MicroRNA519d and microRNA4758 can identify gangliogliomas from dysembryoplastic neuroepithelial tumours and astrocytomas

Mutation analysis of GG and DNT

The BRAF V600E mutation was found in 16 of the 26 (61.5%) GG samples. An initial screen using sanger sequencing identified 11 GG samples that were positive for the BRAF V600E mutation. Samples that were deemed BRAF V600E negative based on sanger sequencing and of which a sufficient amount of DNA was still available were screened further using a next-generation sequencing panel. Based on this panel an additional 5 GG samples were found to be positive for BRAF V600E. One BRAF V600E positive GG also had an additional BRAF mutation (BRAF T559R) on the same allele. A total of 8 DNTs were also screened on the panel of which 1 was found positive for the BRAF V600E, 1 was found positive for FGFR1 mutations (FGFR1 G539R and FGFR1 K656E) and 1 was found positive for CIC G935R mutation of which the pathological implication is unknown.

miR-519d and miR-4758 expression in GG

First, we performed a global expression analysis of 1891 miRNAs using an Exiqon miRCURY LNA™ microRNA Array on 11 frozen brain samples (control cortex, n=5; GGs, n=6). Compared to control cortex, a total of 5 miRNAs: miR-519d, miR-4758, miR-664b, miR-4714 and miR-5681b were differentially expressed in GGs. We further validated these 5 miRNAs using Taqman miRNA assays in a larger cohort (frozen samples, control cortex, n=7; GG, n=14). Only miR-519d and miR-4758 were confirmed to be differentially expressed in GGs compared to control cortex, showing 5-fold and 13-fold upregulation (p=0.0335 and p=0.0153, respectively; Figure 1A-1B). Receiver operating characteristic (ROC) analysis for miR-519d yielded an area under the curve (AUC) of 0.796, p=0.031 and for miR-4758 AUC=0.837, p=0.0139. Furthermore, a positive correlation was observed between the two miRNAs (r=0.731, p=0.0002). We also evaluated the expression of both miRNAs in GG compared to other gliomas, including PA (n=15), diffuse astrocytoma grade II (AII; n=10), diffuse astrocytoma grade III (AIII; n=14) and glioblastoma (GB; n=15; Figure 1A-1B). miR-519d was upregulated in GG compared to all other gliomas, whereas miR-4758 was upregulated compared to AII, AIII, GB but not PA. Furthermore, no difference was found in expression of both miRNAs in BRAF V600E mutated GGs (n=8) compared to wildtype GGs (n=6; p=0.950 and p=0.414; data not shown).

miR-519d and miR-4758 expression by quantitative real-time PCR in GG, DNT and subependymal giant cell astrocytoma

Recent studies indicate a robust stability of miRNAs, supporting the accuracy of miRNA measurements with quantitative real-time PCR, also in formalin-fixed paraffin-embedded (FFPE) tissues [29–31]. Therefore, miR-519d and miR-4758 expression was also evaluated in a large cohort of FFPE samples, including control cortex (n=9), GG (n=14), peritumoural cortex (n=7), DNT (n=9) and subependymal giant cell astrocytoma (SEGA; n=10) samples. Normal appearing peritumoural tissue represents appropriate disease control tissue for GG, since it is exposed to the same seizure activity, drugs and fixation time, while also being of identical age and gender. Evaluation of miR-519d and miR-4758 expression in this cohort confirmed the upregulation of miR-519d in GG compared to control cortex (p=0.041), DNT, SEGA and peritumoural cortex, whereas miR4768 was upregulated in GG compared to control cortex (p=0.035), DNT and SEGA but not when compared to peritumoural cortex (Figure 1C-1D).

miR-519d and miR-4758 cellular distribution by in situ hybridization in GG

The cellular distribution of miR-519d and miR-4758 in peritumoural tissue (n=6) and GG (n=7) was investigated using in situ hybridization (ISH). In peritumoural tissue a weak expression of miR-519d and miR-4758 was observed in neuronal cells, but could not be detected in glial cells (Figure 2A and 2F). Expression of both miR-519d and miR-4758 was found specifically in cells with atypical astroglia morphology and in dysplastic neurons (Figure 2B-2C and 2G-2H). Double labelling confirmed miR-519d and miR-4758 expression in NeuN-positive neurons and GFAP-positive astrocytes in GG (Figure 2D-2E and 2I-2J).

