Abstract

   Background: Signaling pathways mediated by microRNAs (miRNAs) have been
   identified as one of the mechanisms that regulate stroke progression
   and recovery. Recent investigations using stroke patient blood and
   cerebrospinal fluid (CSF) demonstrated disease-specific alterations in
   miRNA expression. In this study, for the first time, we investigated
   miRNA expression signatures in freshly removed human stroke brain
   tissue. Methods: Human brain samples were obtained during craniectomy
   and brain tissue resection in severe stroke patients with
   life-threatening brain swelling. The tissue samples were subjected to
   histopathological and immunofluorescence microscopy evaluation, next
   generation miRNA sequencing (NGS), and bioinformatic analysis. Results:
   miRNA NGS analysis detected 34 miRNAs with significantly aberrant
   expression in stroke tissue, as compared to non-stroke samples. Of
   these miRNAs, 19 were previously identified in stroke patient blood and
   CSF, while dysregulation of 15 miRNAs was newly detected in this study.
   miRNA direct target gene analysis and bioinformatics approach
   demonstrated a strong association of the identified miRNAs with
   stroke-related biological processes and signaling pathways.
   Conclusions: Dysregulated miRNAs detected in our study could be
   regarded as potential candidates for biomarkers and/or targets for
   therapeutic intervention. The results described herein further our
   understanding of the molecular basis of stroke and provide valuable
   information for the future functional studies in the experimental
   models of stroke.

   Keywords: stroke, malignant hemispheric infarction, microRNA, next
   generation microRNA sequencing

1. Introduction

   Ischemic stroke accounts for up to 71% of all stroke cases and 51% of
   all stroke-related deaths worldwide, and is currently among the top
   leading causes of serious, long-term disability [[38]1,[39]2].
   Identification of the molecular markers of stroke provides the
   potential to predict severity and clinical outcome, as well as to
   identify possible targets for therapeutic intervention.
   Stroke-associated ischemic damage involves blood brain barrier (BBB)
   dysfunction, microvascular injury, cytotoxic and vasogenic edema,
   post-ischemic inflammation, oxidative damage, spreading depolarization,
   and ultimately, the death of neurons, glia, and endothelial cells
   [[40]3,[41]4,[42]5,[43]6,[44]7]. In search of the molecular mechanisms
   and signaling cascades regulating these pathological processes in human
   subjects, researchers focus on the differential gene profiling in
   peripheral blood and CSF collected from stroke patients
   [[45]8,[46]9,[47]10]. Besides the identified alterations in gene
   expression, stroke-associated changes in microRNA profiles are now
   regarded as indicators of the risk, occurrence, and severity of the
   disease [[48]11,[49]12,[50]13].

   MicroRNAs (miRNAs) are a diverse class of highly conserved small RNA
   molecules that function as critical regulators of gene expression and
   orchestrate a variety of signaling pathways involved in stroke
   progression [[51]14,[52]15]. Among the wide range of identified miRNAs,
   some are now considered as biomarkers for cerebral ischemia
   [[53]13,[54]14,[55]16]. The stroke-related changes in human gene and
   miRNA expression were detected using the patient blood (whole blood and
   plasma), cerebrospinal fluid, or postmortem brain samples
   [[56]17,[57]18,[58]19,[59]20]. Even though these samples represent
   simple and clinically relevant alternatives to brain tissue, the
   interpretation of the obtained gene and miRNA expression profiling
   results remains problematic. While these data may predict a clinical
   outcome or describe the systems response to disease, they do not
   distinguish the brain tissue-specific alterations from the molecular
   changes associated with systemic immune response or postmortem
   transformations.

   In the present study, we analyzed human brain tissue samples resected
   during a craniectomy and stroke-ectomy procedure in patients with
   severe (malignant) hemispheric stroke. In up to 10% of stroke patients,
   the infarct and associated edema progresses into severe space-occupying
   brain swelling associated with a high mortality rate of up to 80%
   [[60]21,[61]22,[62]23]. Malignant cerebral infarction caused by middle
   cerebral artery (MCA) or internal carotid artery (ICA) occlusion leads
   to an increased intracranial pressure and subsequent herniation and
   brain stem compression. The affected MCA territory includes a portion
   of the frontal and temporal/parietal lobes. Depending on the location
   of the occlusion, ICA-affected brain area could be smaller or wider
   than that of an MCA territory infarction. Recommended treatments for
   malignant edema include early identification of neurological worsening,
   osmotic therapy, and decompressive craniectomy (DC) with dural
   expansion in select patients [[63]24,[64]25,[65]26,[66]27]. Additional
   removal of infarcted tissue (stroke-ectomy) is performed in cases of
   severe swelling in order to quickly decompress deep structures
   [[67]28,[68]29]. This usually aspirated and discarded tissue provided a
   unique source for our human stroke tissue analysis. Herein, for the
   first time, morphological evaluation, as well as gene and microRNA
   profiling was performed in the freshly removed human stroke brain
   tissue. This approach enables the identification of early (as opposed
   to post-mortem) brain tissue-specific alterations in the live brain
   following cerebral stroke. The obtained data could contribute to
   knowledge on the underlying mechanisms of stroke, as well as lead to
   development of miRNA-based targeting therapies in the future.

2. Materials and Methods

2.1. Tissue Sample Collection

   The sample collection procedures and the roles of all participants in
   the study are regulated by the protocol approved by the University of
   New Mexico Human Research Review Committee (IRB approval: UNM-HRPO
   17-031).

