The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board (or Ethics Committee) of Xuanwu Hospital, Capital Medical University. This study included 345 patients diagnosed with AIS by experienced neurologists at the Emergency and Neurology Department of Xuanwu Hospital between November 2018 and September 2019, as well as 37 healthy volunteers. The time elapsed from symptom onset to hospital admission was also confirmed by experienced physicians through detailed medical history taking. This duration will be referred to as “time to hospital admission” and “onset-to-treatment time” in the following text. Inclusion criteria for AIS patients were: (1) diagnosis confirmed via brain MRI or CT; (2) evident neurological deficits (such as motor and/or sensory deficits on the opposite side, impairment of higher cerebral functions and homonymous hemianopia); (3) age $\ge$18 years; (4) availability of complete case and follow-up data; (5) present within 24 h of symptom onset; and (6) consent to participate. The exclusion criteria were: (1) other cerebrovascular diseases such as cerebral hemorrhage diagnosed by CT or MRI, transient ischemic attack and hypertensive encephalopathy; (2) traumatic brain injury, encephalitis, multiple sclerosis or epilepsy; (3) severe infectious disease; (4) neoplasms; and (5) missing clinical baseline data. Ultimately, 205 patients with AIS were included in the regression model (Fig. 1).

Fig. 1.

Flow chart of this study. Abbreviation: miRNA-494, microRNA-494; LASSO, least absolute shrinkage and selection operator; ROC, receiver operating characteristic.

Stroke severity was evaluated using the NIHSS, where scores 0–5 indicated minor stroke, 6–10 mild stroke, 11–14 moderate stroke, and 15–42 severe stroke [21, 22]. After 3 months, the mRS was used to assess patient outcomes. The primary outcome was defined as an excellent outcome (mRS score $\le$1) at 3 months post-AIS onset [23, 24, 25, 26]. Secondary outcome was defined as minor-to-moderate stroke (NIHSS score 15) at admission [27].

According to the Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria [28], AIS patients were classified into four subgroups for descriptive and analytical purposes (no patients fell under “other determined etiology”): large-artery atherosclerosis, cardioembolic infarction, lacunar infarct and undetermined etiology.

Hypertension is defined as a clinic systolic blood pressure $\ge$140 mmHg and/or diastolic blood pressure $\ge$90 mmHg in the absence of antihypertensive medication [29]. Diabetes mellitus is defined as fasting plasma glucose $\ge$7.0 mmol/L and/or two-hour glucose $\ge$11.1 mmol/L after a 75 g oral glucose tolerance test based on optimal WHO criteria [30]. Hyperlipidemia (or dyslipidemia) is defined as the presence of one or more of the following in the serum: total cholesterol $\ge$6.2 mmol/L, triglycerides $\ge$2.3 mmol/L, low-density lipoprotein cholesterol $\ge$4.1 mmol/L, or high-density lipoprotein cholesterol 1.0 mmol/L [31].

Peripheral venous blood from patients on admission and healthy volunteers was collected in ethylenediaminetetraacetic acid (EDTA) anticoagulant tubes. Lymphocytes were separated using the standard Ficoll-Paque Plus method. First, plasma and blood cells were separated through centrifugation. The blood cells were suspended in physiological saline and then carefully added along the wall of the centrifuge tube into the Lymphocyte Separation Medium (Haoyang Biotech, LTS1077, Tianjin, China). Following centrifugation, the lymphocytes from the middle layer were harvested and added to the red blood cell lysis solution (155 mM NH₄Cl, 12 mM NaHCO₃ and 0.1 mM EDTA, 12709101, 12600701, 10104201, Xilong Scientific Co., Ltd., Shantou, Guangdong, China). After vortexing and centrifugation, the pelleted cells were washed with physiological saline and transferred into new RNase-free EP tubes. The lymphocytes were finally suspended in 1 mL of TRIzol (Invitrogen Life Technologies, 15596026CN, Carlsbad, CA, USA) and stored at –80 °C until use. All samples were processed consistently throughout the study.

