which showed that TIGAR was either overexpressed (in 17/24 samples) or downregulated or even undetected (in 5/24 samples) (Fig. 2D).

Tumor cell lines are extremely useful for in vitro evaluation of therapeutic strategies, although they may not reproduce exactly the tumor gene and protein expression, since the tumoral microenvironment is lacking in these systems and the cells can undergo phenotype changes after some time in culture. In this regard, miR-144 targets that were found not to be dysregulated in GB cell lines (IDH1/2) but were overexpressed in GB tumor samples were also evaluated in the cell lines after transfection with miR-144 mimics (Fig. 3B and F). Although a target validation was not performed in this study, miR-144 was successfully overexpressed in both cell lines (Fig. 3A and E), despite the lower expression efficiency recorded for U87 cells as compared to that for DBTRG cells. This appears to be consistent with the fact that transfection of U87 cells with miR-144 mimics did not result in IDH1/2, TIGAR or PDK1 mRNA decrease. However, protein quantification showed a decrease of PDK1 levels in U87 cells overexpressing miR-144 with respect to those transfected with a control ON (Fig. 3C and D), which can reflect a miRNA mechanism of translation repression or cotranslational protein degradation (27). Thus, through miR-144 overexpression, the PDK1 was successfully downregulated in U87 cells, which could explain the decreased glycolytic activity measured in the same conditions (Fig. 4B). This would be in agreement with the reported increase of glycolysis and lactate production, as a consequence of PDK1-mediated inhibition of pyruvate dehydrogenase complex (28). However, miR-144 overexpression resulted in increased ATP production in cells treated with either oligomycin (mitochondrial respiration blockage) or 2-DG (glycolysis blockage) (Fig. 4C), which is compatible with the ability of these cells to change their energy-production route. Although no changes in the TIGAR mRNA or protein levels have been observed in U87 cells (Fig. 3B), downstream effects of miR-144 overexpression with respect to glycolysis depression suggest a non-canonical miRNA mechanism involving the derepression of TIGAR target mRNA (27,29). Target upregulation mediated by miRNAs is dependent on a number of cell-specific conditions, such as those experienced by distinct target mRNAs (30), which were not explored in the context of the present work. Regardless, the increased ATP production in U87 cells after miR-144 overexpression and treatment with oligomycin and 2-DG indicates that these cells are able to provide substrates for glycolysis downstream of TIGAR-mediated glycolysis inhibition. These intermediary molecules are produced through the pentose phosphate pathway (PPP), which is enforced by the TIGAR-mediated inhibition of PFK1 and consequent accumulation of fructose-6-phosphate that is channeled into the PPP (8). In addition, this alternative pathway provides protection from oxidative stress-induced DNA damage and apoptosis (8), which can be on the basis of the resistance these cells present to DCA, a stimulator of TCA cycle that acts through cellular overload with reactive oxygen species (31), and to TMZ, a DNA damage-inducing agent (32). In fact, by inducing the PPP, TIGAR contributes to the formation of NADPH and ribose-5-phosphate, which are essential for protection from oxidative stress and DNA repair (8).

Despite the inability of miR-144 overexpression to render U87 cells susceptible to chemotherapy, the other major GB hallmark, cell invasion, was significantly impaired by this treatment approach. Although a simple cell migration assay showed no significant impairment of mobility of U87 cells transfected with miR-144 mimics and only a significant reduction of DBTRG cell velocity (Fig. 5A–D), a more sophisticated invasion assay, performed in a three-dimensional model of U87 cells, demonstrated impaired ability of cells transfected with miR-144 mimics to invade a collagen matrix, designed to simulate the extracellular environment (Fig. 6A and B). This effect may be due to miR-144 targeting of matrix metallopeptidases MMP2 and MMP9, which are upregulated in GB and whose mRNAs were found to be downregulated as a consequence of miR-144 overexpression (data not shown) or to decreased cell proliferation or increased cell death.

