Seven to eight weeks old female ICR mice (InVivos, NUS) were made diabetic with a single intra-peritoneal injection of streptozotocin (STZ, 75 mg/kg body weight, Sigma-Aldrich, St. Louis, MO, USA), freshly prepared in 0.01 M citrate buffer (pH 4.5). A week later, the blood glucose levels of the mice were tested using a blood glucometer (Abbott's laboratories, Illinois, USA). Mice with non-fasting blood glucose levels of >200 mg/dl were considered diabetic, and were used for mating. Three to four diabetic female mice were caged with one healthy age-matched male mouse. The day that the copulation plug became apparent, was marked as embryonic day 0.5 (E0.5). In addition, time mated adult female pregnant ICR mice were procured from InVivos, NUS. At E13.5, the embryos from diabetic or control mice were collected through cesarean section, after anesthetizing them with pentobarbital (150 mg/kg body weight). The forebrains of the embryos were dissected and cultures of NSCs were set up. Only embryos from mothers with non-fasting blood glucose levels of >300 mg/dl were taken as experimental group. The procedures pertaining to animal usage was in accordance with the guidelines laid by Institutional Animal Care and Use Committee (IACUC), NUS.

The NSCs were isolated from the forebrain region of the embryonic brain. Briefly, the forebrains were dissected from the embryos and mechanically dissociated in 1X PBS solution. The dissociated cells were washed twice in 1X PBS and seeded in DMEM/F12 (1:1, ThermoFisher Scientific, Waltham, MA, USA). The cells were cultured in DMEM/F12 medium with 10 mM/L D-Glucose concentration, supplemented with insulin-transferrin-selenium (ThermoFisher Scientific), 20 ng/ml EGF (Sigma-Aldrich, St. Louis, MO, USA), and an antibiotic antimycotic solution (Sigma-Aldrich) in T-75 flasks (Greiner, Sigma-Aldrich). The cultures formed neurospheres that were maintained at 37°C/5% CO2 for 3–4 days, after which the supernatant containing the neurospheres was collected in 50 ml tubes (Greiner Bio-One GmBH, Germany) and centrifuged at 1,000 rpm/5 min. The supernatant was discarded and the cell pellets were dissociated with TryPLE™ Select (Gibco, Life technologies, Carlsbad, CA, USA) and re-plated into T-75 flasks. After 3–4 days the cultures were passaged again and used for different experiments. NSCs from normal pregnancy were cultured for 48 hours in medium with low glucose (2 mM D-glucose) or high glucose (40 mM D-glucose) concentration to mimic hypo- or hyperglycemia or in 1% oxygen for 4 hours to mimic hypoxia.

Total RNA was isolated from NSCs from the different experimental groups using miRNeasy RNA isolation kit (Qiagen, Germany). Microarray profiling and data analysis were conducted at Exiqon services, Denmark. The quality of total RNA was first verified by Agilent 2100 Bioanalyser profile. Using the miRCURY LNA™ microRNA Hy3™/Hy5™ Power Labeling Kit, (Exiqon, Denmark), 750 ng of total RNA from sample (Hy3™ label), or reference (Hy5™ label) was labeled with a fluorescent label. Subsequently, the Hy3™-labeled samples and a Hy5™-labeled reference RNA samples were mixed pair-wise, and hybridized to the miRCURY LNA™ microRNA Array (6th gen—hsa, mmu & rno, miRBASE V16.0; Exiqon, Denmark). Hybridization was performed using a Tecan HS 4800™ hybridization station (Tecan, Austria), in accordance to the miRCURY LNA™ microRNA array instruction manual. Following hybridization, the miRCURY™ LNA array slides were scanned using the Agilent G2565BA Microarray Scanner System (Agilent Technologies, Inc., USA) and stored in an ozone free environment (ozone level below 2.0 ppb). The images were analyzed using the ImaGene 9.0 software (BioDiscovery, Inc., USA). The quantified signals were background corrected (Normexp with offset value 10, Ritchie et al., 2007) and normalized using the Quantile normalization algorithm. To identify the miRNAs that were statistically different between two NSC groups, a two tailed t-test was used, while one way ANOVA was used for more than two groups.