   Surgical procedure: Brain tissue specimens were collected from the
   ischemic stroke patients with large cerebral infarction who underwent
   decompressive hemicraniectomy with stroke-ectomy procedure. During the
   surgery, a large (≥12 cm) craniectomy was created and the dura opened
   in a stellate fashion. After the dura was opened, the brain was
   inspected and if it is still herniating out of the defect, additional
   infarcted temporal lobe tissue was removed and stored immediately in
   RNAlater solution to stabilize the samples for further analyses.
   Additional stroke-ectomy was performed with suction and bipolar cautery
   until the brain was adequately relaxed.

   Sample handling: The resected tissue was taken to the UNM Human Tissue
   Repository and Tissue Analysis Shared Resource (HTR-TASR, an honest
   broker). Control temporal lobe samples of non-stroke patients were
   received from the HTR-TASR as frozen blocks. These biopsy samples were
   collected during a neuro-oncology surgery, either from the tissue
   adjacent to the tumor, or tissue removed when accessing a deep tumor.
   These temporal lobe samples were identified by the pathologist as
   “normal”, based on the absence of gliosis, necrosis, or tumor cells.
   Similar histologically normal tissue has been widely used as reference
   normal samples for genomic and proteomic analyses in cancer research
   [[69]30,[70]31].

2.2. Human Brain Tissue Processing

   Before sectioning, the tissue specimens were cut longitudinally to
   obtain two mirror samples from the same tissue block. One portion of
   the brain tissue was fixed in 10% neutral buffered formalin and
   subjected to sectioning and histological staining, followed by
   morphological evaluation. Another “mirror” part of each sample was
   stored in RNA later and subsequently frozen at −70 °C, for further
   microRNA analysis.

2.3. Histological Staining

   4 μm paraffin-embedded tissue sections were cut using RM2135 microtome,
   Leica (Buffalo Grove, IL, USA). After the deparaffinization, the
   sections were subjected to hematoxylin and eosin (HE) staining. The
   staining was performed using a protocol for standard automated HE
   histopathology staining with Leica Autostainer XL.

2.4. Immunohistochemistry and Fluorescence Microscopy

   Immunofluorescence staining was performed based on the modified
   protocols for human brain tissue labeling described earlier
   [[71]32,[72]33]. After the regular deparaffinization and rehydration
   steps and prior to immunohistochemical labeling, the human tissue
   samples underwent antigen retrieval by incubating sections for 10 min
   in a solution of Tris-buffered saline (TBS) containing 20 μg/mL
   proteinase K. For blocking and permeabilization, the sections were
   incubated for 2 h at room temperature in TBS with 0.05% tween (TBST)
   containing 2% normal goat serum. Incubation with primary antibody was
   performed for 24 h at room temperature. The following antibodies were
   used for the immunofluorescence labeling: mouse monoclonal pan-neuronal
   marker antibody (1:100, MilliporeSigma, Burlington, MA, USA), rabbit
   polyclonal anti-Iba-1 (1:500, Fujifilm, Wako Chemicals, Richmond, VA,
   USA), and mouse monoclonal anti-GFAP (1:100, BD Biosciences, Franklin
   Lakes, NJ, USA). FITC-, and Rhodamine-conjugated secondary antibodies
   (1:500 concentration, 2 h at room temperature incubation) were from
   Jackson Immunoresearch. DAPI staining was used to visualize nuclei.
   Imaging was performed using Zeiss LSM 800 Airyscan confocal microscope,
   using single-scan and tile-scan image acquisitions.

2.5. Next Generation Sequencing (NGS) Analysis

   Tissue samples were sent to Qiagen (Carlsbad, CA, USA) for RNA
   isolation and miRNAseq library preparation. miRNA and small RNA
   sequencing was performed using Illumina NGS sequencing platform. The
   resulting NGS data were analyzed using the CLC Genomics Workbench
   (version 20.0.2) and CLC Genomics Server (version 20.0.2), developed by
   QIAGEN, Aarhus, Denmark. First, quality and adapter/common sequence
   trimming on the reads was performed, and the trimmed reads were grouped
   according to Unique Molecular Identifiers (UMIs) and aligned to miRbase
   v22. Reads were normalized for expression analysis using trimmed mean
   of M-values method (TMM). miRNA differential expression analysis was
   performed using the EdgeR Bioconductor package 3.14.

   Whole Transcriptome RNA-Sequencing Analysis: The unmapped reads from
   the NGS miRNA analysis were extracted, deduplicated and mapped to the
   genome. Gene expressions were calculated by counting number of reads
   mapping to the annotated gene loci.

   All NGS sequencing data have been deposited in NCBI’s Gene Expression
   Omnibus and are accessible through GEO Series accession number
   [73]GSE155257 at:
   [74]https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE155257
   (accessed on 29 July 2020).

2.6. Statistical Analysis

   Statistical analysis was performed using the EdgeR Bioconductor package
   (version 3.36). p-values for significantly differentially expressed
   miRNAs and genes were estimated by an exact test assuming a negative
   binomial distribution. As a result, a list of miRNAs and their target
   genes with differential expression between the stroke and control
   tissue samples were generated and further analyzed to interpret their
   biological significance. The Empirical analysis of gene expression data
   was implemented as the ‘Exact Test’ for two group comparisons
   incorporated in the EdgeR Bioconductor package. This test was
   implemented to calculate miRNA differential expression, in particular,
   for calculations of “Log-Fold-Change” and “FDR corrected p-values”
   which are the main result of the statistical testing. Only the miRNAs
   and genes with 2-fold change in expression in Stroke group (compared to
   Control group) and FDR p < 0.05 are reported as differentially
   expressed. The volcano and PCA plots were visualized using the Enhanced
   Volcano Bioconductor package (version 1.12.0) and EdgeR Bioconductor
   package respectively.