Contrary to U87 cells, DBTRG cells transfected with miR-144 mimics showed decreased mRNA expression of its targets PDK1, TIGAR and IDH1/2 (Fig. 3F). Accordingly, miR-144 overexpression resulted in PDK1 downregulation. However, the predictable downstream events of this downregulation (decreased acidification of the extracellular environment consistent with decreased glycolysis and lactate production) were not observed. Rather, mitochondrial respiration appeared to be affected (Fig. 4D), which could be a consequence of IDH2 downregulation (Fig. 3F), resulting in decreased production of mitochondrial respiratory chain substrates and NADPH. In fact, these cells showed high dependence on glycolysis for energy production after miR-144 overexpression, as assessed by incubation of transfected cells with 2-DG, resulting in decreased ATP production (Fig. 4F). This is consistent with the slight (although not significant) increase of glycolytic fluxes after miR-144 overexpression (Fig. 4E). Furthermore, IDH1 downregulation mediated by miR-144 overexpression may result in decreased levels of the cofactor NADPH, which is produced by IDH1 in the cytoplasm concomitantly with the oxidation of isocitrate to α-ketogutarate and is required for glutathione reduction (33), hence sensitizing these cells to the oxidative stress-induced toxicity imposed by DCA (Fig. 7B) (31). In this regard it is worth mentioning that, in hypoxic tumors, the TCA cycle and glycolytic flux are disconnected and these tumors rewire the TCA cycle, running it in reverse through IDH1/2-mediated reductive carboxylation, to sustain proliferation (34). In addition to promote chemosensitization to DCA, miR-144 overexpression reduced DBTRG cell migratory ability, as assessed by cell tracking and trajectory determination (velocity and travelled distance) (Fig. 5C and D), which may be due to miR-144-mediated downregulation of MMP9 (data not shown). Unfortunately, DBTRG cells did not form uniform three-dimensional structures, but rather adopted irregular shapes incompatible with an efficient evaluation of collagen matrix invasion, the reason why the cell invasion assay was not performed in these cells.

Conclusion

GB is an extremely malignant type of brain tumor, challenging the therapeutic approaches that address fast tumor growth and high invasion rates. Additionally, relapsing GB tumors often develop chemoresistance. Therapeutic strategies able to reduce tumor growth may be beneficial to patients for some time, but sensitizing recurrent GB cells, characterized by high glycolytic profile, to chemotherapy is a step forward toward a successful treatment option. Furthermore, decreasing the ability of tumor cells to invade adjacent healthy brain tissue is a promising strategy to minimize the rates of tumor relapse after surgery, which currently is unable to resect all malignant cells, leaving some tumor-initiating cells embedded in the brain parenchyma that will be responsible for the re-appearance of the tumor. The delivery of small nucleic acids to GB cells in vivo may be accomplished with the use of stable nucleic acid lipid particles, functionalized with chlorotoxin for selective GB targeting, which were previously shown to effectively target GB in vivo and to mediate a significant therapeutic effect in mice brains without causing systemic toxicity (35).

Overall, this work demonstrates the potential of miR-144 overexpression to reduce GB cell malignancy, both by decreasing cell migration and invasion abilities and by sensitizing resistant tumor cells to chemotherapy, paving the way to a novel and more effective GB therapy.

