Male Sprague–Dawley rats (Charles River), weighing 200–250 g, were housed under diurnal lighting conditions (12 h darkness/light) and in a conditioned room (23°C). The experiments were performed according to the international guidelines for animal research and approved by the Animal Care Committee of “Federico II,” University of Naples, Italy. The animals, during any surgical or invasive procedure, were anesthetized using a mixture of oxygen and sevoflurane at 3.5% (Medical Oxygen Concentrator LFY-I-5A), and the rectal temperature was maintained at 37 ± 0.5°C with a heat-controlled mat (Harvard Apparatus). The rats were randomly assigned into experimental and control groups.

Among the 55 animals used for the present study, seven have been excluded from the statistical analysis. In particular, two died during the surgical procedures of hemorrhage induction, and five were excluded for lack of achievement of stroke models.

Stroke was induced by middle cerebral artery occlusion (MCAO) by introducing a suture filament into the internal carotid artery until the middle cerebral artery (24), modified and readapted in our laboratory (25). Briefly, under an operating stereomicroscope (Nikon SMZ800, Nikon Instruments, Florence, Italy), right carotid bifurcation was identified and exposed by using surgical pincers (Dumont #7, FST), and the external carotid artery near the bifurcation was cut and electrocauterized to create a stump on the artery. A silicon-coated nylon filament (Doccol, CA, USA) was inserted through the stump and gently advanced 19 mm into the right internal carotid artery in order to reach the origin of the middle cerebral artery. All animals were sacrificed after 24 h from MCAO. In the model of transient ischemia, the filament was removed after 100 min, and the animals were sacrificed after 3 and 24 h from reperfusion.

Intracerebral hemorrhagic stroke was induced in rats as previously described (26, 27) and adapted in our laboratory. Briefly, using a stereotaxic apparatus, the anesthetized rats were injected with 0.5 U collagenase 2-μl volume dissolved in phosphate-buffered saline (collagenase type VII from Sigma-Aldrich, catalog number C2399) through a Hamilton syringe (26-gauge needle, 10-μl volume) in a burr hole into the right striatum (3 mm lateral to midline, 0.5 mm anterior to bregma, and 5.2 mm below the surface of the skull) over 5 min. The syringe was kept in place for 10 min and removed over 5 min. The animals were sacrificed after 24 h from collagenase injection.

A catheter was inserted into the femoral artery to measure the arterial blood gases before and after ischemia (Rapid lab 860; Chiron Diagnostic, Medfield, MA, USA). Induction of ischemia was confirmed by monitoring the regional cerebral blood flow (CBF) in the area of the right MCA through a disposable microtip fiber optic probe (diameter, 0.5 mm) applied on the right temporo-parietal region of the skull, connected through a Master Probe to a laser Doppler computerized main unit (PF5001; Periflux system, 5000, Perimed AB, Järfälla, Sweden), and analyzed using PSW Perisoft 2.5 (28). CBF monitoring was continued up to 30 min after the end of the surgical procedure once the occurrence of reperfusion was verified.

Brain regions corresponding to the ischemic core and penumbra were dissected from rats subjected to sham surgery and transient middle cerebral artery occlusion (tMCAO). Total RNA from brain tissues was extracted with Trizol, following the instruction of the supplier (TRI Reagent®-Sigma), and RNA quality was assessed using a Thermo Scientific™ NanoDrop™ One Microvolume UV–vis spectrophotometer. The RNA samples were sent to LC Sciences (Houston, Texas, USA), a global biotechnology company which performed a miRNA microarray analysis, including separation, quality control, labeling, hybridization, and scanning. The array contained 810 mature rat miRNA probes for the whole Rattus norvegicus miRNome based on a database of published miRNA sequences and annotation (Sanger miRbase Release 21.0). In detail, hybridization was performed on a μParaflo microfluidic chip using a micro-circulation pump (Atactic Technologies, Inc., Houston, TX, USA). After the hybridization, the chips were washed, and then the fluorescence data images were collected using a laser scanner (GenePix 4000B; Molecular Devices, LLC, Sunnyvale, CA, USA) and digitized using Array-Pro image analysis software (Media Cybernetics, Inc., Rockville, MD, USA). The chips were scanned at a pixel size of 10 μM with Cy3 Gain and Cy5 Gain at 460 and 470 nm scanning, respectively. The data were analyzed by first subtracting the background and then normalizing the signals using a locally weighted regression filter.

The datasets presented in this study can be found in the online GEO dataset repositories.

A significantly different expression pattern between tMCAO and sham-operated groups was obtained by microarray analysis, as shown in the volcano plot graph (Figure 1, at the top). A clustered heat map was then extrapolated, showing a colorful illustration of miRNA expression profiles across three animals for each experimental condition (Figure 1, at the bottom).