2.7. Evaluation of miR-155 Target Genes

   Human miR-155 Targets RT^2 Profiler PCR Array (Qiagen) was used to
   assess the expression of currently known experimentally verified and
   bioinformatically predicted 84 human genes regulated by miR-155. Three
   RNA samples per stroke and control groups were evaluated. The PCR Array
   Data analysis was performed using an automated PCR Analysis Web Portal
   and GeneGlobe Data Analysis (Qiagen). The p-values are calculated using
   a Student’s t-test followed by a Benjamini-Hochberg correction method.

2.8. Bioinformatics Analysis of miRNA Sequencing, Whole Transcriptome
RNA-Sequencing, and PCR Array Data

   For each differentially expressed miRNA, the experimentally validated
   and predicted target gene list was generated using a multimiR package.
   Subsequently, a degree of association between miRNA target genes and
   specific GO terms from GO Biological Process domain was determined
   using the Gene Ontology Consortium Resource [[75]34,[76]35,[77]36]. GO
   analyses were applied to determine the functional meaning of the miRNA
   sequencing results by detecting: (1) comprehensive sets of functional
   annotation tools to identify enriched biological processes, (2)
   relations that operate between the identified GO terms, and (3)
   association of miRNAs and genes with specific disease. To investigate
   the signaling pathways mediated by the target genes for each of the
   differentially expressed miRNA, GO using “Biological Process” and
   “Molecular Function” were performed. Only pathways with an FDR-adjusted
   p < 0.05 are represented. Additional KEGG pathway enrichment analysis
   from Database for Annotation, Visualization and Integrated Discovery
   (DAVID) bioinformatics tool was applied to determine molecular pathways
   and biological processes associated with miR-155 target genes.

3. Results

   Brain tissue specimens from five non-consecutive patients with
   malignant hemispheric stroke and three non-stroke patients were used in
   our study. Patients were selected based on the need for stroke
   resection for severe swelling in the opinion of the attending
   neurosurgeon. Detailed information about the stroke and control samples
   is provided in [78]Supplemental Table S1. Representative neuroimaging
   is shown in [79]Figure 1. The figure describes a typical case with
   pre-operative and post-operative images demonstrating the large region
   of stroke ([80]Figure 1A,B) and subsequent stroke tissue removal
   ([81]Figure 1C).

Figure 1.

   [82]Figure 1
   [83]Open in a new tab

   Representative pre-operative and post-operative images from the same
   stroke patient. (A): Axial computed tomography (CT) scan at one day
   post stroke demonstrating left sided middle cerebral artery infarct
   with loss of grey-white matter differentiation in a large territory.
   The white dotted line points out the approximate infarct border. (B):
   Post-operative CT scout image demonstrating the large craniotomy defect
   (dotted line). (C): Post-operative axial CT scan demonstrating
   completed infarct, bony decompression, and the region of partial stroke
   resection in the temporal lobe (arrow).

3.1. Morphology of the Brain Tissue Collected from Stroke Patients

3.1.1. Brain Tissue Damage

   The representative images of the temporal lobe tissue obtained from
   non-stroke control and stroke patient (HE staining) are demonstrated in
   [84]Figure 2 and [85]Supplemental Figure S1. The tissue from non-stroke
   patient cortex did not present any signs of edema or vascular damage,
   and comprised normal neurons with abundant lightly eosinophilic
   cytoplasm and good nuclear detail with prominent nucleoli ([86]Figure
   2A,B). All stroke patient samples were characterized by classic
   hypoxic-ischemic changes including bright eosinophilic cytoplasm and
   lack of the nuclear detail in neurons. The tissue was characterized by
   acute neuronal degeneration and classic appearance of the “eosinophilic
   neurons” (also termed “red” neurons), identified by cell body
   shrinkage, darkly stained pyknotic nuclei, and an intensely stained red
   eosinophilic cytoplasm (black arrows, [87]Figure 2C,D). While all
   pyramidal neurons were severely damaged, some of them had a relatively
   intact chromatin pattern, and others comprised the necrotic “ghost”
   pyramidal neurons with absent affinity for hematoxylin and nuclei (blue
   arrows, [88]Figure 2D). Peri-vascular and peri-neuronal widening was
   consistent with edema (stars, [89]Figure 2D).

Figure 2.

   [90]Figure 2
   [91]Open in a new tab

   Morphology of human stroke brain tissue. Morphological evaluation of
   the temporal lobe tissue, HE staining. (A,B), Control non-stroke
   samples; (C,D), Brain tissue collected at 48 h after stroke onset.
   Black arrows: eosinophilic (red) neurons with darkly stained pyknotic
   nuclei. Blue arrows: necrotic “ghost” neurons with an absent nuclear
   detail. Stars: peri-neuronal and peri-vascular space widening
   associated with edema. Neutrophils (arrowheads) were detected in the
   vicinity of large vessels and capillaries (E,F), tissue collected at 60
   h after stroke onset), as well as in the brain parenchyma (G,H), tissue
   collected at 28 h after stroke onset). (E) and higher magnification in
   (F): Margination and pavementing of neutrophils along the blood vessel
   wall; some neutrophils migrated out and are detected in the blood
   vessel vicinity. (G) and higher magnification in (H): Neutrophils are
   detected at a considerable distance from the blood vessels, which
   demonstrates the process of neutrophil invasion into the damaged brain
   tissue. BV- blood vessel. Bars: (A,C): 50 μm; (B,D,E–H): 20 μm.

