Discovery of new MicroRNAs and their mRNA targets in patients with acute ischemic stroke

TL;DR

This study identifies new microRNAs and their target genes in patients with acute ischemic stroke, revealing potential biomarkers and pathways involved in stroke-related molecular responses.

Contribution

The study discovers novel miRNAs and their mRNA targets in acute ischemic stroke patients, offering new insights into stroke biomarkers and molecular mechanisms.

Findings

146 upregulated and 258 downregulated RNAs were identified in stroke patients.

miR-199a-5p showed the highest diagnostic potential with an AUC of 0.89.

miR-4467 and GNAI2 mRNA demonstrated dynamic changes during hospitalization, indicating post-stroke molecular adaptations.

Abstract

In this study, we applied microarray, bioinformatics, and qRT-PCR techniques to identify miRNAs and their target genes in plasma obtained from acute ischemic stroke patients and matching controls.

Microarray analyses were performed with 24-h acute ischemic stroke vs. healthy individuals and CV-risk factors matched control group plasma samples. Statistical analysis of gene expression was performed using TAC and R, with a focus on robust methods suitable for the small sample size, and miRNA target prediction was conducted using a previously established in-house wizbionet R package. Top non-coding regulators of ischemia (miR-18a-5p, miR-4467, miR-199a-5p and miR-3135b) and their predicted target genes (ANKRD12, HIF1A, GNAI2, GRIN1) were detected via qRT-PCR.

146 upregulated and 258 downregulated differentially expressed RNAs were detected by microarray analysis. Using the multiMiR R package for target prediction, 67 upregulated and 125 downregulated mRNAs were mapped. Functional enrichment analysis revealed that upregulated miRNAs were associated with pathways like BDNF and IL-2 signaling, while downregulated miRNAs were linked to neurodevelopmental and NGF pathways. MiR-18a-5p and miR-199a-5p were significantly elevated in stroke patients at both day 1 and day 7 compared to healthy individuals and CV-matched controls (p < 0.05 for all). miR-4467 was lower at day 1 versus both controls (p < 0.05) but markedly increased by day 7 (p < 0.001), remaining higher than in controls (p < 0.05). miR-3135b showed persistent downregulation at both time points (p < 0.05). ROC analysis confirmed diagnostic value for all miRNAs, with the highest AUC for miR-199a-5p (0.89, 95% CI: 0.81–0.97, p < 0.001), followed by miR-3135b (0.88), miR-4467 (0.80), and miR-18a-5p (0.73; p = 0.002). For potential role in dynamic post-stroke molecular responses perspective, utility, miR-4467 increased significantly during hospitalization (p < 0.001). GNAI2 mRNA was significantly elevated at day 1 versus controls (p < 0.05). ANKRD12 was consistently lower at both day 1 and day 7 compared to controls (p < 0.05). GRIN1 was reduced at admission (p = 0.006) but normalized by day 7 (p > 0.05). ROC analysis showed diagnostic significance for ANKRD12, GNAI2, and GRIN1 (AUC = 0.683, 0.698, 0.693).

Our integrated miRNA/mRNA analysis identified distinct molecular signatures in acute ischemic stroke, with 146 upregulated and 258 downregulated RNAs, implicating key neuroinflammatory and neuroprotective pathways, including BDNF, IL-2, and NGF signaling. Among the validated candidates, miR-199a-5p, miR-3135b, miR-4467, and miR-18a-5p demonstrated diagnostic potential, while miR-4467, together with GNAI2 and HIF1A, showed post-stroke dynamic relevance, reflecting early transcriptomic adaptations following ischemic injury.

The online version contains supplementary material available at 10.1186/s12920-025-02302-5.

Figures

Volcano plot with the results of the differential expression analysis of stroke vs. CV-matched control group samples

Workflow of the bioinformatic analysis. DE- differentially expressed

Top 10 enriched terms associated with the targets of at least 3 up- or down-regulated top miRNAs. A Bioplanet signaling pathways, (B) dbGaP phenotypes, (C) ARCHS4 Tissues, (D) Cell Markers. Y axis- Top 15 enriched terms for up- and down-regulated top miRNAs. X axis-genes targeted by the top differentially expressed miRNAs with the same regulation directionality. Numbers on the right sides of the circles symbolizing gene number are related to ranks of the processes; the lower number, the more significant process. Numbers on the left side of the circles symbolize gene number. The figure also sho

Top gene networks affected by ischemic stroke-related miRNAs

MiRNAs and predicted genes expression difference between groups. a miR-18a-5p; b) ROC analysis of miR-18a-5p baseline; c) miR-4467; d) ROC analysis of miR-4467 baseline; e) miR-199a-5p; f) ROC analysis of miR-199a-5p baseline; g) miR-3135b; h) ROC analysis of miR-3135b baseline; i) ANKRD12; j) ROC analysis of ANKRD12 baseline; k) HIF1A; l) ROC analysis of HIF1A baseline; m) GNAI2; n) ROC analysis of GNAI2 baseline; o) GRIN1; p) ROC analysis of GRIN1 baseline; *p* value was calculated with Mann-Whitney U or Wilcoxon test, appropriately. Four-group comparison was calculated by the Kruskal-Wallis

Predicted enrichment panel of miR-3135b, miR-4467, miR-18a-5p, and miR-199a-5p for pathways (A) and tissues (B). Numbers associated with circles symbolizing gene number are related to the ranks of the processes; the lower the number, the more significant the process. The figure also shows whether a given process is significant after p-value adjustment (blue border), significant only before adjustment (grey border), or not significant (yellow border). In addition, a visual legend indicates the number of associated genes using both color and node size

Predicted enrichment panel of miR-3135b, miR-4467, miR-18a-5p, and miR-199a-5p for phenotypes (A) and cell types (B). Numbers associated with circles symbolizing gene number are related to the ranks of the processes; the lower the number, the more significant process. The figure also shows whether a given process is significant after p-value adjustment (blue border), significant only before adjustment (grey border), or not significant (yellow border). In addition, a visual legend indicates the number of associated genes using both color and node size