Materials and Methods

Cell lines and culturing conditions

The U-87 MG (U87) human glioma cell line, kindly provided by Dr Peter Canoll (Columbia University, New York, NY), was maintained in Dulbecco modified Eagle’s medium—high glucose (DMEM-HG; Sigma, D5648), supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS) (Gibco, Paisley, Scotland), 100 U/ml penicillin (Sigma-Aldrich), 100 μg/ml streptomycin (Sigma-Aldrich), 10 mmol/l HEPES and 12 mmol/l sodium bicarbonate. The DBTRG-05 MG (DBTRG) human recurrent glioma cell line, established from a 59 year Caucasian female patient with GB treated with local brain irradiation and multidrug chemotherapy and kindly provided by Dr Massimiliano Salerno (Siena Biotech, Italy), was maintained in Roswell Park Memorial Institute 1640 (RPMI-1640) (Sigma-Aldrich, R4130), supplemented with 10% heat-inactivated FBS (Gibco, Paisley, Scotland), 100 U/ml penicillin (Sigma-Aldrich), 100 μg/ml streptomycin (Sigma-Aldrich) and 12 mmol/l sodium bicarbonate. HA were maintained in DMEM (D5648), supplemented with 2% heat-inactivated FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, 10 mmol/l sodium bicarbonate, and further freshly supplemented with 1% N2 (Sigma-Aldrich) and 1% non-essential amino acids (Sigma-Aldrich). The cells were cultured at 37°C under a humidified atmosphere containing 5% of CO₂ and grown adherent, being detached upon addition of enzyme-free dissociation buffer containing NaCl (8 g/l), Na₂HPO₄ (1.16 g/l), KH₂PO₄ (0.2 g/l), KCl (0.2 g/l) and EDTA (0.16 g/l).

Patient tumor samples

Twenty-four GB samples obtained from the Tumor Bank of the Centro Hospitalar e Universitário de Coimbra were evaluated regarding the levels of miR-144 and of its targets IDH1/2, TIGAR and PDK1. Total RNA was isolated from frozen tumor samples (~10 mg) with the AllPrep DNA/RNA/miRNA Universal kit (Qiagen) according to manufacturers’ instructions for simultaneous purification of genomic DNA and total RNA, including miRNA, from tissues.

Lipoplex preparation

MiRIDIAN hsa-miR-144-3p mimics (5′ UACAGUAUAGAUGAUGUACU 3′ from Dharmacon) and the non-targeting control oligonucleotide (control ON, 5′UUCUCCGAACGUGUCACGUdTdT 3′) were formulated into lipoplexes prepared from DLS liposomes, as previously described (36). DLS stock solution was prepared by mixing 1 mg of dioctadecylamidoglycylspermidine (Promega, Madison, WI) with 1 mg of dioleoylphosphatidylethanolamine (Sigma, Munich, Germany) and then dissolved in 40 μl of 90% ethanol, followed by a 10-fold dilution in sterile H₂O, to achieve a final lipid concentration of 5 mg/ml. The mixture was homogenized by gentle vortexing and incubated for 30 min at room temperature to allow liposome formation. DLS/ON lipoplexes were prepared freshly for every experiment. Briefly, 25 pmol of each ON were mixed with the appropriate volume of DLS suspension, previously diluted in OPTIMEM, to achieve the desired ratio of 95 μg DLS/10 μg ON. The complexes were formed over a 30 min incubation period at room temperature.

Cell transfection

For qRT-PCR, migration and western blot assays, cells were plated onto 12-well plates (Costar), at a density of 6 × 10⁴ cells/well in a final volume of 1 ml of culture medium (DMEM and RPMI supplemented with 10% serum for U87 and DBTRG cells, respectively). For viability assays, U87 and DBTRG cells were seeded at a density of 6 × 10³ and 5 × 10³ cells/well, respectively, onto 96-well plates (Costar), and the plate border wells were filled with sterile water. For cellular bioenergetics determination, cells were plated onto XF24 Cell Culture Microplates (Seahorse Bioscience) at the same density used for 96-well plates. On the next day, the cell medium was replaced with OPTIMEM (Gibco) and the lipoplexes were added to the cells, at a final concentration of 50 nm per well. After 4 h of incubation at 37°C, the transfection was stopped by replacing the OPTIMEM medium with the corresponding complete culture medium. RNA and protein were collected 48 h after cell transfection. Cell migration and cellular bioenergetics were determined 48 h after transfection.

Drug storage and cell incubation

TMZ (Temodar, Merck) was acquired from Selleckchem, and DCA was acquired from Sigma (Sigma, Germany). Stock solutions of TMZ (20 mm) and DCA (8 m) were prepared in DMSO (Sigma, Germany) and sterile water, respectively, and stored at −20°C and 4°C, respectively. Twenty-four hours after transfection with complexes carrying miRNA mimics, U87 and DBTRG cells were incubated with DCA (20 mm) or TMZ (400 μm) for 48 h at 37°C under a humidified atmosphere containing 5% CO₂.

