Identification of Novel Circulating miRNAs in Patients with Acute Ischemic Stroke

Abstract

Ischemic strokes are associated with significant morbidity and mortality, but currently there are no reliable prognostic or diagnostic blood biomarkers. MicroRNAs (miRNAs) regulate various molecular pathways and may be used as biomarkers. Using RNA-Seq, we conducted comprehensive circulating miRNA profiling in patients with ischemic stroke compared with healthy controls. Samples were collected within 24 h of clinical diagnosis. Stringent analysis criteria of discovery (46 cases and 95 controls) and validation (47 cases and 96 controls) cohorts led to the identification of 10 differentially regulated miRNAs, including 5 novel miRNAs, with potential diagnostic significance. Hsa-miR-451a was the most significantly upregulated miRNA (FC; 4.8, FDR; 3.78 × 10^-85), while downregulated miRNAs included hsa-miR-574-5p and hsa-miR-142-3p, among others. Importantly, we computed a multivariate classifier based on the identified miRNA panel to differentiate between ischemic stroke patients and healthy controls, which showed remarkably high sensitivity (0.94) and specificity (0.99). The area under the ROC curve was 0.97 and it is superior to other current available biomarkers. Moreover, in samples collected one month following stroke, we found sustained upregulation of hsa-miR-451a and downregulation of another 5 miRNAs. Lastly, we report 3 miRNAs that were significantly associated with poor clinical outcomes of stroke, as defined by the modified Rankin scores. The clinical translation of the identified miRNA panel may be explored further.

Keywords: biomarkers; ischemic stroke; miRNA; microRNA.

Figure 1

Study design. (A) Serum samples from stroke patients and healthy controls were collected to isolate circulating miRNAs and generate libraries for RNA-Seq. Multiple bioinformatics tools were utilized for analyses and visualization of sequencing data. (B) Study populations included healthy controls, stroke baseline (BL), and stroke follow-up (FU) patients. Each study population was randomly divided into discovery and validation (replication) cohorts for downstream analyses for the identification of differentially regulated miRNAs.

Figure 2

Circulating miRNA profiling of stroke baseline patients compared with healthy controls. (A) Volcano plot shows overall differentially regulated miRNAs in stroke baseline patients from the discovery cohort; red dots represent upregulated, blue represent downregulated, and gray dots represent unchanged miRNAs based on significance (FDR < 0.05) and log₂FC > 2 (dotted vertical lines). (B) Bar plot represents the number of upregulated (red bar) and downregulated (blue bar) miRNAs using the specified cutoffs. (C) Column graph shows the Log₂FC + standard error of the mean (SEM) of the 3 upregulated and 25 downregulated miRNAs in the discovery cohort. (D) Volcano plot, (E) bar plot, and (F) column graph show the 1 upregulated and 10 downregulated miRNAs in the validation cohort.

Figure 3

Validated differentially regulated circulating miRNAs between stroke baseline patients and healthy controls. (A) Venn diagram shows the total number of overlapping downregulated and upregulated miRNAs in the stroke baseline versus healthy control comparison between discovery (gray) and validation (light blue) cohorts. (B) Column graph shows the Log₂FC + standard error of the mean (SEM) of 1 upregulated and 9 downregulated validated miRNAs. (C) Box and whiskers plots show the difference in counts per million (CPM) in stroke baseline (BL) and healthy controls (HC) in discovery (gray) and validation (light blue) cohorts of the 10 differentially regulated, validated miRNAs in stroke BL versus healthy controls. Mean with minimum and maximum values, and upper and lower quartiles are depicted for each data set with significant comparisons annotated by an asterisk (*) on top (p < 0.0001).

Figure 4

Diagnostic capacity of differentially regulated circulating miRNAs in stroke patients. The orthogonal partial-least-squares-discriminant analysis (OPLS-DA) was performed using the top differentially regulated miRNAs (n = 27) in the discovery cohort data. The classifier was trained on data from all participants in (A) discovery cohort (n = 142) and tested on the (B) validation cohort (n = 143). Scatter plots show the predictive component to discriminate stroke cases from healthy controls (green dots—x-axis) versus the orthogonal component representing a multivariate confounding effect that is independent of stroke (blue dots—y-axis). (C) ROC curve analysis generated an overall sensitivity of 0.94, specificity of 0.99, and AUC of 0.97.

Figure 5

Circulating miRNA profiling of stroke follow-up patients compared with healthy controls. (A) Volcano plot shows overall differentially regulated miRNAs in stroke follow-up patients from the discovery cohort; red dots represent upregulated, blue represent downregulated, and gray dots represent unchanged miRNAs based on significance (FDR < 0.05) and log₂FC > 2 (dotted vertical lines). (B) Bar plot represents the number of upregulated (red bar) and downregulated (blue bar) miRNA using the specified cutoffs. (C) Column graph shows the Log₂FC + standard error of the mean (SEM) of the 2 upregulated and 24 downregulated miRNAs in the discovery cohort. (D) Volcano plot, (E) bar plot, and (F) column graph show the 6 upregulated and 10 downregulated miRNAs in the validation cohort. (G) Venn diagram shows the total number of overlapping 6 downregulated and 1 upregulated miRNA in the stroke follow up versus healthy control comparison between discovery (gray) and validation (light blue) cohorts. The 7 validated miRNAs are also listed.

Figure 6

Associations between circulating miRNAs and clinical outcomes of ischemic stroke. Circulating miRNA profiles of stroke BL patients were compared with healthy controls based on the 90-day clinical follow up (mRS scores). Stroke BL patients with good outcome (discovery cohort; n = 77 and validation cohort; n = 84) and poor outcome (discovery cohort; n = 18 and validation cohort; n = 12) were compared with healthy controls (discovery cohort; n = 46 and validation cohort; n = 47). (A) Venn diagram lists the differentially regulated and overlapping circulating miRNAs in healthy controls versus stroke good outcome and stroke poor outcome patients. (B) Box and whiskers plots show the difference in counts per million (CPM) in stroke poor outcome (PO) and healthy controls (HC) in discovery (gray) and validation (light blue) cohorts of the 3 differentially regulated miRNAs, which were unique to stroke poor outcome patients. Mean with minimum and maximum values, and upper and lower quartiles are depicted for each data set with significant comparisons annotated by an asterisk (*) on top (p < 0.0005).