Blood samples were withdrawn from the tail vein of anesthetized rats before the ischemia induction and at different time intervals from reperfusion by using a 1-ml syringe with a 23-G needle and collected in BD Vacutainer tubes (K3 EDTA 5.4 mg). To separate the plasma, the blood samples were centrifuged in the same collecting tubes at 1,500 × g (2,900 rpm) for 8 min at room temperature in a ALC PK 120 centrifuge. The supernatant plasma was transferred to sterile tube and centrifuged in an Eppendorf centrifuge at 11,000 rpm to purify the sample from any cellular residues. Prior to RNA extraction, the absorbance at 415 nm of a 50-μl aliquot for each sample was measured in a Bio-rad Microplate Reader to evaluate the presence of free hemoglobin because of a previous hemolytic process. The data reported in the literature suggest working on plasma samples with absorbance values below 1.0 OD. This restriction is necessary in order to obtain an evaluation of microRNAs that are present in exosome or free in plasma and released after ischemic lesion, excluding those present in the red and white blood cells and released following hemolysis.

RNA samples were prepared from the ischemic core, corresponding to the striatum, and penumbra region, corresponding to the temporo-parietal brain surrounding cortex. These brain regions were dissected from rats subjected, respectively, to sham surgery and to tMCAO. The sham-operated rats are, in fact, healthy animals. This experimental group has been chosen in order to compare the miRNA expression to that of a control group that underwent similar conditions in terms of anesthesia and rat manipulation. Total RNA from brain tissues was extracted with Trizol (TRI Reagent®-Sigma) following the instruction of the supplier, and RNA quality was assessed using a Thermo Scientific™ NanoDrop™ One Microvolume UV–vis spectrophotometer.

The expression levels of 15 microRNAs selected from the microarray analysis performed on ischemic brain were singularly evaluated by real-time PCR in plasma of rats subjected to permanent or transient focal cerebral ischemia or to intracerebral collagenase-induced hemorrhage (ICH) at different duration times (0.5, 3, and 9 h from ischemic stroke induction).

miRNA isolation from plasma samples was performed with miRNeasy Serum/Plasma Kit (Qiagen) according to the protocol of the manufacturer. For miRNA analysis in the plasma samples, not specific concentration but precise volumes (5 μl) of RNA were retrotranscribed by using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and Taqman probes (Thermo Fisher Scientific), following TaqMan Small RNA Assays Protocol (16°C for 30 min, 42°C for 30 min, and 85°C for 5 min). Quantitative real-time polymerase chain reaction was performed with TaqMan Universal PCR Master Mix II (Applied Biosystems) in a 7500 Fast Real-Time PCR System (AB Applied Biosystems). The cDNA samples were amplified simultaneously in triplicate in one assay run, following the protocol for Taqman assays: 50°C for 2 min, 95°C for 10 min, 40 cycles of amplification of 95°C for 15 s, and 60°C for 1 min. The results were analyzed and exported with 7500 Fast System SDS Software.

The TaqMan probes used are the following: mmu-miR-210^* (ID: 462444_mat), hsa-miR-150 (ID: 000473), hsa-miR-30e (ID: 002223), hsa-miR-181b (ID: 001098), hsa-miR-191 (ID: 002299), hsa-let-7b (ID: 002619), hsa-miR-30a-5p (ID: 000417), hsa-miR-16 (ID: 000391), hsa-miR-27b (ID: 000409), mmu-miR-212 (ID: 002551), hsa-miR-218 (ID: 000521), hsa-miR-126 (ID: 002228), hsa-let-7c (ID: 000379), hsa-miR-101 (ID: 002253), hsa-miR-9^* (ID: 002231), and miRNA Control Assay U6 snRNA (ID: 001973).

The extensive file of predicted or verified targets of all aberrantly expressed microRNAs in plasma from rats subjected to brain ischemia of different entity, permanent or transient, or origin, hemorrhagic or occlusive, indicates that a large group of genes may be potentially dysregulated since the prenatal period of life. However, for the present study, the attention has been mainly focused on those genes coding plasma membrane proteins controlling ion influx or efflux, whose regulation has been investigated and acknowledged in stroke mechanisms and in several neuroprotective approaches (Supplementary Table 1). Four different web servers, each one based on a specific algorithm, were used to predict the targets of those microRNAs that have been found to be dysregulated in plasma from rats subjected to brain ischemia of different entity, permanent or transient, or pathophysiology, hemorrhagic or occlusive.