3.1.2. Leukocyte Infiltration

   Pathological processes associated with stroke are exacerbated following
   hypoxia-induced blood-brain barrier damage and infiltration of
   different types of leukocytes, including macrophages and neutrophils.
   Recruitment of circulating leukocytes in the ischemic brain contributes
   to neuroinflammation, BBB damage, and further loss of brain tissue
   following stroke [[92]4]. Neutrophils are among the first cells
   infiltrating the brain after ischemic stroke; they are detected in the
   blood vessels within first 7 h and peak at 1 to 3 days after stroke
   offset. Monocyte infiltration is detected within the first 24 h
   post-ischemia, peak at 4 days, and some of these cells persist for
   weeks and acquire features of tissue macrophages [[93]37,[94]38]. We
   did not observe any notable monocyte/macrophage infiltration in the
   obtained samples, which was expected based on the early (17–72 h after
   stroke onset) time point of tissue collection. We, however, detected
   vascular damage and so called neutrophil margination and “pavementing”
   along the endothelium, a process that occurs before the leukocyte
   extravasation ([95]Figure 2E,F). A significant accumulation of
   neutrophils, both in the perivascular space and in the vicinity of the
   blood vessels was observed (arrowheads, [96]Figure 2E,F). Neutrophils
   were also detected in the brain parenchyma, at a considerable distance
   from the blood vessels (arrowheads, [97]Figure 2G,H). This observation
   is important given the controversy around the question of whether
   neutrophils are detectable in the stroke brain parenchyma
   [[98]38,[99]39].

3.1.3. Immunofluorescence Microscopy

   Immunofluorescence staining of the same stroke tissue samples
   demonstrated that small (unipolar and bipolar) cortical neurons
   ([100]Figure 3A) and spindle-like neurons ([101]Figure 3B) retained
   their morphology and cytoarchitecture. Pan-neuronal marker antibody
   staining revealed a sophisticated neuronal network with still intact
   processes and intercellular connections ([102]Supplemental Figure S2).
   Astrocytes were characterized by the low GFAP expression and damaged
   appearance ([103]Figure 3C), while microglia were represented by normal
   looking spindle-shaped cells ([104]Figure 3D).

Figure 3.

   [105]Figure 3
   [106]Open in a new tab

   Immunofluorescence analysis of human stroke brain tissue IF staining of
   human stroke brain tissue collected at 48 h after stroke onset. A, B:
   Pan-neuronal antibody staining (red) visualizing small unipolar and
   bipolar (A) and spindle-shaped (B) neurons. (C): GFAP (green)
   immunostaining for astrocytes. (D): Iba-1 (red) immunostaining for
   microglia. Bars: (A,C,D): 20 μm, and (B): 50 μm.

   To summarize, morphological evaluation demonstrated that human stroke
   samples used in our study represented a severely damaged cortical
   tissue characterized by early ischemic changes. While tissue edema was
   intense and tissue damage was irreversible, the observed morphological
   and cytological characteristics pointed to a transitional stage between
   a severe acute hypoxic injury and tissue necrosis.

3.2. miRNAs Differentially Expressed in Stroke Brain Tissue

   NGS analysis of stroke and non-stroke control human brain tissue
   specimens was performed to assess stroke-related changes in expression
   of 1815 miRNAs. At the first stage of elimination from the list of
   miRNAs dysregulated in stroke tissue, a p < 0.05 cutoff resulted in a
   group of 172 miRNAs with impaired (>1.5 fold increased or decreased)
   expression in stroke patient vs. non-stroke control samples. Of these
   miRNAs, 75 (58 upregulated and 17 downregulated) miRNAs were previously
   reported to be differentially expressed in the whole blood, serum, or
   CSF samples from stroke patients ([107]Supplemental Table S2
   [[108]40,[109]41,[110]42,[111]43,[112]44,[113]45,[114]46,[115]47,[116]4
   8,[117]49,[118]50]). Further FDR adjusted p-value cutoff resulted in a
   list of 34 miRNAs differentially expressed in stroke brain tissue.
   Volcano plot on [119]Figure 4A displays magnitude changes of miRNAs in
   stroke vs. control samples. Principal Component analysis (PCA) was used
   to reduce the dimension of large data sets and explore sample clusters
   arising naturally based on the expression profile ([120]Figure 4B).

Figure 4.

   [121]Figure 4
   [122]Open in a new tab

   Visualization of miRNA expression in stroke and control tissue. (A):
   Volcano plot showing the magnitude of the difference in expression
   values of the samples in Control and Stroke groups. The graph is
   constructed by plotting the −log[10](FDR corrected p-value) on the
   y-axis, and the log[2](fold-change) on the x-axis from the group of 172
   miRNAs with impaired expression in stroke patient vs. control samples.
   Grey circles represent samples with FDR p-value > 0.05, green circles
   represent miRNAs that have a fold-change > 1.5 and FDR p-value > 0.05,
   blue circles represent miRNAs that have an FRD p-value < 0.05 and a
   fold-change > 1.5, and red circles represent miRNAs that have an FDR
   p-value < 0.05 and fold-change >2.0. (B): Principal component analysis
   (PCA) plot for stroke and control samples. The PCA was performed on 34
   miRNAs identified to be differentially expressed between stroke and
   non-stroke control samples using the miRNAs that have the largest
   coefficient of variation based on TMM normalized counts. Each asterisk
   represents a sample. Red asterisk–stroke samples from patients S1–S5;
   black–control non-stroke samples C1–C3. (C): A heat map demonstrating
   expression levels of 50 microRNAs in three different control samples.
   Each column represents one miRNA and each row represents one sample.
   The color represents the difference of the count value to the row mean.
   N = 5 stroke and 3 control samples.