Full text

Introduction

Diseases of civilization are a major concern in modern medicine, as leading causes of death worldwide. As reported by the World Stroke Organization (WSO), ischemic stroke continues to be the second most common cause of mortality and the third-leading cause of combined death and disability globally [1]. Patients with acute stroke need a quick diagnosis and a timely treatment, within 3–6 h from the onset of symptoms. Longer reaction times substantially limit the effectiveness of treatment and chances of a full recovery. The stroke diagnosis starts with the recognition of stroke warning signs; however, it is not clear if a patient has a stroke or other neurological diseases that mimic stroke, including brain mass lesions (tumor, abscess, hemorrhage) [2]. Current diagnostic methods rely on radiologic brain imaging. Standard computer tomography (CT), magnetic resonance imaging (MRI), and transcranial Doppler imaging can be used. Unfortunately, most patients do not arrive at the hospital in time, or the radiological diagnosis is delayed, precluding them from accessing effective reperfusion treatments [3]. For this reason, faster and effective diagnostic methods are still needed. MicroRNAs (miRNAs) are a class of small non-coding molecules approximately 21 nucleotides in length that primarily modulate gene expression at the post-transcriptional phase. Recent research has highlighted their potential in diagnosing various diseases, including their role as diagnostic and prognostic biomarkers for ischemic stroke, supported by findings from in vitro, animal, and human studies. Advances in RNA-sequencing, along with robust statistical and experimental methods, are increasingly enabling the precise distinguishing of disease-related genetic variants and genes, offering promising tools for improved management of stroke patients [4–7]. In addition, the detection of circulating miRNAs may be easily carried out in human plasma samples collected from ischemic stroke patients, with the potential to speed up the diagnosis, thus allowing a larger proportion of patients to qualify for the most effective therapies available thus far, namely, thrombolysis and thrombectomy, within the narrow therapeutic window.

Despite growing evidence for the role of circulating miRNAs in ischemic stroke, the diagnostic utility and temporal dynamics of specific miRNA/mRNA networks in the acute phase remain insufficiently understood. Most previous studies have focused on limited candidate miRNAs without detecting their target genes or assessing their changes over time. The present study addresses this gap by integrating plasma miRNA profiling with bioinformatic target prediction and qRT-PCR validation to identify novel regulators that may serve as early diagnostic biomarkers and mechanistic links in acute ischemic stroke pathophysiology. Specifically, we used microarray technology to detect differentially expressed non-coding RNAs in plasma samples collected from patients during the acute phase of ischemic stroke and compared them with age-, sex-, and cardiovascular risk factor-matched controls and healthy individuals. In addition to the microarray analysis, we validated the top miRNAs and their predicted target mRNAs using qRT-PCR.

Methodology

Participants

The study cohort included 28 individuals diagnosed with acute ischemic stroke, classified based on clinical features consistent with the WHO’s criteria and confirmed through brain imaging (CT or MRI) [8]. The study followed the TOAST (Trial of Org 10172 in Acute Stroke Treatment) classification criteria for patient selection. We included all cases of ischemic stroke caused by large-vessel atherosclerosis, as well as a subset of undetermined etiology cases that met specific criteria: presence of ≥ 50% carotid artery stenosis ipsilateral to the cerebral infarction and absence of atrial fibrillation. During hospitalization, all patients underwent Duplex Doppler imaging of extracranial arteries and 24-hour Holter ECG monitoring [9–11]. Patients with a peripheral arterial occlusive disease or cancer were excluded.

The control group comprised 35 patients matched for age and gender, who had no neurodegenerative disease history, prior history of stroke, or transient ischemic attack (TIA). 46% of these patients had confirmed stable coronary artery disease (CAD) alongside concurrent cardiovascular risk factors. The study protocol and informed consent form were developed in accordance with the Declaration of Helsinki and received approval from the Ethics Committee of the Medical University of Warsaw, Warsaw, Poland (KB/148/2017; KB/112/2016). Written informed consent was obtained from all participants or their legal representatives. All blood collections were performed between September 2019 and June 2020 at the Institute of Psychiatry and Neurology and the Banacha Hospital of the Medical University of Warsaw.

Blood sample collection and handling for patients and control group

Blood samples were obtained using citrate anticoagulant tubes (S-Monovette, Sarstedt) from patients with acute ischemic stroke at two different intervals: first within 24 h of stroke symptoms, and second after 7 days of hospital admission. Healthy individuals and CV-risk factors matched control group provided a single blood sample. All samples underwent immediate plasma separation (within 20 min of collection) by centrifugation, after which the plasma was aliquoted into specialized low-DNA-binding Eppendorf tubes and preserved at -80 °C for subsequent laboratory investigations.

Small RNA extraction and quality control

Total RNA was extracted from all 130 plasma samples (including healthy individuals, CV-risk factors matched controls, and stroke patients with dual sampling at 24 h and 7-days) using the miRVana miRNA isolation (Invitrogen) kit with phenol following the manufacturer’s instructions. RNA quality assessment was performed through absorbance measurements using both the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and NanoDrop 1000 spectrophotometer (Thermo Scientific, Wilmington, DE).

Microarray analysis

Microarray assays

Microarray analyses were performed only with 24-h acute ischemic stroke (n = 5) vs. control samples (n = 4). In total, 9 samples were analyzed. However, 1 sample in the stroke group was excluded as it did not pass the pre-specified threshold for quality control (Supplementary 1, "Participants" section). Small RNA expression profiling was performed using the Human Transcriptome Array 2.0 ST platform (Affymetrix Inc., Santa Clara, CA, USA). The raw microarray data (CEL files) generated by the Affymetrix system were subsequently processed and analyzed using the Affymetrix Expression Console software. FlashTag™ Biotin HSR RNA Labeling Kits (Applied Biosystems™) were used to prepare target for GeneChip™ miRNA 4.0 Arrays. 100 ng of total RNA sample was adjusted to 8 µL of volume and used for poly (A) tailing reaction according to the manufacturer’s instructions. FlashTag™ Biotin HSR Ligation was then performed with the use of T4 DNA Ligase at room temperature incubation for 30 min. All the prepared mixes were used to hybridize miRNA 4.0 GeneChip™ arrays at 48 °C and 60 RPM rotation for 16 h in the Hybridization Oven 645 (Affymetrix, Santa Clara, CA, USA). Wash and stain of arrays were performed with Fluidics Station 450 using the fluidics script appropriate for the Fluidics Station 450 (Protocol FS450_0002 according to Affymetrix instructions for 100 format array). Array scanning was performed following GeneChip Command ConsoleTM v6.1 user guide (Applied Biosystems™, P/N 702569). Expression data were extracted from CEL files by Transcriptome Analysis Console v4.0. Initial bioinformatics processing was performed with GeneChip^®^ Operating Software (GCOS) utilizing the Robust Multi-array Average (RMA) normalization method. The complete gene expression dataset generated in this investigation, including comprehensive metadata regarding experimental procedures and computational analysis parameters, is publicly available in the Gene Expression Omnibus (GEO) repository under accession number GSE255070.