miRNA target expression in GG

In order to identify potential targets of miR-519d and miR-4758 5 different databases were utilised. Targets were considered relevant if found in at least 3 of the 5 databases. Based on these target prediction tools, AKT3, CDKN1A (P21), JAK1, PTEN, ERBB3, ERBB4, RB1 and TIMP2 were identified as potential targets of miR-519d. Both clinical and experimental studies have shown the potential contribution of miR-519d dysregulation in hepatocellular carcinoma, breast cancer and cervical cancer [32–34]. AKT3, CDKN1A and PTEN have all been shown to be downregulated by miR-519d, potentially explaining the oncogenic features of this miRNA [35, 36]. Based on target prediction tools the relatively unknown miRNA, miR-4758 was predicted to target CDK2 and CDKN1B (P27), raising the possibility that miR-4758 also has oncogenic properties. Evaluation of the mRNA expression of miR-519d and miR-4758 targets (Figure 3) showed a downregulation of ERBB3 (p=0.0005), AKT3 (p=0.0015), PI3KCA (p=0.0015), RB1 (p=0.0009), PTEN (p=0.0041), TP53 (p=0.023) and CDKN1B (p=0.0065) in GG (n=14, except for ERBB4 where n=13) compared to control cortex (n=7; Mann-Withney U test). Furthermore, targets ERBB4 (p=0.032) CDKN1A (p=0.048) and JAK1 (p=0.0124) were found to be upregulated in GG compared to control cortex. The expression of TIMP2 (p=0.68), CDK4 (p=0.39), CDK2 (0.91) and MDM2 (p=0.63) did not change.

miRNA target expression in cell culture

Given the lack of GG cell lines, we used the human pediatric low grade astrocytoma cell line Res259 (LGG2) obtained from a pediatric AII [37] to investigate the effects of miR-519d and miR-4758 upregulation. After transfection with miRNA mimic, overexpression of the specific miRNA was observed under basal condition (data not shown). Quantitative real-time PCR analysis was performed to evaluate the effects of miR-519d and miR-4758 transfection on mRNA expression of a subset of targets related to the PI3K/AKT pathway, involved in the regulation of cell cycle progression [38]. CDKN1A was downregulated after miR-519d transfection, whereas transfection with miR-4758 did not affect CDKN1A expression compared to control (Figure 4A). Co-transfection of miR-519d with miR-4758 rescued the effect of miR-519d transfection. The expression of miR-4758 target CDKN1B was not affected by transfection with miR-4758 or miR-519d alone (Figure 4B). However, after co-transfection with miR-4758 and miR-519d CDKN1B was upregulated compared to control. Furthermore, PTEN mRNA expression did not change after transfection with either one of the miRNAs (Figure 4C).

Effects of miR-519d and miR-4758 modulation on the cell cycle

We further evaluated the role of miR-519d and miR-4758 on cell cycle progression using flow cytometry analysis in Res259/LGG2. Overexpression of miR-519d showed a decreased G₁ phase (13%, p<0.01) whereas, transfection with miR-4758 resulted in an increased G₂/M phase. Overexpression of miR-519d, but not miR-4758, increased the S phase; miR-4758 mimic co-transfected with miR-519d mimic was able to counteract the effect of miR-519d alone (Figure 4D).

Subjects

The majority of the cases included in this study were obtained from the archives of the Academic Medical Center of Amsterdam and the University Medical Center Utrecht. A further DNT case was included from the University College London Institute of Neurology. SEGA tissue was obtained from the following institutes: University Hospital Erlangen, Children's Memorial Health Institute in Warsaw, Hacettepe University in Ankara, and the University Hospital de Santa Maria (CHLN) in Lisbon. A total of 99 surgical tumour specimens were examined (Table 1; n=35 GNTs (WHO grade I); n= 64 astrocytomas; SEGA (TSC1/TSC2 mutated) n=10; PA (WHO I) n=15; AII (IDH-1 mutant) n=10; AIII (IDH-1 mutant) n=14 and GB (IDH-1 WT) n=15). Eight surgical specimens contained sufficient amount of peritumoural tissue (normal-appearing cortex/white matter adjacent to the tumour; Table 1). All tumour cases were reviewed independently by eight neuropathologists, and the diagnosis was confirmed according to the revised WHO classification of tumours of the central nervous system [1, 7]; only entities with consensus agreement between the neuropathologists were included (CD34 positive GGs and CD34 negative DNTs, all IDH1 negative). Additional molecular diagnosis was performed on GNT samples with sufficient amounts of DNA to identify mutations in BRAF and FGFR1 genes. Control cortical specimens were obtained at autopsy from 17 patients. For the cortical specimens attention was taken to provide equal grey/white matter tissue components. For control tissue, all autopsies were performed within 24 hours after death. Tissue was obtained and used in accordance with the Declaration of Helsinki and the AMC Research Code provided by the Medical Ethics Committee and approved by the science committee of the UMC Utrecht Biobank.