   PCA analysis demonstrated that control samples were segregated into a
   separate cluster, while, with the exception of sample 4, stroke samples
   clustered together. A portion of a heat map on [123]Figure 4, C shows
   the profiles of 50 miRNAs, demonstrating a similarity between the
   control samples from three different non-stroke patients. miRNA
   expressions were not statistically different between control samples.

   Of the detected 34 differentially expressed miRNAs, 19 (17 upregulated
   and 2 downregulated) were identified as stroke-related miRNAs with
   >2-fold changed expression in stroke samples. Based on the available
   literature, the aberrant expression of these miRNAs and their family
   members is associated with stroke progression and outcome ([124]Table
   1). miRNA-based regulation of various cellular and biological processes
   is mediated via their direct target genes and proteins. To explore the
   functional significance of the differentially expressed miRNAs, we
   identified their experimentally verified and predicted target genes and
   subsequently applied GO pathway enrichment analyses, as described in
   Methods. In agreement with previous reports, our analyses detected that
   biological processes significantly influenced by the identified miRNAs
   (including neuronal death, inflammatory response, blood coagulation,
   glutamate and acetylcholine secretion, synaptic plasticity, and
   vascular permeability) are critical for stroke progression and outcome.
   Importantly, the upregulation of miR-1246, miR-4516, miR-320a-3p,
   miR-320c, miR-204-3p, miR-17-5p, miR-16-5p, and miR-423-5p detected in
   stroke brain tissue is in agreement with the reported induction of
   these miRNAs and their family members in stroke patient blood or CSF
   [[125]51,[126]52,[127]53,[128]54,[129]55,[130]56,[131]57,[132]58].

Table 1.

   Differential (>2-fold change, FDR p < 0.05) expression of miRNAs in
   human stroke brain tissue—significantly dysregulated stroke-related
   miRNAs. Red—upregulated and blue—downregulated miRNAs.
   Name Fold-Change p-Value FDR p-Value Reported in:
   hsa-miR-1246
   hsa-miR-4516
   hsa-miR-182-5p
   hsa-miR-320d
   hsa-miR-1255b-5p
   hsa-miR-320c
   hsa-miR-183-5p
   hsa-miR-196b-5p
   hsa-miR-204-3p
   hsa-miR-17-5p
   hsa-miR-193b-5p
   hsa-miR-16-5p
   hsa-miR-320b
   hsa-miR-423-5p
   hsa-miR-320a-3p
   hsa-miR-505-5p
   hsa-miR-652-3p
   hsa-miR-135a-3p
   hsa-miR-196a-5p 334.46
   32.68
   15.44
   10.80
   9.53
   9.29
   8.87
   5.64
   3.23
   3.20
   2.87
   2.78
   2.54
   2.46
   2.41
   2.34
   2.19
   −9.00
   −31.65 0
   0
   3.98 × 10^−6
   5.73 × 10^−11
   0.0001
   2.30 × 10^−11
   0.0001
   0.0011
   0.0008
   0.0005
   0.0004
   0.0002
   0.0017
   0.0015
   0.0008
   0.0014
   0.0015
   0.000536
   3.33 × 10^−16 0
   0
   0.0002
   8.67 × 10^−9
   0.0059
   3.79 × 10^−9
   0.0059
   0.0347
   0.0251
   0.0180
   0.0142
   0.0090
   0.0465
   0.0416
   0.0263
   0.0416
   0.0416
   0.0180
   1.01 × 10^−13 [[133]51,[134]52,[135]53]
   [[136]54]
   [[137]18]
   [[138]18,[139]59]
   [[140]60]
   [[141]51]
   [[142]18]
   [[143]18]
   [[144]57]
   [[145]56]
   [[146]61]
   [[147]58]
   [[148]18]
   [[149]52]
   [[150]55]
   [[151]18]
   [[152]18]
   [[153]57]
   [[154]18]
   [155]Open in a new tab

Newly Detected Group of miRNAs Dysregulated in Stroke Brain Tissue

   NGS analysis of stroke and non-stroke patient samples detected 15
   differentially expressed miRNAs with no previously reported association
   with human stroke ([156]Table 2). The list included 5’ or 3’ strands of
   the mature miRNAs.

Table 2.

   Differential (>2-fold change, FDR < 0.05) expression of newly detected
   miRNAs in human stroke brain tissue. Red—upregulated,
   blue—downregulated miRNAs.
                      Name       Fold-Change p-Value FDR p-Value
   hsa-miR-10395-3p
   hsa-miR-4732-5p
   hsa-miR-7704
   hsa-miR-9901
   hsa-miR-4772-3p
   hsa-miR-412-5p
   hsa-miR-142-3p
   hsa-miR-619-5p
   hsa-miR-1270
   hsa-miR-3615
   hsa-miR-539-3p
   hsa-miR-3912-5p
   hsa-miR-6500-3p
   hsa-miR-5683
   hsa-miR-892a 17.13
   16.35
   15.42
   14.91
   8.35
   6.75
   5.14
   5.08
   2.96
   2.62
   −2.08
   −2.65
   −2.83
   −3.68
   −3.88 1.00 × 10^−7
   0.0002
   1.61 × 10^−10
   0.0001
   0.0005
   1.10 × 10^−7
   0.0001
   0.0001
   0.0008
   0.0005
   0.0012
   0.0001
   0.0002
   0.0007
   0.0019 9.08 × 10^−6
   0.0062
   2.24 × 10^−8
   0.0048
   0.0172
   9.08 × 10^−6
   0.0060
   0.0050
   0.0254
   0.0175
   0.0363
   0.0051
   0.0090
   0.0243
   0.0514
   [157]Open in a new tab

   Bioinformatic analysis using the GO database revealed the association
   of this group of miRNAs with multiple diseases and pathological
   conditions, including coronary artery disease, stroke, amyotrophic
   lateral sclerosis, and Parkinson’s disease. Functional GO pathway
   enrichment analyses detected a significant association of these miRNAs
   with various stroke-related biological processes. As summarized in
   [158]Figure 5, the identified miRNAs are significantly associated with
   biological processes and signaling pathways, which play an important
   role in stroke progression and recovery.