Microarray data normalization and statistical analysis

Microarray data underwent statistical analysis utilizing TAC software. Further analyses were conducted in R, utilizing signal information obtained from the TAC output (specifically, columns containing extension rma-dabg.chp Signal). Tests were performed using log2-transformed data across all comparison groups (A-B). The TAC software was used for statistical analysis with criteria including gene-level fold change thresholds and FDR-corrected p values, while R was employed for additional statistical tests such as logistic regression, Mann-Whitney test, and co-expression analysis, with a focus on Mann-Whitney test results due to the limited sample size and the robustness of TAC’s Tukey-biweight averaging method Additional statistical analysis parameters and methods, as well as in silico analysis models, are described in detail in the supplement (Supplementary 1, "Blood sample collection and handling for patients and control group" section).

Bioinformatic analysis

Gene list selection

We screened the DisGeNet database for genes associated with ischemic stroke [12]. We identified 1450 genes, 145 of which were mentioned in at least 10 articles. Genes were selected based on a combined ranking obtained using the OneR algorithm and interaction network analysis. The application of the machine-learning–based OneR algorithm enabled prioritization of target genes according to the number of regulatory differentially expressed (DE) miRNAs. Six numerical lists of DE miRNAs were analyzed: DE_miR_COUNT, DE_miR_up_COUNT, DE_miR_down_COUNT, DE_Top_miR_COUNT, DE_Top_miR_up_COUNT, and DE_Top_miR_down_COUNT. Each list was divided into four clusters, and top target genes were selected based on their consistent assignment to cluster 1 across the miRNA lists. Among the identified genes, ANKRD12 ranked highest for up-regulated top miRNAs. It was followed by HIF1A, an ischemic stroke–related gene that showed the strongest overall ranking among up-regulated top miRNA lists, and GRIN1, which ranked highest among down-regulated top miRNA lists. GNAI2, identified as the top regulatory gene for down-regulated top miRNAs, ranked between these genes. Network analysis further highlighted HIF1A (three interactors, two of which are stroke-related) and GRIN1 (four interactors) as key nodes associated with ischemic stroke, supporting their biological relevance.

MiRNA’s target predictions and ranking

In order to identify the targets of differentially expressed miRNAs, we used our wizbionet R package and previously developed pipelines [13]. We conducted targeted screening utilizing the multimiR package, opting for the top 20% predictions from all accessible databases [14]. Our analysis encompassed targets for all mature versions, and in cases where a differentially expressed miRNA was identified as a precursor miRNA (pre-miRNA), we generated both − 3p and − 5p versions. Subsequently, upon identifying the targets of differentially expressed miRNAs, we carried out data aggregation and summarization, obtaining insights into the total number of targets affected by each DE miRNA, including those associated specifically with ischemic stroke. Gene-miRNA interactions were aggregated using our wizbionet R package [15]. In effect, we obtained information on how many miRNAs from each list (all, up, down-regulated) of interest are regulating given gene and stacked the list of those miRNAs. In the next step, we used our R function from wizbionet based on a machine learning algorithm, OneR, which divided each numerical column with the number of targets into four clusters [13]. Further, we counted the occurrence of the top 1 clusters for each gene for each given miRNA list and sorted them from highest to lowest. Top genes were selected based on the number of targets from each miRNA list assigned to cluster 1.

Interaction network construction

Interaction network between top targets of DE miRNAs was constructed using targeted genes list and human interactome data from StringApp for Cytoscape [16, 17]. On the network were mapped associations with ischemic stroke gene list, number of all DE miRNAs, up and downregulated.

Enrichment analysis

Enrichment analysis of differentially expressed miR targets was done with the EnrichR database API plugin [18]. This analysis employed Fisher’s exact test with the Benjamini-Hochberg correction, using a precomputed background as the reference for each term in each gene set library. EnrichR provides analytical access to more than 190 ontological databases [19]. In all analyses, the Benjamini-Hochberg corrected p value cutoff was set as lower than 0.05. Visualization of the top ontological results was performed in R using ggplot2 and ggrepel libraries.

Reverse-transcription and detection of miRNAs/mRNAs expression level

Small RNA was subsequently reverse-transcribed using the TaqMan miRNA Reverse Transcription Kit (Applied Biosystems) according to the protocol. MiRNAs as well as mRNA expression levels were quantified via quantitative real-time PCR (qRT-PCR) using TaqMan miRNA/mRNA Assay kits on a CFX384 Touch Real-Time PCR Detection System (BioRad Inc., Hercules, CA, USA). Additionally, all samples were normalized to the synthetic spike-in cel-miR-39, which was added in equal amounts during RNA isolation. In addition, one randomly selected sample was used as an internal reference across all qRT-PCR runs [20, 21]. GAPDH was used for mRNA expression analysis as a normalizer. All reactions were performed in triplicate, with miRNA expression calculated using the 2-ΔΔCT method by using exogenous and endogenous normalizers [22, 23].