Quantitative real time PCR (qRT-PCR)

Total RNA, including miRNAs, was extracted from cells using the miRCURY Isolation Kit (Exiqon), according to manufacturer’s instructions for cultured cells. The total RNA extracted was quantified using the NanoDrop 2000 Spectrophotometer (Thermo Scientific). Complementary DNA (cDNA) synthesis for miRNA quantification was performed using the Universal cDNA Synthesis Kit (Exiqon). cDNA was synthesized from 10 ng of total RNA in a 10 μL reaction, according to the following protocol: 60 min at 42°C, followed by reverse transcriptase heat-inactivation for 5 min at 95°C. The obtained cDNA was then diluted 1:40 (v/v) with RNase-free water and stored at −20°C. For mRNA quantification, cDNA synthesis was performed using 500 ng of total RNA in a 10 μL reaction, employing the NZY First-Strand cDNA Synthesis Kit (NZYtech, Lisbon, Portugal) and applying the following protocol: 10 min at 25°C, 30 min at 50°C and 5 min at 85°C for reverse transcriptase inactivation. Subsequently, the samples were incubated with NZY RNase H (Escherichia coli) for 20 min at 37°C in order to degrade the RNA template in cDNA:RNA hybrids. The produced cDNA was diluted 1:20 (v/v) with RNase-free water and stored at −20°C.

Quantitative PCR was performed in a StepOnePlus Thermocycler (Applied Biosystems) using 96-well microtiter plates. For miRNA quantification, the miRCURRY LNA TM Universal RT microRNA PCR system (Exiqon) was employed using SYBR Green Master Mix (Exiqon). The primers used were acquired from Exiqon (Table 1). Non-coding small nuclear RNA (snRNA) U6 (RNU6) was used as control. A master mix was prepared for each primer set, containing 1 μm each primer, FW and RV. All reactions were performed in duplicate using the Exiqon qPCR protocol (according to supplier instructions).

Table 1

Primer sequences used for miRNAs qPCR analysis and respective target sequences

MiRNA	Target sequence
hsa-miR-19a-3p	UUGCAAAUCUAUGCAAAACUGA
hsa-miR-183-5p	UAUGGCACUGGUAGAAAUUCACU
hsa-miR-155-5p	UUAAUGCUAAUCGUGAUAGGGGU
hsa-miR-23a-3p	AUCACAUUGCCAGGGAUUUCC
hsa-miR-144-3p	UACAGUAUAGAUGAUGUACU
hsa-miR-200c-3p	UAAUACUGCCGGGUAAUGAUGGA

For mRNA quantification, the SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) was used. The primers for the tested genes were designed using the bioinformatics primer designing tool Primer-BLAST, which relies on Primer3 to design PCR primers for a specific PCR target and BLAST and global alignment to select primer pairs, and purchased from Invitrogen. Hypoxanthine phosphoribosyltransferase 1 (HPRT1) was used as reference gene (Table 2). For each reaction, performed in duplicate, a final concentration of 1 μm for each primer pair was used in the Bio-Rad qPCR protocol (according to supplier instructions).

Table 2

Primer sequences used for qPCR analysis and respective amplicon sizes

Gene		Primer sequence (5′-3′)	Amplicon
IDH1	Forward	ATGGTGACGTGCAGTCGG	76 bp
	Reverse	GGACAAACCAGCACGCT
IDH2	Forward	AGACCTCATCAGGTTTGCCCA	111 bp
	Reverse	TTCACATTGCTGAGGCCGT
PDK1	Forward	GGCTATGAAAATGCTAGGCG	93 bp
	Reverse	CTGTCCTGGTGATTTTGCATT
TIGAR	Forward	CTGACTGAAACTCGCTAAGG	105 bp
	Reverse	CAGAACTAGCAGAGGAGAGA
HPRT1	Forward	CTGCGTAACTCCATCTGA	103 bp
	Reverse	ACCGTAATTGGCATCGT