Total RNA (including small RNAs) was isolated from various experimental groups of NSCs using the Qiagen miRNEASY kit (Qiagen GmbH, Hilden, Germany), according to the manufacturer's instructions. The extracted total RNA was used for mRNA and miRNA qPCR analysis.

The cDNA synthesis was performed using 2 μg of RNA, 2 μl oligodT, 200 U of moloney murine leukemia virus (M-MLV) reverse transcriptase, 5 U of RNasin (Promega, Madison, WI, USA), and 2 mmol/L of dNTPs, in a 25 μl reaction volume. Firstly, 2 μg of RNA was mixed with 2 μl of oligodT and incubated at 70°C for 5 min in Bio-Rad™ T100 thermal cycler, following which the tubes were kept on ice for 1–2 min. Subsequently, 5X reaction buffer (5 μl), reverse transcriptase (1 μl), RNAase inhibitor (0.7 μl), dNTPs (0.5 μl), and nuclease free water, were added to the tubes and the cDNA synthesis was performed, by incubating the tubes at 42°C for 50 min, 95°C for 5 min and stored at 4°C. Gene expression was analyzed in Applied Biosystems (Model 7900 HT, Applied Biosystems, Foster City, CA, USA) using 10 μl reaction mix containing 5 μl SYBR green (Qiagen), 1 μM/L of each primer, 1 μl cDNA (from individual samples), and 3 μl nuclease free water in 96 well-FAST optical plates. Duplicates of each sample were loaded and an average was taken for further analysis. The 2^−ΔΔCt method was used to calculate the fold change in mRNA expression (Livak and Schmittgen, 2001).

To analyze miRNA expression, cDNA conversion of miRNA (from individual samples) was carried out using Universal cDNA synthesis kit II (Exiqon, Denmark). The expression of miRNA-30b and miRNA-30d were quantified by RT-PCR with Exilent SYBRGreen master mix (Exiqon) and microRNA primers for miRNA-30b or miRNA-30d (Exiqon) in 96 well-FAST optical plates (7900 HT, Applied Biosystems). Duplicates of each sample were loaded and an average was taken for further analysis. U6 snRNA (Exiqon) was used as the control. The 2^−ΔΔCt method was used to calculate the fold change in miRNA expression.

Total protein was extracted from NSCs using M-PER extraction reagent (ThermoScientific) and quantitated using the Bradford method (Bio-Rad, Hercules, CA, USA). Thirty micrograms of protein from each sample was denatured at 95°C for 5 min and loaded into 8 or 10% SDS-PAGE gel. Following electrophoresis, gels were blotted onto PVDF membranes and blocked in 5% non-fat milk for 1 h at room temperature. Primary antibodies were diluted in 5% milk and the PVDF membranes were incubated with the appropriate primary antibodies, rabbit anti-Sirt1 (ab12193, Abcam, Cambridge, United Kingdom) or mouse anti-Sirt1 (ab50517, Abcam), rabbit anti-Map2 (H-300, Santa Cruz Biotechnology, Dallas, Texas, USA), rabbit anti-Gfap (MAB 360, Millipore, Billerica, Massachusetts, United States), mouse anti-CNPase (MAB 326, Millipore), or mouse anti- beta actin (A1978, Sigma-Aldrich) overnight at 4°C. Subsequently, HRP-conjugated secondary antibodies [Anti-mouse HRP-31430; Anti-rabbit HRP-31460 (ThermoScientific)] were added and incubated for 1 h at room temperature. Chemiluminescence was detected using either pico or femto (ThermoScientific) substrates and signals were captured on X-ray films. The bands on the X-ray film was scanned and quantified using GS-800 calibrated densitometer (Bio-Rad, USA). Rabbit anti-Sirt1 was used for all experiments except to check Sirt1 silencing efficiency where mouse anti-Sirt1 was used.