Mutation analysis

GG samples were screened using sanger sequencing as described previously [41]. Next generation sequencing (NGS) was performed on 10 GGs and 8 DNTs using a customised Ion AmpliSeq™ Neurology Panel (ThermoFisher Scientific, Waltham, Massachusetts, USA) for targeted multi-gene amplification, as previously reported [18]. This panel consists of the following genes; AKT1, AKT3, ATRX, BRAF, CDK6, CIC, CTNNB1, DDX3X, DEPDC5, EZH2, FGFR1 (exon 12 and exon 14), FUBP1, H3F3A, HIST1H3b, HIST1H3c, IDH1, IDH2, KDM6A, mTOR, MYB, MYBL1, NPRL2, NPRL3, PIK3CA, PIK3R1, PIK3R2, PTCH1, PTEN, SMARCA4, SMARCB1, SMO, SUFU, TP53. FGFR1 mutations below 5%, duplications and fusion genes cannot be detected with this panel.

Tissue preparation

Brain tissue from control and tumour patients was fixed in 10% buffered formalin and embedded in paraffin. Paraffin-embedded tissue was sectioned at 5 μm and mounted on pre-coated glass slides (Star Frost, Waldemar Knittel GmbH, Brunschweig, Germany). Sections of all specimens were processed for hematoxylin and eosin (HE), staining as well as for immunohistochemical stainings for a number of neuronal and glial markers to confirm the diagnosis. Additional tissue from controls (n=8) and from patients with GG (n=15), PA (n=15), AII (n=10), AIII (n=14) and GB (n=15) was snap frozen in liquid nitrogen and stored at -80°C until further use (RNA isolation for microRNA array and/or quantitative real-time PCR).

In situ hybridization

ISH for miR-519d and miR-4758 was performed using 5′ - 3′ double digoxygenin (DIG)-labeled Superior probes (miR-519d-3p; DIG-CacTcuAaaGggAggCacTuuG-DIG; miR-4758-3p: DIG-GagGguGguCagCagGugGggCa-DIG; Ribotask ApS, Odense, Denmark). The hybridizations were done on 5 μm sections of paraffin-embedded materials as described previously [30]. The probes were hybridized at 58°C for 1 h, and the hybridization was detected with alkaline phosphatase (AP)-labeled anti-DIG (Roche Applied Science, Basel, Switzerland). NBT (nitro-blue tetrazolium chloride)/BCIP (5-bromo-4-chloro-3′-indolyphosphate p-toluidine salt) was used as chromogenic substrate for AP. Negative control assays were performed without probes and without primary antibody (sections were blank). For double-staining, combining immunohistochemistry with ISH, the sections were first processed for ISH and then processed for immunohistochemistry with glial fibrillary acidic protein (GFAP; monoclonal mouse, Sigma, St. Louis, Mo, USA; 1:4000), and NeuN (neuronal nuclear protein; mouse clone MAB377; Chemicon, Temecula, CA, USA; 1:2000). Signal was detected using the chromogen 3-amino-9-ethylcarbazole (Sigma-Aldrich, St. Louis, MO, USA).

RNA isolation

For RNA isolation, cells or frozen tissue were homogenized in Qiazol Lysis Reagent (Qiagen Benelux, Venlo, The Netherlands). Total RNA, including the miRNA fraction, was isolated using the miRNeasy Mini kit (Qiagen Benelux, Venlo, the Netherlands) according to manufacturer's instructions. The concentration and purity of RNA were determined using a Nanodrop spectrophotometer (Thermo Fisher Scientific, Wilmington, DE, USA). FFPE material was processed for RNA isolation using QuickExtract™ FFPE RNA Extraction Kit (Epicentre, Madison, WI, USA) according to manufacturer's instructions. The concentration of RNA was determined using a Qubit® 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA).