The no template control (NTC) and the no reverse transcriptase control (noRT) were assessed, for each primer set, in all experiments performed. The threshold value for threshold cycle determination (Ct) was defined as 10 000 and the baselines adjusted for each sample. Fold changes of miRNAs and mRNA levels were determined according to the Pfaffl method using the levels of RNU6 or SNORD44 (miRNAs) and HPRT1 (mRNA) as internal controls, taking into consideration the different amplification efficiencies of each primer set (ranging from 80% to 120%) obtained from a standard curve generated by making serial dilutions. The amplification efficiency for each target or reference miRNA and mRNA was determined according to the formula: E = 10^–1/S – 1, where S is the slope of the obtained standard curve.

Western blot analysis

Total protein extracts were obtained using lysis buffer (50 mm NaCl, 50 mm EDTA, and 1% Triton X-100) supplemented with a protease inhibitor cocktail (Sigma), 10 μg/ml DTT, and 1 mm PMSF. Extracts in lysis buffer were subjected to three freeze–thaw cycles, and protein content was determined using the Bio-Rad DC protein assay (Bio-Rad). For each sample, 20 μg of total protein was resuspended in loading buffer (20% glycerol, 10% SDS and 0.1% bromophenol blue), incubated for 5 min at 95°C, and loaded onto a 10% polyacrylamide gel. After electrophoresis, the proteins were blotted onto a PVDF membrane according to standard protocols. After blocking in 5% nonfat milk, the membrane was incubated overnight at 4°C with the appropriate primary antibody (anti-PDK1 1:1000, anti-TIGAR 1:1000, Cell Signaling), and thereafter with the appropriate HRP secondary antibody (1:10 000, anti-rabbit, Invitrogen) for 2 h at room temperature. Equal protein loading was shown by reprobing the membrane with an anti-β-actin (1:10 000, Sigma) and with the appropriate secondary antibody. Membranes were revealed by incubation with enhanced chemiluminescence substrate for 5 min and images taken in ChemiDoc (BioRad) were analyzed with ImageLab (BioRad).

Cellular bioenergetics analysis

Mitochondrial OCR, which is a measurement of electron flux through the mitochondrial respiratory system resulting in oxygen reduction to water, and ECAR, which is an indirect measurement of the glycolytic capacity of the cells, leading to lactate production, were measured in U87 and DBTRG cells using an XF24 Extracellular Flux Analyzer. Twenty-four hours before the assay, 1 ml of XF Calibrant Solution was placed into each well of the sensor hydration microplate and the sensor cartridge placed onto the microplate and incubated overnight in a non-CO₂ incubator at 37°C. The XF Base Medium was then supplemented with 25 mm glucose (Sigma) and 4 mm l-glutamine (Sigma), for U87 cells, while for DBTRG cells the XF Base Medium was supplemented with 2 mm l-glutamine (Sigma), to achieve nutrient concentrations of normal culture medium for each cell line. Media pH was adjusted at 7.35 ± 0.05. Cells were gently washed twice with 1 ml of XF assay medium at 37°C without disturbing the cell monolayer and, then, incubated with XF assay medium at 37°C for 1 h in a non-CO₂ incubator. Solutions of oligomycin (1 μm), FCCP (0.3 μm), rotenone (1 μm) and 2-deoxy-D-glucose, 2-DG (1.2 m) were prepared in XF Base Medium. Each compound was loaded into its respective cartridge. To compensate for the dilution effect in the wells due to the successive injection of the compounds, volume adjustments were made. Several parameters were calculated from bioenergetics analysis, according to Figure 8, namely, mitochondrial respiration, taken as the difference between OCR of resting cells cultured in medium containing glucose and OCR after rotenone injection; ATP-coupled respiration, determined as the difference between OCR of cells cultured in medium containing glucose and OCR after oligomycin injection; proton leak evaluated as the difference between OCR after oligomycin injection and rotenone injection and non-mitochondrial respiration, determined as the remaining OCR after rotenone injection. Parameters determined from ECAR measurements included glycolytic flux, determined as the difference between ECAR of resting cells cultured in medium containing glucose and after 2-DG injection; glycolytic capacity, assessed as the difference between ECAR measurements after oligomycin stimulation and after 2-DG inhibition; glycolytic reserve representing the difference between ECAR of resting cells and ECAR after oligomycin stimulation; and non-glycolytic acidification, determined after 2-DG injection.