NSCs from normal pregnancy were used to perform gain/loss-of-function studies. NSCs were trypsinized as mentioned previously and counted using the Bio-Rad TC 20 Automated cell counter, before seeding in 6-well-plates at a cell density of 1 × 10⁶ cells/well. Transfection was carried out using XTremeGene SiRNA transfection reagent, (Roche, Basel, Switzerland) for siRNA probes (Sirt1-4390771; negative control-4390843, Ambion, ThermoFisher Scientific) in Opti-MEM media at a final concentration of 50 nM. Transfection of miR-30b mimics (Ambion, ThermoFisher Scientific) and negative control probes (Ambion, ThermoFisher Scientific) were performed using lipofectamine RNAiMAX (ThermoFisher Scientific) in OptiMEM media at a final concentration of 10 nM. Opti-MEM medium was changed to NSC culturing media after 6 h post transfection. RNA was extracted at 48 h post transfection while protein extraction was performed 72 h post transfection.

Mouse microglial BV-2 cells were maintained in DMEM containing 10% fetal calf serum (FCS) at 37°C in a humidified incubator under 5% CO₂. BV2 cells were co-transfected with miR-30b mimics or negative control mimics (at 10 nM) and plasmid vector containing luciferase gene and 3′UTR of Sirt1 gene (308 ng) (MmiT054166-MT06) using Lipofectamine® RNAiMAX (Thermofisher Scientific). The cells were cultured for 48 h after which luciferase activity was assayed according to the manufacturer's instructions using LUC-Pair Duo-Luciferase Assay kit 2.0 (LPFR-P010, Genecopoeia, Rockville, USA). The luminescence intensity was measured using a luminometer (Spectramax M5) and firefly luciferase activity was normalized to renilla luciferase activity.

The neurospheres were harvested and transferred to poly-L-lysine coated coverslips and poly-L-ornithine coated coverslips (the dissociated cells) and cultured in 24 well-plates for appropriate time points. The cells were fixed with ice cold 4% paraformaldehyde (PFA) for 30 min and the cells were permeabilized with phosphate buffered saline (PBS) containing 0.1% Triton-X 100 (PBS-T) for 20 min. Following that, the cells were blocked with 5% normal goat serum for 1 h before incubating with mouse anti-Sirt1antibody (1:500, ab50517, Abcam) overnight at 4°C. The next day, cells were incubated with Cy3-conjugated anti-mouse IgG secondary antibody (1:100, Chemicon, Temecula, CA, USA) for 1 h at room temperature. Finally, the nucleus was counterstained with DAPI and the coverslips were mounted on glass slides with fluorescent mounting medium (DAKO, USA). Images were captured with Olympus FV1000 confocal microscope.

Pathway analysis using Ingenuity Pathway Analysis software (IPA, Qiagen) was performed to identify the mRNA targets of the differentially expressed microRNAs using expression levels of miRNAs in each of the NSC groups with a significant fold change of <−1.2 or >1.2 (p < 0.05) as input. IPA's microRNA target filter was applied to each dataset, and core analysis was performed to identify the canonical biological pathways. Comparison analysis was done to identify the common biological pathways among the different groups.

At least three independent experiments were performed and the data was represented as mean ± SD. To identify the miRNAs that were statistically different between two NSC groups, a two-tailed t-test was used, while one way ANOVA was used for more than two groups. Student's t-test was done using Microsoft Excel spread sheet and data was considered significant when p < 0.05 (Control vs. Experimental, n = 3–4).

In order to determine whether maternal diabetes alters the expression of miRNAs in NSCs of the developing brain, global miRNA expression profiling was carried out in NSCs isolated from embryos of control and diabetic pregnant mice using miRCURY LNA™ miRNA Array (containing ~2,383 capture probes). Since maternal diabetes can cause fetal hyper/hypoglycemia (Schwartz and Cornblath, 2000) or hypoxia (Teramo et al., 2013), NSCs derived from control pregnancy were exposed to high glucose to mimic hyperglycemia (HG) or low glucose to mimic hypoglycemia (LG) or 1% oxygen to mimic hypoxia (OX) in vitro and used for the microarray analysis.