Used chemicals, reagents, and plastics

(i) Consumables: S-Monovette Citrate #05.1165 blood collection tubes; Eppendorf™ DNA LoBind™ Tubes #13-698-791; (ii) RNA extraction kits: mirVana™ miRNA Isolation Kit, with phenol #AM1560, Invitrogen; Cel-miR-39-3p mimic: MC10956, catalog #4,464,066; (iii) small RNA quality control with Bioanalyzer: Agilent, High Sensitivity RNA ScreenTape Sample Buffer #5067–5580; Agilent, High Sensitivity RNA ScreenTape # 5067–5579; Agilent, High Sensitivity RNA ScreenTape Ladder #5067–5581; Agilent Small RNA kit #5067 − 1548 (iv) The GeneChip™ Human Transcriptome Array 2.0, #902,162; FlashTag™ Biotin HSR RNA Labeling Kits, Applied Biosystems™, #901,910; GeneChip™ miRNA 4.0 Array, Applied Biosystems™ #902,411. (v) cDNA synthesis: TaqMan™ Advanced miRNA cDNA Synthesis Kit #A28007. (vi) RT-qPCR primers: 478551_mir, miR-18a-5p; 478902_mir, miR-4467; 478231_mir, miR-199a-5p; 478011_mir, miR-3135b; 478640_mir, miR-1226; 478018_mir, miR-3185; 478293_mir, Cel-miR-39-3p; Hs01074529_m1, ANKRD12; Hs00153153_m1, HIF1A, Hs01064686_m1, GNAI2; Hs00609557_m1; GRIN1; Hs05003736_s1, AGO1; Hs02786624_g1, GAPDH - Applied Biosystems; (vii) PCR reaction: TaqMan™ Fast Advanced Master Mix for qPCR #4,444,558.

Statistical analysis and results illustration

Demographic and categorical clinical variables were expressed as proportions of patients and analyzed using χ² tests. Normality testing was performed using the Shapiro-Wilk test. Continuous variables were reported either as mean ± standard deviation or median (interquartile range), with between-group comparisons conducted using either Student’s t-test or Mann-Whitney U test as appropriate. Longitudinal comparisons (day 1 vs. day 7 post-stroke) employed paired t-tests or Wilcoxon signed-rank tests based on distribution characteristics. For biomarker evaluation, we generated receiver operating characteristic (ROC) curves to assess the diagnostic potential of miRNA expression. All miRNA expression values were log10-transformed prior to statistical analysis. We considered p values < 0.05 statistically significant for all two-tailed tests. Three-group comparisons were performed using either one-way ANOVA (with Tukey’s HSD post hoc test) or Kruskal-Wallis test, determined by data distribution. For multiple biomarker comparisons, FDR correction was used. All analyses were conducted in SPSS Statistics version 22.0 (IBM Corp., Chicago, IL, USA). Graphs were refined using Adobe Illustrator version 24.0.2, and the graphical abstract was created with BioRender (biorender.com). ChatGPT and Grammarly were used to assist in improving grammar and correcting typographical errors.

Results

Ischemic stroke patients and control group demographics

The characteristics of participants and patient demographics are summarized in Table 1. Common cardiovascular risk factors included hypertension (64%) and current smoking (39%), which were prevalent in most patients. The demographic is presented in Table 1. The stroke group had a higher prevalence of atrial fibrillation (p = 0.023), prior TIA (p = 0.001), and prior ischemic stroke (p < 0.001) compared to the control group. Laboratory findings revealed elevated WBC (p = 0.008) and hs-CRP (p = 0.038) in patients with stroke.

Table 1. Participants characteristicsHealthy individual(n = 28)CV-risk factor matching control(n = 46)Acute ischemic stroke(n = 28)PvalueHealthy individuals group compared to the acute ischemic stroke groupP valueCV-risk factor matching group compared to the acute ischemic stroke groupBaseline characteristics Gender (male, n, %)15 (54%)22 (58%)15 (53.6%)0.5840.632 Age (years)65.29 ± 4.1165.00 ± 9.2166.39 ± 15.920.7230.634 BMI kg/m^2^26.14 ± 3.3925.81 ± 3.0724.87 ± 3.730.1370.244Comorbidities Hypertension18 (64%)33 (72%)18 (64%)0.0610.502 Heart Failure2 (7%)13 (28%)3 (10.7%)0.6860.639 Atrial fibrillation0 (0%)0 (0%)3 (10.7%)0.0750.023 Diabetes9 (32%)10 (22%)6 (21%)0.3650.975 Smoking3 (11%)15 (33%)11 (39.3%)0.0140.606 History of myocardial infarction0 (0%)11 (24%)4 (14%)0.0380.318 Prior TIA0 (0%)0 (0%)6 (21%)0.0100.001 Prior CAD0 (0%)21 (46%)8 (29%)0.0020.075 Prior ischemic stroke0 (0%)0 (0%)8 (27%)0.002< 0.001Laboratory data HCT-40.8 (39.0-43.8)40.2 (36.3–42.8)0.078 WBC (x10^9^/L)-6.4 (5.6–7.4)8.3 (6.6–9.8)0.008 High-sensitivity C-reactive Protein (mg/dL)-1.7 (0.98–2.83)2.2 (1.2-4.0)0.038 Fibrinogen (mg/dL)-372.0 (323.3-417.8)359.5 (247.8-386.6)0.095 INR-1.0 (0.9–1.1)1.0 (1.0-1.1)0.144 NIHSS at admission (1–4) (5–15)--15 (54%)13 (46%)-Concomitant treatment before the acute ischemic stroke Diuretics-10 (22%)8 (28%)0.506 Statins-38 (83%)27 (93%)0.078 ACEi/ARB5 (18%)28 (61%)19 (66%)< 0.0010.545 B-blockers6 (21%)28 (61%)11 (39%)0.1460.071

Continuous variables were reported either as mean ± standard deviation or median (interquartile range), with between-group comparisons conducted using either Student’s t-test or Mann-Whitney U test as appropriate. p values marked with bold indicate statistically significant differences between the groups < 0.05.

Microarray analysis of plasma MiRNAs

Results of the differential expression analysis

Statistical evaluation of the microarray results detected 404 DE probes (146 upregulated, 258 downregulated). Among these, target prediction analysis was performed for 192 miRNAs (67 upregulated and 125 downregulated) that were successfully mapped using the multiMir R package (Supplementary 1, "Ischemic stroke patients and control group demographics" section).