Figure 5.

   [159]Figure 5
   [160]Open in a new tab

   Pathway enrichment analysis for newly detected miRNAs. Diagram showing
   selected groups of GO terms (with the corresponding FDR p-value)
   significantly associated with the identified miRNAs (depicted in the
   box). The annotations describe molecular functions and signaling
   pathways, as well as general and brain-specific biological processes.
   Abbreviations: BMP- bone morphogenetic protein; MAPK—mitogen-activated
   protein kinase; PDGF—platelet-derived growth factor; RIG-I—retinoic
   acid-inducible gene receptor I; TGF-beta—transforming growth factor
   beta.

   [161]Supplemental Figure S3 demonstrates some of the multiple predicted
   and experimentally validated target genes regulated by the identified
   miRNAs. In order to validate the dysregulation of newly detected
   miRNAs, we assessed the expression of their known direct target genes
   in stroke and control brain tissue. [162]Table 3 presents the list of
   the miRNAs and their target genes; 20 genes with the greatest up- or
   downregulation in stroke vs. control tissue are displayed. As expected,
   microRNAs and their target genes had the opposite expression pattern; a
   number of genes were targeted by several newly detected microRNAs
   ([163]Table 3, stars). We were unable to verify the expression of
   target genes for miR-10395-3p, miR-9901, and miR-412, due to a very
   limited number of their known predicted targets. Among the deregulated
   target genes were: SMAD genes implicated in regulating inflammation,
   myelination, and vascular function after stroke [[164]62]; WNT1, the
   regulator of Wnt signaling pathway mediating neurogenesis and vascular
   remodeling [[165]63]; MAPK81P2, a critical activator of
   mitogen-activated protein kinase (MAPK) pathway contributing to
   neuroinflammation and neuronal survival after stroke [[166]64]; NOTCH2
   influencing neuronal viability [[167]65]; PDGFRA, regulating
   platelet-derived growth factor signaling contributing to post-stroke
   atherosclerosis [[168]66]; ADAMTS5, a member of matrix
   metalloproteinase family, regulating degradation of extracellular
   matrix proteins, tissue repair, and remodeling [[169]67]; FGF12, a
   member of fibroblast growth factor (FGF) family implicated in
   neuroprotective effect after stroke [[170]68]; and NOS1 encoding
   neuronal nitric oxide synthase 1, associated with stroke-induced
   neurotoxicity [[171]69].

Table 3.

   Differentially (>2-fold change, FDR p < 0.05) expressed miRNA target
   genes in human stroke brain tissue. Red—upregulated and
   blue—downregulated miRNAs and genes. Target genes shared by at least 2
   different miRNAs are marked with a star.
   miRNA Differentially Expressed Target Genes
   hsa-miR-4732-5p UHMK1, TOX *, TMEM183A *, TMED8, TGOLN2, SMAD2, SAMD12,
   RORB, RAP2A, RALYL*, PURG *, PTCH1 *, PSD3 *, PRPF40A, PKIA *, PHKA1 *,
   PDE4D, PCDH17, PARM1, PAPOLA *
   hsa-miR-7704 SPATA17, LDLRAP1, IWS1, DERL3, SEMA6A *, CACNA1E, ATG12,
   OPA1, PAK3, EPS8, CHMP3, POLR2L, GPR85, C4orf33, LETMD1, MEIS2 *,
   SPATA6, KLF12 *, ROBO1, PTCH1 *
   hsa-miR-4772-3p CTDSP2, TMEM183A *, PURA *, RHOU, CSRNP3, ARHGAP31 *,
   SH3BGRL2, CADM2 *, LRRC28 *, METTL15 *, PHKA1 *, TAOK1 *, GUCY1A2,
   MLLT3 *, CHIC1 *, RNF141 *, TOMM20 *, ZDHHC21 *, HS6ST3 *, PURG *
   hsa-miR-142-3p PDGFRA, TSPAN3, MTF2, FOXP1, REPS2 *, MTX3, MEIS2 *,
   PIK3CA, MAP2K6, CLSTN2, HMGCLL1 *, FGF12 *, SNTB2 *, NUCKS1, PURA *,
   CADM2 *, MAGI2, CHIC1 *, LRP1B, TOX *
   hsa-miR-619-5p EIF4EBP2, CADM2 *, LRRC28 *, METTL15 *, ATRNL1, NKAIN2,
   TAOK1 *, MLLT3 *, GALC, CHIC1 *, KDM3B, AMBRA1, IL1RAPL1, RNF141 *,
   TOMM20 *, RPS6KA6, ZDHHC21 *, KCNH5, HS6ST3 *, PURG *
   hsa-miR-1270 LRRTM4, THADA, PCDH9, KLF12 *, STX17, MMP16, HMGCLL1 *,
   CNOT7, PURB, RALYL *, PTAR1, PKIA *, NOVA1, OLFM3, NUBPL, FGF12 *,
   SNTB2 *, PAPOLA *, DPP6 *, CNR1
   hsa-miR-3615 REPS2 *, SEMA6A *, ARHGAP15, MAMLD1, FOXB1, CCDC80, MEIS2
   *, SLITRK5, PSD3 *, CBX5, DIRAS2, PKIA *, BTBD3, DPP6 *, ARHGAP31 *,
   TAOK1 *
   hsa-miR-539-3p AQP7 *, ADAMTS5, HAVCR1 *, GOLGA8G, ZNF266, TMCO5A,
   NOTCH2 *, PLCL1 *, HCN4, RYR2, PHKB *, ABCA13, LYN *, C18orf25, WIPF2,
   GATA3 *, FBLN2, TMPRSS4 *, TNFAIP3, ZC3H4 *
   hsa-miR-3912-5p FZD8 *, WNT1, PID1 *, ADAM8, GATA3 *, KLHDC7B, JAG2,
   DLST, RBMS3, FIG4, TMEM63A, IQSEC3
   hsa-miR-6500-3p FZD8 *, HIC1, AQP7 *, MRGPRF, DMRT1, CTCFL, HSPA1A,
   UNC45B, SLCO5A1, SBSPON, FOXJ3, HAVCR1 *, PID1 *, NR5A1, COX20,
   NOXRED1, SCNN1G *, UBAP1, PLCL1 *, LYN *
   hsa-miR-5683 LSM4, LOXL4, PPM1N, ARSI, GATA3 *, NUTM2D *, TMOD4, NLRP2,
   ZNF217, NGB, HOXD3, CYTH3, DACT3, CTPS2, MOB2, CD74, CLDN10, GALNT1,
   FZD8 *, MAPK8IP2
   hsa-miR-892a NOTCH2 *, SCNN1G *, ACTG2, PHKB *, LYN *, ZBTB42, CASC4,
   NUTM2D *, FTCDNL1, IRX2, SLC16A3, RCL1, TMPRSS4 *, DPEP1, VSIG10, ZC3H4
   *, SAP30BP, POTEG, NOS1, IRAK3
   [172]Open in a new tab