Figure 8

Schematic representation of OCR (A) and ECAR (B) recorded using the XFe Seahorse Bioanalyzer with sequential injection of oligomycin, FCCP, 2-DG and rotenone, and extrapolated bioenergetics parameters (see text for details).

OCR and ECAR readings were normalized to total protein levels in each well, which were obtained by the BioRad DC Protein Assay. Briefly, lysis buffer (50 mm NaCl, 50 mm EDTA, and 1% Triton X-100) was added to each well of XF24 Cell Culture Microplate (Seahorse Bioscience) and cells were subjected to freeze-thaw cycles. Cells were transferred to microcentrifuge tubes; the cell lysates were centrifuged for 5 min at 14000g and the supernatant was collected. The protein content of the lysates was determined according to the manufacturer’s instructions, based on a standard curve for BSA. The plate, containing known BSA concentrations, was incubated for 15 min in the dark at room temperature and the absorbance was measured at 750 nm in a microplate spectrophotometer (SpectraMax Plus 384, Molecular Devices).

For each condition of cellular energetics analysis, ATP was quantified in a VICTOR Multilabel Plate Reader (Perkin Elmer) using the software Wallac 1420. Briefly, supernatants obtained from lysates were loaded onto a white-walled 96-well plate and 20 μL of luciferase (Sigma) diluted 1:1000 in lysis buffer and 20 μL of 400 μm luciferin (Sigma) in lysis buffer, pH 8.0, were added to each well. The luminescence data obtained were normalized using a standard curve for ATP obtained from nine ATP solutions over the concentration range of 10⁻¹² to 10⁻⁴ m ATP and a blank, without ATP).

Cell migration and trajectory analysis

Cell migration experiments were performed using the μ-Slide Chemotaxis (Ibidi) as previously described (37). Twenty-four hours before the experiment, μ-Slide Chemotaxis (Ibidi) plates with plugs and cell growth media were placed in the incubator at 37°C and 5% CO₂ in a humidified sterile 10 cm Petri dish, for gas equilibration. Forty-eight hours after transfection, cells were detached from 12-well plates by incubation with dissociation buffer for 5 min at 37°C, resuspended and counted. Subsequently, 1.8 × 10⁴ cells in a final volume of 10 μL were applied onto the central compartment of a μ-Slide Chemotaxis (Ibidi) plate, following provider’s instructions. Cells were allowed to adhere to the slide for 2 h in an incubator at 37°C and 5% CO₂, and then complete cell medium was applied to each side compartment. Thereafter, cell migration was recorded for 14 h under a microscope with a Plan-Apocromat 20×/0.8 air objective and CCD digital camera (AxiocamHRm), equipped with a 37°C and 5% CO₂ incubator, and photographs were taken every 5 min. The trajectory of each cell was determined using the Image J software (v. 1.48, Wayne Rasband, NIH, USA) with the Cell Tracking plugin. Trajectories were analyzed using the Chemotaxis and Migration tool from Ibidi, where single cell trajectory was tracked by selecting the center of mass through all the time points. Cell velocity was calculated by the displacement from the initial to the end point of the total trajectory time. For each experimental condition, images were acquired in four different random regions and a minimum of 20 cells were tracked. Experiments were independently repeated three times.