Identification of the top MiRNAs associated with ischemic stroke

In order to identify the most significantly altered miRNAs, we filtered out the data based on the following criteria: p value in Mann-Whitney test less than 0.05, DE miRNA was expressed in both conditions (parameter from TAC software), and absolute Fold Change (FC) was at least 2. We identified 126 such miRNAs; 23 of them were upregulated and 103 were downregulated. 23 upregulated and 50 downregulated top miRNAs were mapped by multimir for target predictions. The volcano plot of all microarray results is shown in Fig. 1.

Fig. 1. Volcano plot with the results of the differential expression analysis of stroke vs. CV-matched control group samples

Results of the enrichment analysis for the top MiRNAs

To elucidate the functional impact of miRNAs linked to ischemic stroke, we conducted enrichment analyses on predicted target genes of the top DE miRNAs. Due to the high number of targets regulated by analyzed miRNAs, we decided to focus only on those that were regulated by at least 20% of miRNAs with the same directionality of expression changes. Performed separate enrichment analyses for 7162 targets of upregulated top miRNAs and 153 downregulated. The workflow of the bioinformatic analyses is shown in Fig. 2.

Fig. 2. Workflow of the bioinformatic analysis. DE- differentially expressed

Pathways

Our analysis revealed that targets of the top upregulated miRNAs showed significant enrichment in key Bioplanet pathways, including BDNF signaling, Interleukin-2 signaling, cancer pathways, FSH-mediated apoptosis regulation, axon guidance, and TGF-beta-dependent extracellular matrix regulation. Conversely, targets of the top-downregulated miRNAs were predominantly associated with developmental biology, neuronal system function, and NGF signaling pathways **(**Fig. 3A).

Fig. 3. Top 10 enriched terms associated with the targets of at least 3 up- or down-regulated top miRNAs. A Bioplanet signaling pathways, (B) dbGaP phenotypes, (C) ARCHS4 Tissues, (D) Cell Markers. Y axis- Top 15 enriched terms for up- and down-regulated top miRNAs. X axis-genes targeted by the top differentially expressed miRNAs with the same regulation directionality. Numbers on the right sides of the circles symbolizing gene number are related to ranks of the processes; the lower number, the more significant process. Numbers on the left side of the circles symbolize gene number. The figure also shows whether a given process is significant after p-value adjustment (blue border), significant only before adjustment (grey border), or not significant (yellow border). In addition, a visual legend indicates the number of associated genes using both color and node size

Phenotypes and genotypes

We found that targets of the most important upregulated miRNAs were most significantly modulated with Body Height, and CV-related terms including Body Mass Index, Echocardiography, Cholesterol-HDL, Blood Pressure, Coronary Disease and Cholesterol-LD. For down-regulated miRNAs, significantly associated targets included: Lymphocytes, Creatinine, Attention Deficit Hyperactivity Disorder with Hyperactivity and Triglycerides (Fig. 3B).

Tissues and cell markers

Enrichment analysis focused on tissues and cells affected by the top ischemic stroke-related miRNAs showed that targets of the upregulated ones were significantly modulated in neuronal epithelium, prefrontal cortex, and myoblast. We also observed enrichment of the pancreatic islet term. Targets of top-downregulated miRNAs were related to breast (bulk tissue) and nervous tissue, including prefrontal cortex and spinal cord (Fig. 3C).

Analysis of cell markers showed sole association of targets of the top up-regulated miRNAs with Natural Killer T (NKT) cell: Fetal Kidney, Gonadal Endothelial cell: Fetal Gonad, Meiotic Prophase Fetal Germ cell: Fetal Gonad. Among the top terms, we also observed blood-related terms: Vascular Progenitor cell and Vascular Stem cell. While targets of top-downregulated miRNAs were enriched in terms related to the following adult cells: Purkinje cell (Brain), Endocrine cell (Pancreas), SLC16A7 + cell (Lung), Endocrine Progenitor cell (Pancreas), and Endometrial Progenitor cell (Endometrium) (Fig. 3D).

Identification of the top targets affected by ischemic stroke-related MiRNAs

The aim of this analysis was the identification of the targets that are the most affected by DE miRNAs based on their number and directionality of regulation. Our analysis demonstrated that ANKRD52 and AGO1 were common targets of all DE miRNA types. Among the most strongly regulated targets, ANKRD12 and HIF1A showed particular susceptibility to upregulated miRNAs, while GNAI2 and GRIN1 were most responsive to downregulated miRNAs (Fig. 4).

Fig. 4. Top gene networks affected by ischemic stroke-related miRNAs

Expression alterations, and the potential roles of the analyzed miRNAs and predicted mRNAs in dynamic post-stroke molecular responses

miR-18a-5p and miR-199a-5p showed significantly elevated expression in stroke patients at both day 1 and day 7 compared to healthy individuals and CV-matched controls (p < 0.05 for all). In contrast, miR-4467 expression was significantly lower at day 1 compared to both control groups (p < 0.05 for both), but markedly increased by day 7 versus day 1 (p < 0.001), remaining higher than in both control groups (p < 0.05 for both). miR-3135b demonstrated persistent downregulation at both time points compared to both control groups (p < 0.05 for all). ROC curve analysis revealed significant diagnostic performance for all baseline miRNAs in acute ischemic stroke. miR-199a-5p demonstrated the highest discriminatory power (AUC = 0.89, 95% CI: 0.81–0.97, p < 0.001), followed by miR-3135b (AUC = 0.88), miR-4467 (AUC = 0.80), and miR-18a-5p (AUC = 0.73), all with p < 0.001 except for miR-18a-5p (p = 0.002).

From potential role in dynamic post-stroke molecular responses point of view, miR-3135b, miR-199a-5p, and miR-18a-5p did not show prognostic utility, as the paired changes between day 1 and day 7 were not statistically significant in stroke patients (p = 0.059, p = 0.171, and p = 0.264, respectively). Among the four analyzed miRNAs, only miR-4467 showed post-stroke dynamic, being significantly increased during hospitalization within 7 days (p < 0.001) **(**Fig. 5).