The samples were removed from –80 °C and allowed to thaw on ice; 200 µL of chloroform (10006818, Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) was added to each EP tube, followed by vigorous shaking for 15 seconds and incubation at 25 °C for 15 minutes until phase separation was observed. The tubes were centrifuged (4 °C, 12,000 rpm, 15 minutes). Then, the upper aqueous phase was carefully transferred to a new RNase-free EP tube, and 500 µL of isopropanol (80109218, Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) was added. The mixture was incubated overnight at –20 °C. The next day, the tubes were centrifuged again (4 °C, 12,000 rpm, 10 minutes). The supernatant was discarded, and the pellets were washed twice with 1 mL of 75% ethanol (10009218, Sinopharm Chemical Reagent Co., Ltd.). After removing the ethanol, the tubes were placed on a clean bench and air-dried at room temperature. Subsequently, 40 µL of diethylpyrocarbonate (DEPC)-treated water (Invitrogen Life Technologies, AM9920) was added to each tube to dissolve the RNA. The RNA purity and concentration were determined using a spectrophotometer, measuring absorbance ratios at 260/280 and 230/260. RNA integrity was assessed by agarose gel electrophoresis to check for any degradation.

The level of miRNA-494 in the peripheral blood lymphocytes of patients and volunteers was measured using RT-qPCR. First, cDNA was synthesized from total RNA using M-MuLV Reverse Transcriptase (P7040L; Enzymatics, Beijing, China). RT-qPCR was performed with a 2X PCR master mix (AS-MR-006-5, Arraystar, Rockville, MD, USA) on a QuantStudio5 Real-time PCR System (Applied Biosystems, Foster City, CA, USA). The PCR program was set as follows: initial denaturation at 95 °C for 10 minutes, followed by 40 cycles of PCR (95 °C for 10 seconds, 60 °C for 60 seconds with fluorescence collection). The mRNA levels were normalized to the level of U6, and the relative expression of each mRNA was calculated using the 2^{-Δ⁢Δ⁢CT} method. The following primers were used: U6, F: 5^′ GCTTCGGCAGCACATATACTAAAAT 3^′ and R: 5^′ CGCTTCACGAATTTGCGTGTCAT 3^′; has-miR-494-5p, GSP: 5^′ GGGAGGTTGTCCGTGTTGT 3^′ and R: 5^′ GTGCGTGTCGTGGAGTCG 3^′.

All data were analyzed using R software (version 4.2.3, R Core Team, R Foundation for Statistical Computing, Vienna, Austria), IBM SPSS Statistics 25 (SPSS Inc., Chicago, IL, USA), and GraphPad Prism 8.2.1 (GraphPad Software, La Jolla, CA, USA). The criteria for type one error was set at $\alpha$ = 0.05. We assessed the normality of the data distribution using the Shapiro–Wilk test for samples with n 50 or the Kolmogorov–Smirnov test for samples with n $\ge$ 50. For normally distributed data, the central tendency of continuous variables is described by the mean, with the standard deviation (SD) expressing the variability. For non-normally distributed data, the central tendency is represented by the median and the variability by the interquartile range. Categorical variables are presented as proportions. Depending on the data distribution, continuous variables were analyzed using either the Student’s t-test or the Mann–Whitney U test. Categorical variables were analyzed using either Pearson’s chi-square test or Fisher’s exact test. A logistic regression model was employed to determine the odds ratios (ORs) and 95% confidence intervals (95% CIs) to assess the association between miRNA-494 levels in lymphocytes and patient outcome at 3 months. This same analysis was conducted to evaluate the relationship between miRNA-494 and disease severity on admission. The receiver operating characteristic (ROC) curve and Youden index were used to convert miRNA-494 into a dichotomous variable [32]. Five-fold cross-validation of least absolute shrinkage and selection operator (LASSO) regression was used to screen valid clinical indicators. Sensitivity analysis was performed using multiple adjusted multivariate logistic regression models to test the robustness of miRNA-494 findings [33, 34]. For the prediction model linking miRNA-494 findings and outcomes, we included the admission NIHSS score, admission mRS score, systolic blood pressure (SBP), low-density lipoprotein cholesterol (LDL) level, and platelet count (PLT) as inclusion indicators for the follow-up logistic regression model. For the model relating miRNA-494 to disease severity on admission, we added the neutrophil-to-lymphocyte ratio (NLR), PLT on admission, history of coronary heart disease, and atrial fibrillation to the regression model. The variance inflation factor (VIF) is used to measure the severity of multicollinearity in a multiple linear regression model. When the VIF $\le$10, it is considered that there is no multicollinearity.

Initially, 205 patients with AIS and 37 healthy volunteers were analyzed for differences in miRNA-494 RNA levels in peripheral lymphocytes. The results indicated significantly higher expression in patients with AIS than in healthy individuals (Fig. 2a, p 0.001). The ROC curve demonstrated the predictive ability of miRNA-494 levels in lymphocytes for diagnosing AIS. The area under the curve (AUC) for miRNA-494 levels predicting AIS was 0.774, with a 95% confidence interval of 0.701–0.848 (p 0.0001) (Fig. 2b). The miRNA-494 levels in lymphocytes were also analyzed across different stroke subtypes. As shown in Fig. 2c and Table 1, no significant statistical differences were observed among the groups (p = 0.218, Kruskal-Wallis).