TargetScan is a computational method to predict the targets of conserved vertebrate miRNAs, integrating the model of miRNA–mRNA interaction on the basis of thermodynamics and sequence alignment analysis between miRNA binding sites among different species (29). The latest updated 7.2 version examined the function of non-canonical binding sites identified in recent studies and considers 14 different features of the microRNA, microRNA site, or mRNA to predict those sites within mRNAs that are most effectively targeted by microRNAs (30).

MiRDB is an online database for miRNA target prediction and functional annotations (31). All targets were predicted by a bioinformatics tool, MirTarget, which was developed by analyzing thousands of miRNA-target interactions from high-throughput sequencing experiments (32).

PicTar uses the criteria of co-expression in space and time of miRNAs and their targets through combinations of different microRNAs (33). This algorithm requires that the binding stability of the putative miRNA-target interaction, measured by thermodynamic binding energy, is higher than a specified threshold.

miRmap is an open source software library which employs thermodynamic, evolutionary, probabilistic, or sequence-based features (34), making it currently the most comprehensive miRNA target prediction resource. Only miRNA-mRNA interactions with a miRmap score above 70 were considered for the present study.

As regard the microarray experiments, Student's t-test analysis was conducted for individual comparisons between the two experimental groups. The false discovery rate was P < 0.05 and served as the cut-off criteria. The data were log₂ transformed and median centered by Cluster 3.0 software (Informer Technologies Inc., Los Angeles, CA, USA). Real-time PCR results are expressed as fold change (2^−ΔΔCt) compared to the control group set to 1, following the instructions provided by the literature (35). Briefly, the difference between the Ct values of the gene of interest and the internal control (ΔCt) is calculated for both control sample and target sample. Then, the difference between the ΔCt of the target sample and the control sample (ΔΔCt) is calculated. The fold change of gene expression of target samples compared to the control sample is calculated as 2^−ΔΔCt. Values are expressed as means ± standard deviation. Statistical analysis was performed with GraphPad Prism 5.0 (GraphPad Software, Inc., San Diego, CA) using one-way analysis of variance followed by Newman–Keuls post-hoc test for groups of more than two. Statistical significance was accepted at 95% confidence level (p < 0.05).

Among the 15 microRNAs selected from the microarray analysis, 11 miRNAs showed no significant difference in plasma samples at all times assessed after cerebral ischemia compared to the control group (Figures 2A–C,E–J,L,O). By contrast, miR-16-5p, miR-101a-3p, miR-218a-5p, and miR-27b-3p resulted to be upregulated in plasma samples withdrawn up to 3 h of permanent middle cerebral artery occlusion (pMCAO) (Figures 2D,K,M,N). These miRNAs did not display an overexpression after 9 h of permanent ischemia.

Altered miRNAs in the plasma of rats subjected to ischemic stroke were evaluated also in plasma from ICH rats (Figure 3); among them, miR-16-5p and miR-27b-3p significantly increased 30 min after hemorrhagic stroke, but the levels were already restored at 3 h (Figures 3D,N). Conversely, miR-101a-3p and miR-218a-5p did not show expression changes at any time evaluated (Figures 3K,M). Moreover, some microRNAs whose plasma levels were not modulated by ischemic stroke displayed upregulation after ICH. Indeed let-7c-5p was upregulated only after 30 min from the onset of hemorrhagic stroke (Figure 3G), whereas the let-7b-5p, miR-150-5p, and miR-181b-5p levels significantly increased up to 9 h (Figures 3E,F,H).

Reperfusion affected the expression of miR-16-5p, miR-101a-3p, and miR-27b-3p. In particular, in the plasma of rats subjected to tMCAO, the upregulation of miR-16 and miR-101a-3p was delayed compared to the expression levels assessed after pMCAO, showing a statistical increment after 9 h from reperfusion (Figure 4). By contrast, miR-218 and miR-27b resulted to be not modified by tMCAO compared to the control group (Figure 4). Furthermore, miR-30a, let-7b, let-7c, miR-212, and miR-9, whose expression was not modulated by permanent ischemia, were upregulated after reperfusion (Figures 4A,E,G,L,O).

microRNA–target interaction analysis showed that, among the proteins involved in the control of ionic homeostasis, miR-150-5p activity has been linked, among others, to the control of immune response after stroke. In addition, we identified sodium/calcium exchanger, NCX1, and zinc transporter, ZnT6, and transient receptor potential cation channel, TRPM7, as putative additional targets. All these membrane proteins are implied in cellular ionic homeostasis, and their activity has been strongly linked to stroke pathophysiology (36–38). Similarly, among miR-180 targets linked to stroke pathophysiology, the membrane channel acid sensing ionic channel, ASIC1, and the Na+/H+ exchanger it should be mentioned (Tables 1, 2).