Heat map in Figure 1A shows the hierarchical clustering performed on all samples with regard to the top 50 miRNAs with the highest standard deviations (Table 1). Only those miRNAs that were expressed/ detected in all NSC groups were included.

Since our study involved several experimental groups, we performed Principal Component Analysis (PCA) on the top 50 miRNAs with highest standard deviations, to identify differences and similarities between the experimental groups. This method was used to trim down large data sets and cluster samples based on their expression profile, which results in separation of samples in different regions of the plot. The PCA analysis revealed that the NSCs from diabetic pregnancy were distinct (based on the expression profile) from the other experimental groups (Figure 1B) and hence were chosen for further analysis. Moreover, the LG group was excluded for further analysis since they did not cluster closely in the PCA plot.

Ingenuity Pathway Analysis (IPA) was performed to identify biological pathways regulated by differentially expressed miRNAs. miRNAs with a fold change <1.2 or >1.2 and with p < 0.05 were selected in each of the NSC groups. A total of 25 miRNAs [Control (Con) vs. HG], 54 miRNAs (Con vs. OX), and 104 miRNAs [Con vs. Diabetic (Dia)] were found to be differentially expressed (Tables 2–4). The number of differentially expressed miRNAs in each group and miRNAs common between the groups are shown in Figure 1C. A comparison revealed that seven miRNAs (let7^*, let-7a-5p, miR-17^*, miR-320, miR-668, miR-767, miR3068-3p), were commonly deregulated in NSCs from diabetic pregnancy, hypoxia and HG groups when compared to the control (Table 5). The expression levels of these deregulated miRNAs were similar across the groups (i.e., they were either down regulated or up regulated in all three groups when compared to the control) except for let 7^*, suggesting that at least six miRNAs had crucial roles in response to glucose or glucose-induced hypoxia. In order to identify the significance of the seven commonly deregulated miRNAs, IPA pathway analysis was carried out. From this analysis, it appeared that majority of the gene targets of five out of seven miRNAs were from the axonal guidance pathway (Figures 2A,B, Table 6).

Furthermore, three miRNAs were common between Con vs. HG and Con vs. Dia groups (Supplementary Table 1), 11 miRNAs were common between Con vs. HG and Con vs. Hypoxia groups (Supplementary Table 2), and 26 miRNAs were common between Con vs. Dia and Con vs. Hypoxia groups (Supplementary Table 3). The expression levels of the majority of miRNAs were similar between the groups compared and large numbers of miRNAs were common between Con vs. Hypoxia and Con vs. Dia groups, suggesting that these miRNAs could be biomarkers of high glucose/hypoxia.

Using microRNA target filter, the mRNAs predicted to be targets of the differentially expressed miRNAs were selected and core analysis was performed for each group vs. the control. The top five canonical pathways predicted in each group, signify that maximum miRNAs deregulating that pathway are altered by glucose or glucose-induced hypoxia (Table 7). Comparison analysis revealed that axonal guidance signaling was the top canonical pathway regulated by the differentially expressed miRNAs in NSCs exposed to different conditions (Table 8). miRNAs deregulated in NSCs from diabetic pregnancy and their targets in axonal guidance signaling pathway are represented in Figure 3. The expression of specific genes (Robo1, Ntn1, Nrp1, Ntng1, Gsk3b, Rac1, Rock2, Clf2, Efnb3) from this pathway were quantitated by qPCR, and five genes (Robo1, Ntn1, Nrp1, Ntng1, Efnb3) were found to be down regulated in NSCs from diabetic pregnancy when compared to the control (Figure 4), suggesting that hyperglycemia deregulates axonal guidance signaling.