In addition to validating miRNAs identified in the microarray analysis, we also assessed the expression of their predicted target genes in our cohort. GNAI2 mRNA levels were significantly elevated on the first day post-stroke compared to both control groups (p < 0.05). Conversely, ANKRD12 expression was lower in stroke patients both at admission and on day 7 compared to both controls (p < 0.05 for all). GRIN1 levels were significantly reduced on the day of hospital admission compared to the CV-matched controls (p = 0.006), but increased by day 7 post-stroke. By day 7, GRIN1 expression did not differ significantly from either control group (p > 0.05 for both). ROC curve analysis revealed significant diagnostic performance of ANKRD12, GNAI2, and GRIN1 baseline mRNAs for acute ischemic stroke (AUC: 0.683; 0.698; 0.693, respectively).

From potential role in dynamic post-stroke molecular responses perspective, HIF1A mRNA levels showed no significant difference at day 1 compared to both control groups (p > 0.05 for both), but decreased markedly by day 7 relative to day 1 (p < 0.001), and were significantly lower than in both controls (p < 0.05 for both). Similarly, GNAI2 levels were also reduced on day 7 compared to both day 1 and control groups, supporting the potential post-stroke dynamic of HIF1A and GNAI2 in our study **(**Fig. 5).

Fig. 5. MiRNAs and predicted genes expression difference between groups. a miR-18a-5p; b) ROC analysis of miR-18a-5p baseline; c) miR-4467; d) ROC analysis of miR-4467 baseline; e) miR-199a-5p; f) ROC analysis of miR-199a-5p baseline; g) miR-3135b; h) ROC analysis of miR-3135b baseline; i) ANKRD12; j) ROC analysis of ANKRD12 baseline; k) HIF1A; l) ROC analysis of HIF1A baseline; m) GNAI2; n) ROC analysis of GNAI2 baseline; o) GRIN1; p) ROC analysis of GRIN1 baseline; p value was calculated with Mann-Whitney U or Wilcoxon test, appropriately. Four-group comparison was calculated by the Kruskal-Wallis test. Abbreviations: Log10, log10 transformation; miR, microRNA; p, value; AUC, area under the curve; CI, confidence interval; ROC, Receiver operating characteristic

Prediction of the role of MiRNAs determined by qRT-PCR in ischemic stroke

hsa-miR-3135b: Enrichment analysis of the terms associated with the miRNAs determined with qRT-PCR showed a strong association of downregulated hsa-miR-3135b with cancerous pathways, including Glioma, Endometrial cancer, and Prostate cancer. We also observed enrichment of a cholesterol-related phenotype in various tissues, including the brain cortex and pancreatic islets, as well as cell markers such as SLC16A7 + cells (Lung) and PROM1High Progenitor cells (Large Intestine). In total, we identified 7 ischemic stroke-related genes mentioned in at least 10 publications (CCBE1, ERBB2, IL23R, IL2RA, POU2F2, PTPN2, TP53) as targeted by hsa-miR-3135b. hsa-miR-4467: On the other hand, also downregulated hsa-miR-4467 showed a much lower number of targets. In total, we have not identified any ischemic stroke-related genes mentioned in at least 10 publications as targeted by hsa-miR-4467. hsa-miR-18a-5p: We identified 12 ischemic stroke-related genes mentioned in at least 10 publications (ABCC2, EMP2, FAS, FCGR2B, IGF1, IL23R, LMNA, MAPK14, RIT1, SMAD4, TP53, VEGFA) as targeted by hsa-miR-18a-5p. hsa-miR-199a-5p: We identified 16 ischemic stroke-related genes mentioned in at least 10 publications (ACE, CCL2, CNBP, CSF3, EDN1, ERBB2, IL13, IL23R, KRAS, MAPK14, PKD2L1, PTGS2, PTPN14, SMAD4, TGFBR1, VEGFA) as targeted by hsa-miR-199a-5p **(**Figs. 6 and 7).

Fig. 6. Predicted enrichment panel of miR-3135b, miR-4467, miR-18a-5p, and miR-199a-5p for pathways (A) and tissues (B). Numbers associated with circles symbolizing gene number are related to the ranks of the processes; the lower the number, the more significant the process. The figure also shows whether a given process is significant after p-value adjustment (blue border), significant only before adjustment (grey border), or not significant (yellow border). In addition, a visual legend indicates the number of associated genes using both color and node size

Fig. 7. Predicted enrichment panel of miR-3135b, miR-4467, miR-18a-5p, and miR-199a-5p for phenotypes (A) and cell types (B). Numbers associated with circles symbolizing gene number are related to the ranks of the processes; the lower the number, the more significant process. The figure also shows whether a given process is significant after p-value adjustment (blue border), significant only before adjustment (grey border), or not significant (yellow border). In addition, a visual legend indicates the number of associated genes using both color and node size

Discussion

Our analysis showed alteration of circulating miRNAs after acute ischemic stroke. Since many known biomarkers related to ischemic stroke are not specific, there is a need for novel biomarkers to allow early diagnosis and avoid irreversible brain injury. Specific miRNAs are promising biomarkers to improve both diagnosis and prognosis of ischemic stroke. In fact, we were able to identify a set of miRNAs whose circulating levels were significantly affected in stroke patients. Most importantly, we found both up-regulated (miR-18a-5p, miR-199a-5p) and down-regulated (miR-4467, miR-3135b) small RNAs. Additionally, these miRNAs presented different kinetics, potentially allowing for the identification of stroke at different time points, while also offering the chance to determine the time of the acute event. In fact, while miR-18a-5p is significantly up-regulated at day 1, with a slight further increase over the next days, miR-3135b shows an analogous modulation in the opposite direction. miR-4467 exhibits substantial downregulation at day 1, with a rapid return to baseline values by day 7.