Cell invasion determination

Cell invasion experiments were performed in U87 cells grown in three-dimensional spheroids, using the Gravity Plus Hanging Drop 96-well plates (InSphero) embedded into a collagen matrix, as described (38). Briefly, a total of 2.5 × 10⁴ cells in 40 μL of DMEM medium was loaded into each well of a Gravity Plus Hanging Drop 96-well plates (InSphero). After 3 days, round and homogeneous spheroids were formed. Each spheroid was transferred into a low-adhesion microcentrifuge tube containing OPTIMEM, and DLS/control ON or DLS/miR-144 lipoplexes were added to the spheroids, at a final ON concentration of 50 nm. After 4 h of incubation at 37°C, the transfection was stopped by adding complete DMEM medium. Forty-eight hours after transfection, each spheroid was collected with a p200 cut pipette tip and placed in the center of each well of a 24-well plate containing the collagen matrix. The collagen matrix was prepared by gently mixing PureCol bovine collagen type I solution (3 mg/ml; Advanced BioMatrix) with Minimum Essential Medium (10×) and NAOH 0.1 m (5×) and held on ice, to achieve a final collagen concentration of 2.2 mg/ml. Twenty-four well plates, each well being covered with 400 μl of liquid collagen, were incubated for 2 min at 37°C to allow initiation of collagen polymerization prior to spheroid embedding. Thirty minutes following spheroid embedding (after complete collagen polymerization), 400 μ of complete DMEM were added to each well. Cell invasion was monitored by digital photography using a Leica DM IL LED Fluo inverted light microscope (Leica DFC450C camera) at room temperature, with the Leica Application Suite (LAS V4.4). Images were acquired every 24 h (Day 0 = picture taken 3 h after collagen embedding) using a 4×/0.10 objective. Invasion areas were quantified using Image J software (v. 1.48, Wayne Rasband, NIH, USA).

Cell density evaluation

Cell density was evaluated through the quantification of the cellular protein content using the colorimetric SRB assay. Briefly, 72 h after transfection and 48 h following drug treatment, cells plated in 96-well plates were fixed in 1% acetic acid in methanol and stored at −80°C. After at least 1 h of freezing, the solution was removed and the cells left drying for 20 min at 37°C, followed by incubation with 0.5% SRB solution containing 1% acetic acid in water for 1 h at 37°C. Subsequently, cells were washed thoroughly with a solution of 1% acetic acid in water, to remove all the surplus SRB dye, and dried at 37°C. Protein-bound SRB was solubilized with 10 mm Tris base buffer (pH = 10). After 2 h of incubation under mild agitation, the absorbance of the resulting solution was measured at 510 nm in a SpectraMax Plus 384 (Molecular Devices). Cell density was expressed as percentage of control cells (non-treated), according to the equation

$$ Cell\ density\ \left(\% of\ control\right)=\frac{Abs_{\left( treated\ cells\right)}}{Abs_{\left( control\ cells\right)}}\times 100. $$

Statistical analysis

All data are presented as mean ± standard deviation (SD) of at least three independent experiments performed in duplicate, unless stated otherwise. One-way analysis of variance (ANOVA) combined with the Tukey post-hoc test was used for multiple comparisons (unless stated otherwise) and considered significant when P < 0.05. Statistical differences are presented at probability levels of P < 0.05, P < 0.01 and P < 0.001. Calculations were performed using standard statistical software (Prism 6, GraphPad, San Diego, CA, USA).

Acknowledgements

A.M.C. is recipient of a fellowship from the FCT with reference SFRH/BPD/99613/2014. The authors would like to thank Dr Margarida Caldeira (Center for Neuroscience and Cell Biology) for the assistance with the Seahorse analyzer, and Professor Paula Agostinho (Center for Neuroscience and Cell Biology) for kindly lending us the VICTOR luminometer.

Conflict of Interest statement. None declared.

Funding

European Regional Development Fund (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 Fundação para a Ciência e a Tecnologia (FCT), I.P. (under projects POCI-01-0145-FEDER-016390:CANCEL STEM and POCI-01-0145-FEDER-007440).