   To summarize, based on the performed bioinformatic and RNA-sequencing
   analyses, we concluded that the aberrant expression of the newly
   detected miRNAs has a significant functional implication and thus,
   these miRNAs could prominently influence the progression and outcome of
   human stroke.

3.3. miR-155 Expression in Stroke Brain Tissue

   Due to our long-standing interest in the role of miR-155 in human
   stroke, we performed a separate investigation of its deregulation in
   stroke brain tissue. miR-155 is an evolutionarily conserved
   multifunctional miRNA, expressed in all human tissues and implicated in
   contributing to various pathological processes [[173]70,[174]71]. A
   number of our previous studies have been focused on the role of this
   miRNA in ischemic stroke. In a rodent model of stroke, an elevated
   miR-155 expression in the brain was linked to higher stroke severity,
   while miR-155 inhibition reduced the infarct size, preserved BBB
   integrity, and supported the functional recovery [[175]72,[176]73]. In
   the present study, miR-155 was significantly (>1.5-fold, p = 0.04)
   elevated in human stroke tissue ([177]Supplemental Table S2). However,
   due to a high variability among five stroke samples, an average miR-155
   transcript expression change did not reach the significance cutoff
   after the FDR p < 0.05 adjustment. Interestingly, miR-155 levels tended
   to be higher in samples S1, S2 and S5 collected within 48–72 h after
   stroke onset, and lower in samples S3 and S4 collected within 17–28 h
   after stroke. In order to evaluate stroke-related changes of miR-155
   activity, we performed miR-155 target gene profiling of samples S1, S2
   and S3 ([178]Figure 6). miR-155 Targets PCR Array analysis detected a
   significant (>1.5-fold, FDR p < 0.05) downregulation of 50 miR-155
   direct target genes in stroke samples compared to controls ([179]Figure
   6A,C, blue). Interestingly, a significant decrease in target gene
   expression was also detected in the S3 patient sample characterized by
   a moderate increase in miR-155 ([180]Figure 6B). Based on the whole
   transcriptome data, additional 23 miR-155 target genes (not included in
   the PCR Array panel) were significantly downregulated in stroke samples
   ([181]Figure 6A, green). GO and KEGG pathway enrichment analyses
   revealed that the detected group of genes is involved in the regulation
   of critical signaling pathways, including TGF-β, ERK, NF-kB, Wnt, MAPK,
   ErbB, Ras, PI3K-Akt, and VEGF signaling cascades. Moreover,
   bioinformatic analysis identified several stroke-related and
   brain-specific biological processes regulated by this set of genes,
   including glial cell differentiation, endothelial morphogenesis,
   leukocyte activation, macrophage cytokine production,
   neurodegeneration, and synaptic function. Raw data demonstrating
   miR-155 direct target expression are uploaded in GEO database. Based on
   the obtained results, we concluded that: (1) even moderate miR-155
   expression changes induce a significant downregulation of its direct
   targets, and (2) 73 detected genes differentially expressed in stroke
   tissue could be regulated on the RNA level via miR-155-induced mRNA
   degradation.

Figure 6.

   [182]Figure 6
   [183]Open in a new tab

   miR-155 target gene analysis. (A) A list of miR-155 direct target genes
   significantly downregulated in stroke tissue, compared to control
   samples. Blue—genes identified by PCR Array, green—additional genes
   identified by whole transcriptome sequencing. (B) A portion of a
   heatmap (after non-hierarchical clustering) displays a group of target
   genes with significantly lower (green) expression in the stroke
   samples. The heat map was generated based on delta Ct values. Genes
   with higher expression levels are shown in red, whereas genes with
   lower expression levels are shown in green. Genes with average
   expression levels are shown in black. (C) Scatter plot compares the
   normalized gene expression between the stroke and control samples. The
   central line indicates unchanged gene expression. Red circles identify
   genes significantly upregulated in stroke tissue, blue—genes
   significantly downregulated in stroke group. The graph plots the log10
   of normalized gene expression levels in control samples (x-axis) versus
   stroke samples (y-axis). N = 3 samples per control and stroke groups.