Fig. 2.

miRNA-494 level was elevated in patients with AIS. (a) Comparison between healthy volunteers and AIS patients, using Mann-Whitney U test (****, p 0.0001), bars mean median and interquartile range. (b) ROC curve for miRNA-494 in lymphocytes to predict AIS, healthy group (n = 37), AIS group (n = 205). (c) miRNA-494 expression in different subtypes of AIS. Abbreviation: AIS, acute ischemic stroke; AUC, area under curve; LAAS, large artery atherosclerosis; CEI, cardioembolic infarct; LAC, lacunar infarct; UDE, undetermined etiology.

After excluding patients with missing clinical indicators, we analyzed the baseline characteristics of the 205 patients, including demographics, past medical history, and various clinical indicators on admission, grouped by outcome (Table 1). The mean age of the patients was 64.60 $\pm$ 13.31 years, and 54 patients (26.34%) were female. The median NIHSS score on admission was 6, with a significant difference between the two outcome groups (p 0.001). Regarding past medical history, significant differences were observed between the groups in patients with coronary heart disease (p = 0.029) and atrial fibrillation (p = 0.038) (Table 1).

Preliminary statistical analysis, including comparisons between the two outcome groups and correlation analysis, showed no significant correlation or association between miRNA-494 and the outcome at 3 months or with the mRS and NIHSS on admission. No trends were observed in the distribution plot of the miRNA-494 percentile divided by the interquartile range (Supplementary Fig. 1a–f). Additionally, miRNA-494 did not reach statistical significance in predicting the prognosis of all 205 patients using logistic regression (OR = 1.266, 95% CI: 0.794–2.018, p = 0.322).

To further explore whether miRNA-494 levels in peripheral lymphocytes predict AIS prognosis, we analyzed various subgroups. Univariate logistic regression was employed to calculate the OR, 95% CI, and p-values of miRNA-494 levels in each subgroup, including sex, age (cutoff of 65 years), time to hospital admission (cutoff of 6 h), NIHSS score (cutoff of 5), and treatment acceptance, such as thrombolysis and mechanical thrombectomy. These cutoff values were based on baseline statistical data and clinical considerations. The results indicated that only the group aged under 65 years demonstrated a statistically significant prognostic ability of miRNA-494 levels in peripheral lymphocytes (Fig. 3, p = 0.028).

Fig. 3.

Logistic regression analysis among subgroups of patients with acute ischemic stroke. Univariate logistic regression on miRNA-494 levels and excellent outcome at 3 months of AIS patients, in each subgroup. Higher miRNA-494 levels in peripheral lymphocytes predicted better outcome. Horizontal axis was the level of miRNA-494. Abbreviation: NIHSS, National Institutes of Health Stroke Scale. *Time means onset-to-treatment time of patients.

We focused on patients aged under 65 years (n = 105). The distribution of miRNA-494 in lymphocytes among the six groups according to the mRS score at 3 months is shown in Fig. 4a, revealing a clear trend where lower mRS scores were associated with higher levels of miRNA-494. Additionally, miRNA-494 expression in lymphocytes was significantly negatively correlated with the mRS score at 3 months (Fig. 4b). Univariate logistic regression analysis indicated that the OR for miRNA-494 levels predicting an excellent outcome for patients 3 months post-stroke was 2.800 (95% CI: 1.120–7.002, p = 0.028) (Table 2). The ROC curve analysis identified the optimal cutoff point for miRNA-494 at 1.057, which yielded a sensitivity of 68.4% and specificity of 55.2% for AIS (Fig. 4c, AUC = 0.627, 95% CI: 0.516–0.739). This level was then categorized as a dichotomous variable with a cutoff point of 1.057. We assessed the association of miRNA-494 levels with AIS outcomes using univariate logistic regression analysis, revealing that miRNA-494 $>$1.057 independently predicted an excellent outcome at 3 months in patients younger than 65 years (Table 2, OR = 2.672 [95% CI: 1.158–6.168], p = 0.021).

Fig. 4.

Analysis of prognosis of patients with AIS aged under 65. (a) Age 65 years old, distribution of miRNA-494 percentile among the 6 groups of mRS score at 3 months (n = 105). (b) Scatter plot and correlation, age 65 years old (n = 105), miRNA494 level was negatively correlated to mRS score after 3 months (p = 0.041). (c) ROC curve of miRNA-494 to predict 3 months outcome in age 65 years (n = 105). Abbreviation: mRS, modified Rankin Score.