From the microarray results, a total of 104 miRNAs were found to be differentially expressed in NSCs from embryos of diabetic pregnancy when compared to the control. Among the differentially expressed miRNAs, members of miRNA-30 family (miRNA 30a, b, c, d, and e) were chosen for further studies as they have been found to be: (a) upregulated significantly (fold change ranging from 1.15–1.52) in NSCs from embryos of diabetic pregnancy (Supplementary Table 4); (b) involved in neurodevelopmental disorders (Mellios and Sur, 2012; Hancock et al., 2014; Sun et al., 2014; Han et al., 2015). Further, quantitative RT-PCR analysis was performed to validate the expression levels of miRNA-30 family (miRNA-30 b, c, d, and e) in NSCs from embryos of diabetic and control pregnancy. While miR-30b and miR-30d were significantly up regulated in NSCs from embryos of diabetic pregnancy when compared to the control, miRNA-30c and miRNA-30e showed an increasing trend (Figure 5A). Gene targets of the miRNA-30 family were predicted using IPA (Supplementary Figure 1).

From the pathway analysis, Sirtuin1 (Sirt1), which is predicted to be one of the targets of miRNA-30b was selected for further analysis as it has been shown to be involved in NSC differentiation and fate determination during brain development (Hisahara et al., 2008; Prozorovski et al., 2008). Quantitative RT-PCR and western blot showed significant down regulation of the Sirt1 gene and protein expression (Figures 5B–D), respectively, in NSCs from embryos of diabetic pregnant mice when compared to normal pregnant mice.

The NSCs from normal pregnancy were transfected with miRNA-30b or miRNA-30d mimics, in order to overexpress these miRNAs. Scrambled nucleic acids were transfected in NSCs from normal pregnancy and they were used as the negative control. Following transfection, the expression of miRNA-30b and miRNA-30d increased 60-fold (Figure 6A) and 40-fold (Figure 6B), respectively, when compared to negative control. Western blot analysis revealed that overexpression of miRNA-30b but not miRNA-30d could significantly decrease the expression of Sirt1 protein suggesting that Sirt1 may be a target of miRNA-30b (Figures 6C,D). In addition, immunostaining and confocal imaging following miR-30b overexpression revealed decreased expression of Sirt1 in miRNA-30b overexpressed cells when compared to negative control (Figure 6E).

In order to confirm that Sirt1 is a direct target of miR-30b, a 3′UTR plasmid assay was performed. Sirt1 plasmid containing the 3′UTR of Sirt1 gene was co-transfected with miRNA-30b mimic or negative control mimic in BV2 cells. Following transfection, the luminescence intensity was found to be significantly decreased in miRNA-30b mimic transfected cells (Figure 7A) suggesting that miRNA-30b binds to its complementary site in the 3′UTR of Sirt1 and directly regulates the expression of Sirt1.

The expression levels of neuronal and glial markers were determined following overexpression of miRNA-30b in NSCs in order to understand the role of miRNA-30b on NSC fate specification. Overexpression of miRNA-30b resulted in significant decrease in the expression of MAP2 (neuronal marker) and CNPase (early oligodendrocyte marker) proteins and increase in the expression of GFAP (astrocyte marker) protein when compared to that in negative control (Figures 7B,C).

In order to verify that miRNA-30b increased astrocytes via down regulating its specific target Sirt1, Sirt1 was silenced using siRNA in NSCs and the expression of MAP2, GFAP, and CNPase proteins were analyzed. Sirt1 silencing efficiency was found to be 40% (Figures 8A,B). In addition, immunostaining and confocal imaging confirmed decreased Sirt1 expression after siRNA mediated Sirt1 silencing (Figure 8C). siRNA-mediated knockdown of Sirt1 in NSCs led to an increase in GFAP with a concomitant decrease in MAP2 proteins similar to miRNA-30b overexpression (Figures 8D,E). However, knockdown of Sirt1 also increased the expression of CNPase (in contrast to miRNA-30b overexpression).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.