MicroRNA microarrays

A screening for miRNAs was performed using the miRCURY LNA™ microRNA array (7th gen, Cat # 208500, Exiqon, Vedbaek, Denmark) by the Exiqon miRNA array service. Briefly, 5 μg of total RNA from 6 GG and 5 control cortex samples were labelled using the miRCURY LNA™ microRNA Hi-Power Labeling Kit, Hy3™/Hy5™ and hybridized on the miRCURY LNA™ microRNA Array. After washing, the slides were scanned and analysed using ImaGene^® 9 (miRCURY LNA™ microRNA Array Analysis Software, Exiqon, Vedbaek, Denmark). The quantified signals were corrected for background (Normexp with offset value 10 [42]), and normalized using the global Lowess (Locally Weighted Scatterplot Smoothing) regression algorithm.

MicroRNA target prediction

Using the online database miRWalk (version 2.0; http://zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk/) a total of 5 different databases (miRWalk, Microt4, miRanda, miRDB and Targetscan) were used to identify predicted targets of miR-519d and miR-4758. Targets were considered relevant if found in at least 3 of the 5 databases.

Quantitative real-time PCR analysis

miRNA (miR-519d-3p, miR-4758-3p, miR-664b-3p, miR-4714-5p, miR-5681b and a reference gene, U6B small nuclear RNA gene, Rnu6B) expression was analyzed using Taqman microRNA assays (Applied Biosystems, Foster City, CA, USA). cDNA was generated using the Taqman MicroRNA reverse transcription kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions, and the PCRs were run on a Roche Lightcycler 480 thermocycler (Roche Applied Science, Basel, Switzerland).

To evaluate miRNA targets (TIMP2, AKT3, ERBB3, ERBB4, PI3KCA, TP53, MDM2, CDK4, RB1, CDK2, CDKN1A (P21), CDKN1B (P27), JAK1 and PTEN), 1 μg of cell culture derived total RNA or 500 ng of human brain material-derived total RNA were reverse-transcribed into cDNA using oligo dT primers (Supplementary Table 1). EF1α was used as a reference gene. PCRs were run as described previously [43] on a Roche Lightcycler 480 thermocycler (Roche Applied Science, Basel, Switzerland).

Quantification was performed using the computer program LinRegPCR in which linear regression on the Log (fluorescence) per cycle number data is applied to determine the amplification efficiency per sample [44, 45]. The starting concentration of each specific product was divided by the starting concentration of reference genes and this ratio was compared between groups.

Cell transfection

The human pediatric low grade astrocytoma cell line (WHO grade II: Res-259; LGG2) was kindly provided by Dr Chris Jones (Institute of Cancer Research, Sutton, UK). Cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM)/HAM F10 (1:1) (Gibco, Life Technology) supplemented with 50 units/ml penicillin, 50 μg/ml streptomycin and 10% fetal calf serum (FCS) in a humidifier incubator at 37°C with 5% CO₂. Oligonucleotides were delivered to the cells using Lipofectamine® 2000 transfection reagent (Life Technologies, Grand Island, NY, USA) in a final concentration of 50 nM for a total of 24 hours. Cells were transfected with mimic pre-miRNA (Applied Biosystems, Carlsbad, CA, USA) for miR-519d-3p and/or miR-4758-3p. Cells treated with lipofectamine without mimic were used as a control. Cells were washed twice with PBS before harvesting.

Cell-cycle analysis

At 24 hours after transfection, cells were collected and fixed in 90% ethanol at 4°C for another 24 hours. Fixed cells were then washed twice with PBS, resuspended in 200 μl PBS containing 1 mg/ml propidium iodide and 1 g/ml RNase and incubated for 10 min at 37°C. Cell cycle analysis was performed using the Fluorescent-Activated Cell Sorter Canto II (BD FACSCanto II, BD Biosciences, San Jose, CA, USA).

Statistical analysis

Statistical analyses were performed with Graphpad Prism® software (Graphpad software Inc., La Jolla, CA, USA). Continuous variables were described with mean and ranges; categorical variables with proportions and percentages. The non-parametric Mann-Whitney test was used to assess differences between two groups. For multiple groups, the non-parametric Kruskal–Wallis test was used followed by a Mann-Whitney test to assess differences between groups. The spearman's rank correlation test was used to assess the correlation between miR-519d-3p and miR-4758-3p. Group differences and correlations were considered significant if p<0.05.