Following univariate analysis, we combined miRNA-494 levels with other clinical indicators to determine whether the predictive power remained reliable when accounting for additional factors. LASSO regression was employed to screen potential variables from 24 candidates, selecting miRNA-494 levels in lymphocytes, NIHSS score, mRS score on admission, SBP, PLT, and LDL as covariates for sensitivity analysis. As a continuous variable, miRNA-494 alone was statistically significant in predicting AIS prognosis, and the inclusion of SBP, PLT, and LDL did not significantly influence the results (Table 2, OR = 3.165 [95% CI: 1.222–8.199], p = 0.018). However, when including the NIHSS score, the predictive effect of miRNA-494 was less stable (Table 2, OR = 2.565 [95% CI: 0.803–8.195], p = 0.112). Nevertheless, as a categorical variable, miRNA-494 $>$1.057 not only maintained its independent predictive ability but also proved useful for predicting excellent outcomes when the model was adjusted for other clinical data (Table 2, OR = 2.901 [95% CI: 1.031–8.159], p = 0.044).

We further analyzed whether miRNA-494 and miRNA-494 $>$1.057 could serve as predictive factors in patients with or without recanalization therapy. In patients who did not undergo thrombolysis or mechanical thrombectomy, high levels of miRNA-494 significantly predicted excellent outcomes after 3 months, with OR values of 6.458 (95% CI: 1.871–31.922, p = 0.009), 3.449 (95% CI: 1.303–10.987, p = 0.022), and 8.938 (95% CI: 2.123–62.910, p = 0.010) (Table 3). Additionally, miRNA-494 $>$1.057 as a categorical variable also demonstrated predictive efficacy among patients not receiving either therapy (OR = 4.800 [95% CI: 1.551–16.234], p = 0.008; OR = 2.833 [95% CI: 1.147–7.439], p = 0.028; OR = 5.200 [95% CI: 1.480–20.733], p = 0.013) (Table 3). However, in patients who underwent either recanalization treatment, the univariate logistic regression analysis of miRNA-494 for the 3-month mRS did not yield significant results (Table 3).

The relationship between miRNA-494 levels and admission NIHSS scores in 205 patients was analyzed using statistical methods and correlation analysis. In terms of severity, miRNA-494 levels showed a general downward trend in patients with more severe strokes (Fig. 5a). Based on the distribution of miRNA-494 levels and stroke severity, combined with clinical criteria and previous studies, the 205 patients with stroke were classified into minor-to-moderate and severe stroke groups, with an NIHSS score threshold of 15 [35, 36]. miRNA-494 levels in peripheral lymphocytes were significantly higher in patients with minor-to-moderate strokes compared with those with severe strokes (Fig. 5b, OR = 1.10 [95% CI: 0.74–1.42] vs. OR = 0.76 [95% CI: 0.53–1.29], p = 0.026). To establish a reliable association between miRNA levels and stroke severity, logistic regression and LASSO regression were applied to the 205 patients. miRNA-494 levels, coronary heart disease, atrial fibrillation history, PLT on admission, and NLR were identified as key variables. A multivariate logistic regression analysis using these variables showed that elevated miRNA-494 levels independently indicated minor-to-moderate stroke (Table 4, OR = 2.586 [95% CI: 1.024–6.533], p = 0.044).

Fig. 5.

Analysis of stroke-severity of patients with AIS. (a) Division by interquartile range, under 4 classifications of severity according to NIHSS score on admission, the distribution of miRNA-494 levels percentiles of AIS patients (n = 205). (b) miRNA-494 level of two groups, cutoff by 15 score of NIHSS, Mann-Whitney U test, p = 0.026. (c) ROC curve of miRNA-494 levels to indicate NIHSS score on admission (n = 205). Abbreviation: ROC, receiver operating characteristic.

We also calculated the predictive value of miRNA-494 levels for minor-to-moderate strokes using ROC curve analysis. When miRNA-494 was treated as a dichotomous variable with a cutoff of 0.858, it had a sensitivity of 68.7% and a specificity of 61.5% for predicting minor-to-moderate stroke (Fig. 5c, AUC = 0.635 [0.786–0.615]). The association between dichotomous miRNA-494 levels and stroke severity was further evaluated using multivariate logistic regression. The OR for miRNA-494 levels $>$0.858 was 3.514 (1.501–8.230, p = 0.004), and after adjusting for other variables, miRNA-494 $>$0.858 remained a significant predictor of stroke severity (Table 4, OR = 3.881 [95% CI: 1.403–10.738], p = 0.009).