Limited data show that miR-4467 is upregulated in cerebrospinal fluid (CSF) of patients with Alzheimer’s Disease and may be a potential diagnostic biomarker of neurodegenerative diseases [24, 25]. Moreover, increased miR-4467 was also observed in patients with schizophrenia, a complex, multifactorial brain disorder [26]. Interestingly, our findings show that miR-4467 was significantly downregulated on the first day of stroke, whereas it increased to baseline on the seventh day of stroke. The potential explanation of the difference in miR-4467 expression is the nature of the neurological disease process. Recent studies show that the path mechanism involved in the acute phase of ischemic stroke and thus acute neuronal injury differs from chronic neurodegeneration. The pathological processes associated with acute ischemic brain tissue injury, such as inflammation, oxidative stress, apoptosis, or temporal changes in cell signaling, appear rapidly and may be involved in secondary injury processes, but importantly, can also be involved in reparative and neuroprotective processes [27, 28]. While neurodegenerative diseases exhibit common underlying pathological mechanisms, their chronic nature typically leads to the progressive degeneration of specific, susceptible neuronal populations [29]. Thus, miR-4467 serves as a promising biomarker for the early phase of ischemic stroke.

We identified hsa-miR-18a-5p as one of the most upregulated miRNAs in stroke patients. Limited data show that hsa-miR-18a-5p may be involved in the pathogenesis of ischemic stroke and other cerebrovascular diseases. Current studies show that TLR8, which plays a pivotal role in the inflammation cascade and ischemic cerebral injury, is upregulated in patients with ischemic stroke and is directly targeted by hsa-miR-18a-5p [30]. Upregulation of hsa-miR-18-5p could alleviate OGD/R-induced cell injury via inhibition of TLRs and NF-κB pathway in an in vitro model [30]. Therefore, the neuroprotective effect of hsa-miR-18a-5p could be potentially used as a prognostic biomarker in ischemic stroke patients; however, further in vivo studies are needed. Additionally, in the Framingham Heart Study cohort, miR-18a-5p-a1 was the most significant miRNA for survival after stroke [31]. Besides its biomarker potential, miR-18a-5p could also be a possible target for novel therapeutics. Recent data show that miR-18a improves human brain endothelial cells’ functioning in cerebral arteriovenous malformations [32]. In our study, we found that miR-18a expression was significantly lower in healthy individuals compared to both CV risk-matched controls and stroke patients at day 1 and day 7. This finding aligns with previous observations linking miR-18a to CAD and myocardial infarction. A previous study investigated the role of the long noncoding RNA CASC2 and its regulatory axis involving miR-18a/SIRT2 in acute myocardial infarction. The authors found that CASC2 expression was significantly decreased in AMI patients compared to controls, as well as in rat models of AMI. In cellular experiments, they demonstrated that CASC2 exerts a protective effect on cardiomyocytes by suppressing miR-18a. Downregulation of CASC2 led to an increase in miR-18a, which enhanced oxidative stress by promoting reactive oxygen species (ROS) generation. Conversely, inhibition of miR-18a mitigated this damage. These findings indicate that CASC2 protects against AMI by downregulating miR-18a and thereby reducing oxidative stress through the SIRT2/ROS pathway [33]. Conversely, a previous study reported contradictory findings, showing that miR-18a expression was significantly lower in CAD patients than in healthy controls. Using next-generation sequencing followed by PCR validation, this study revealed that several miRNAs from the miR-17-92 cluster, including miR-18a, miR-92a, miR-106b, and miR-17, were differentially expressed between CAD patients and controls. Moreover, these miRNAs correlated with specific lipid parameters: miR-17 was associated with elevated LDL levels, whereas miR-92a and miR-106b were linked to HDL [34]. Importantly, although hsa-miR-18a-5p was identified as one of the most upregulated miRNAs in our cohort, its AUC value of 0.73 reflects moderate diagnostic performance and warrants cautious interpretation. Given the limited and somewhat inconsistent evidence regarding the role of miR-18a, more detailed analyses, particularly in larger human cohorts, are needed.

Enrichment analysis of the terms modulated by the small RNAs showed a strong association of upregulated miRNAs with interleukin-2 signaling pathways. Interleukin-2, which is an immunomodulatory lymphocyte secreted by T lymphocytes, plays a pivotal role in ischemic stroke through the multiplication and differentiation of T and B cells, inflammatory reaction, and information transmission [35]. The IL-2/IL-2 antibody complex (IL-2/IL-2Ab) exhibits neuroprotective effects and may improve ischemic stroke prognosis by T cells regulation [35]. Moreover, Elevated serum sIL-2Rα concentrations and reduced IL-2 levels in ischemic stroke patients correlated significantly with worse clinical outcomes, identifying these biomarkers as novel independent predictors of disease prognosis [36].

In the present study, we found that miR-199a was significantly expressed on day one in the plasma of stroke patients and remained high even after seven days. However, we cannot establish any prognostic role of miR-199 in these patients. The data on the role of miR-199 in ischemic stroke are scant. In an ischemic stroke animal model, α-tocopherol treatment exerted neuroprotective effects while significantly reducing miR-199a-5p expression levels [37]. Consistent with these findings, miR-199a-5p was similarly implicated in ischemic stroke rats. A study demonstrated that miR-199a-5p silencing improves cognitive function and decreases hippocampal neuronal apoptosis via AKT signaling pathway activation [38]. Moreover, both ANRIL overexpression (a long non-coding RNA) and miR-199a-5p inhibition conferred protection against ischemia-induced cellular injury by enhancing cell viability through the CAV-1-mediated MEK/ERK signaling pathway [39]. In addition to this, numerous studies indicated that the activated AKT signaling pathway can improve ischemic stroke [38, 40–42].

While ischemic stroke represents a relatively uncommon complication of acute myocardial infarction, it remains one of its most clinically significant consequences [43]. Multiple studies have demonstrated that inhibiting miR-199a-5p alleviates pulmonary hypertension and myocardial injury, thereby improving cardiovascular function. These cardioprotective effects may indirectly influence outcomes in ischemic stroke [44–48]. In line with our results, De Rosa et al., reported the high expression of miR-199a in ischemic and non-ischemic heart failure patients [49]. Conversely, both miR-199a-3p and miR-199a-5p contribute to redundant regulatory networks controlling the endothelial NOS/NO pathway. Elevated NO levels exacerbate infarct volume and cerebrovascular damage, while sustained post-stroke hyperglycemia promotes excessive NO and peroxide accumulation, resulting in microvascular impairment and worsened clinical outcomes [50, 51]. This suggests miR-199a-5p as an injury-related miRNAs. However, at the same time, it poses an issue, which is to address possible differential diagnoses such as acute myocardial injury. Moreover, we acknowledge that the discrepancies between our findings and previous studies may be due to the reliance on animal and in vitro models in earlier research. Further analyses are therefore required in human ischemic stroke populations.