4. Discussion

   The underlying mechanisms of malignant cerebral infarction include
   general cellular and molecular processes associated with all types of
   strokes. Besides the known stroke-induced pathophysiological cascades
   leading to brain tissue damage and BBB disruption, additional
   contributing factors to malignant infarction and brain swelling include
   the involvement of the autonomic nervous system [[184]24,[185]74], and
   cortical spreading depolarization [[186]22,[187]75,[188]76,[189]77].
   Based on the morphological evaluation of freshly removed stroke tissue,
   we propose that miRNA sequencing results obtained in our study could
   reflect molecular mechanisms attributed to (1) stroke-associated brain
   edema and tissue damage, and (2) transition from ischemic damage and
   selective neuronal death to tissue necrosis and infarction. We propose
   that our findings contribute to better understanding of these
   concomitant processes occurring at the early stages of stroke.

   The main limitation of our study is the small number of analyzed
   samples. The reason for this is that DC with stroke-ectomy is a very
   rare procedure that is restricted to a select group of patients. To
   minimize the data variability, we analyzed only the temporal lobe
   samples from patients with an MCA/ICA territory infarction. This
   further reduced the number of specimens included in the analysis. To
   control false discovery rate, greater than 2-fold change and FDR p <
   0.05 cutoff were applied to all miRNA and mRNA sequencing data reported
   in [190]Table 1, [191]Table 2, and [192]Table 3. We believe that low
   FDR p-values, which often ranged between 1.61 × 10^−10 and 0.01, are
   the indicators of the strength and reliability of our study.

4.1. Dysregulation of miRNAs with Known Association with Stroke

   Our investigation established a strikingly similar expression pattern
   of 8 miRNAs detected in our brain tissue samples and blood/CSF samples
   reported by other groups. Therefore, we propose that the impaired
   expression of these miRNAs is associated not only with malignant
   infarction, but with cerebral stroke in general. Based on our analysis
   and reports by other groups, miR-1246, miR-4516, miR-320a-3p and
   miR-320c, miR-204-3p, miR-17-5p, miR-16-5p, and miR-423-5p, were
   upregulated both in stroke brain tissue and patient blood/CSF. Thus,
   signatures of these miRNAs in the blood or CSF samples accurately
   reflect their dynamics in the brain tissue. We, therefore, propose that
   these miRNAs could be regarded as potential candidate biomarkers for
   stroke prognosis and outcome. This statement is supported by the data
   from previous studies demonstrating the functional significance of
   several miRNAs from the identified group. miR-1246 is dramatically
   increased in CSF of patients with larger infarction [[193]53].
   Upregulation of miR-320 family miRNAs is associated with the modulation
   of aquaporin family proteins and stroke-induced edema
   [[194]55,[195]78,[196]79]. Increased plasma miR-16 is associated with
   large infarction [[197]58]. Based on these reports and our data, we
   propose that profiling of these microRNAs in the blood or CSF could
   provide a timely and valuable information for choosing a potentially
   successful therapeutic approach.

   Among the identified significantly dysregulated miRNAs, miR-182-5p,
   miR-183-5p, miR-320b and d, miR-505-5p, and miR-196a-5p had an opposite
   shift in expression levels in our stroke brain tissue samples and blood
   samples detected by other groups. These discrepancies reflect the
   dynamics of miRNA exchange between the brain tissue and circulating
   blood, and could be associated with a different sample collection time.
   This interesting observation will be addressed in our future studies
   where miRNA profiling will be performed in the brain tissue and blood
   samples from the same patient. Due to a long-standing interest in
   miR-155, we performed a separate investigation of its activity in human
   stroke samples. Our findings revealed that, despite the moderate
   increase in miR-155 expression in some samples, activity of this miRNA
   resulted in a significant downregulation of 72 direct target genes.
   Based on bioinformatics analysis, 30 of the detected genes represent
   the shared targets between miR-155 and at least one miRNA from the list
   provided on [198]Table 1. Therefore, further investigation is needed to
   properly evaluate miR-155 activity in the stroke brain tissue.
   Dysregulation of miR-155 and its possible correlation with the time of
   stroke onset will be a subject of our upcoming study involving a larger
   group of stroke patients.

4.2. Newly Detected miRNAs Dysregulated in Human Stroke Tissue

   A group of miRNAs differentially expressed in stroke brain tissue
   included 15 significantly dysregulated miRNAs with no previously
   reported changes in stroke patient blood or CSF. Most of them are the
   recently identified miRNAs with yet unexplored functions. Based on our
   bioinformatics approach, they can regulate molecular processes that are
   critical for stroke progression. We suggest that dysregulation of these
   newly detected miRNAs could reflect: (1) brain tissue-specific changes,
   and (2) miRNA signatures specific to malignant infarction and brain
   swelling. The role of this novel group of miRNAs will be further
   investigated in the functional studies utilizing in vitro or animal
   models of stroke.

5. Conclusions

   In this first analysis of human stroke brain tissue biopsy samples, we
   described morphology and cytoarchitecture of the temporal lobe cortex
   at 17–72 h after the hemispheric infarction. microRNA profiling
   detected 34 miRNAs aberrantly expressed in stroke-tissue, which could
   influence stroke progression and outcome. We anticipate that the
   obtained data will contribute to knowledge of the molecular basis of
   stroke, and provide a foundation for the miRNA-based intervention
   strategies focused on stroke outcome and prevention of malignant brain
   swelling.

Acknowledgments