Our study revealed persistently reduced miR-3135b levels in stroke patients, which remained low even 7 days post-stroke. Emerging evidence implicates the hsa-miR-3135b/REL/SOD2 axis in the progression of ischemic cerebral infarction, supported by elevated REL and SOD2 expression in patients compared with controls [52]. Notably, miR-3135b also associates with coronary artery calcification and serves as a diagnostic marker for obstructive coronary artery disease [53]. While traditional vascular risk factors (diabetes, hypertension, atherosclerosis) contribute to ischemic stroke pathogenesis [54–56]. Our findings corroborate Sun et al.‘s demonstration of significantly lower miR-3135b expression in stroke patients compared to controls, reinforcing its potential role in cerebrovascular disease [57]. In our study, we also observed that miR-3135b expression was significantly higher in healthy individuals compared to both CV risk-matched controls and stroke patients at both time points. Consistent with our findings, a previous study examined the relationship between circulating miRNAs and high-risk clinical traits in patients with non-ST-segment elevation acute coronary syndrome (NSTE-ACS). Whole-genome miRNA sequencing was performed on blood samples from 199 NSTE-ACS patients, and associations were analyzed with 13 high-risk traits, including the validated GRACE mortality risk score. The analysis revealed that chronic heart failure and the GRACE score were the most significant traits, with closely related profiles of associated miRNAs. Statistical analysis showed lower circulating levels of miR-3135b were significantly associated with both the presence of chronic heart failure and higher GRACE mortality risk. The observed underexpression of miR-3135b in these conditions highlights its potential prognostic importance and supports the need for further investigation [58].

Apart from the miRNA analysis, we have predicted the genes that regulate ischemia process through bioinformatic analysis, and thereafter, we have validated their expression in our cohort. ANKRD12 gene (Ankyrin Repeat Domain 12) primarily acts as a transcriptional corepressor, and its role was mainly studied in colorectal, breast, and ovarian cancer [59, 60]. There are a few reports that ANKRD12 mutation plays a crucial role in neurodegenerative diseases like Parkinson’s Disease [61]. To date, there are no studies evaluating the role of ANKRD12 in ischemic stroke.

Moreover, Deng et al. evaluated the role of GNAI2 (G Protein Subunit Alpha I2) and miR-30c-5p in cerebral ischemia-reperfusion (I/R) injury in in vitro and animal models. GNAI2 upregulation was associated with enhanced inflammation and apoptosis. Inhibition of GNAI2 attenuated the influence of downregulated miR-30c-5p, showing the neuroprotective effect on cerebral ischemia/reperfusion injury [62].

FosDT (Fos downstream transcript) is a long non-coding RNA that plays a critical role in post-ischemic gene regulation [63]. The mechanism is associated with increased binding to chromatin-modifying proteins Sin3a and coREST (corepressors of the transcription factor REST). One of the key REST target genes affected by this mechanism is GRIN1, which encodes a subunit of the NMDA receptor. After ischemic stroke, increased FosDT expression contributes to the suppression of GRIN1. However, silencing FosDT restores GRIN1 expression, improves motor recovery, and reduces infarct volume, suggesting that FosDT-mediated repression of GRIN1 may exacerbate ischemic brain injury [63]. It is important to note that, thanks to our target prediction analysis based on microarray miRNA profiling, we have demonstrated for the first time the alteration and potential diagnostic and post-stroke dynamic relevance of GRIN1 and ANKRD12 gene expression in the human population following acute ischemic stroke.

Limitations

Our study has several limitations that should be acknowledged. First, the modest sample size of the discovery microarray cohort, which may limit the generalizability of the findings. Second, although two different control groups were used for qRT-PCR validation, the ischemic stroke cohort lacked an external validation cohort to confirm the diagnostic utility of the identified miRNAs. Third, restricted longitudinal sampling limited to day 1 and day 7, which does not fully capture post-stroke molecular dynamics; therefore, future studies should incorporate predefined serial sampling at multiple time points (e.g., Days 1, 3, 7, 14, and 30), which constrains the assessment of the long-term post-stroke dynamic relevance of these biomarkers. Finally, while the mRNA targets of the identified miRNAs were predicted using bioinformatic platforms (multiMiR and Wizbionet) and partially validated by qRT-PCR, no functional experiments (e.g., luciferase reporter assays or miRNA/mRNA binding studies) were performed to confirm direct regulatory interactions.

Conclusion

For the first time, in our analysis, we have validated the role of ANKRD12 and GRIN1 genes in patients with ischemic disease by presenting their target genes via in silico prediction network analysis. Besides, our integrated miRNA/mRNA analysis identified distinct molecular signatures in acute ischemic stroke, with 146 upregulated and 258 downregulated RNAs, implicating key neuroinflammatory and neuroprotective pathways, including BDNF, IL-2, and NGF signaling. Among the validated candidates, miR-199a-5p, miR-3135b, miR-4467, and miR-18a-5p demonstrated diagnostic potential, while miR-4467, together with GNAI2 and HIF1A, showed post-stroke dynamic relevance, reflecting early transcriptomic adaptations following ischemic injury. Measurement techniques for miRNA are developing and may, in the near future, enable the use of these biomarkers in clinical practice. Moreover, to conclusively confirm the direct regulatory interactions between the identified miRNAs and their predicted mRNA targets, we recommend that further researchers perform functional validation experiments. Luciferase reporter assays or other miRNA/mRNA binding studies would be the logical next step to build upon the promising bioinformatic and qRT-PCR data we have gathered. Accordingly, this study should be viewed as exploratory, and the identified miRNAs are not intended to replace or outperform established clinical or imaging markers but rather to contribute to the understanding of post-stroke molecular responses.

Supplementary Information

Supplementary Material 1.

Supplementary Material 2.