Circulating Circular RNAs as Biomarkers for the Diagnosis and Prediction of Outcomes in Acute Ischemic Stroke

Patients with suspected ischemic stroke were recruited within 72 hours of symptom onset from the Neurology and Emergency Departments of Zhongda Hospital and Jiangsu Province Hospital. The enrollment period was November 2017 to February 2019. Blood samples were collected immediately on admission before any treatment. All patients had a final diagnosis of ischemic stroke as defined by an acute focal neurological deficit in combination with a diffusion-weighted imaging-positive lesion on magnetic resonance imaging or a new lesion on a delayed computed tomography scan. Healthy controls (HCs) were recruited from the physical examination centers of 2 hospitals. For all samples involved in our research, we excluded AIS patients and HCs with active malignant diseases or neurological and psychiatric diseases, those who underwent surgery within the last 3 months, and those who took prior medication with low molecular weight or unfractionated heparin within the last month. After inclusion, clinical and radiological data were collected on standardized forms. Vitals were collected at hospital admission.

The characteristics and baseline information of 3 cohorts are listed in Table I in the online-only Data Supplement. A circRNA microarray analysis involving 10 798 circRNAs was conducted in the discovery cohort (GEO accession number: GSE133768; Figure 1A). Next, 68 downregulated circRNAs and 10 upregulated circRNAs (fold change > 4, P<0.05) were validated in an independent validation cohort using quantitative polymerase chain reaction (Table II and Table III in the online-only Data Supplement). Among them, 3 circRNAs were differentially regulated and showed consistent directionality with the results of the circRNA microarray (Figure I in the online-only Data Supplement). For the replication cohort, the copy numbers per microliter of plasma were calculated based on the Ct values. Three circRNAs (circFUNDC1 [circular RNA FUNDC1], circPDS5B [circular RNA PDS5B], and circCDC14A [circular RNA CDC14A]) showed significantly elevated expression levels in AIS patients compared with HCs (Figure 1B). Next, receiver operating characteristic curves were constructed to compare the copy numbers of circRNAs in AIS patients and HCs (Figure 1C). The combination of 3 circRNAs resulted in an area under the curve of 0.875, corresponding to a specificity of 91% and a sensitivity of 71.5% in AIS diagnosis. Moreover, there is positive correlation between circRNA levels and cerebral infarct volume (Figure 1D).

Next, we further explored the significance of circRNA levels in predicting stroke outcome. Poor significance was observed merely depending on the baseline levels of circRNAs (Figure II in the online-only Data Supplement). Because there were opposite variation tendencies in stroke patients with good outcomes (Figure 2A) versus poor outcomes (Figure 2B), we further explored the association between the changes of circRNA levels within the first 7 days of hospitalization and stroke outcome. Receiver operating characteristic curves were calculated based on the change in values of the circRNAs between admission and the seventh day after admission to predict outcome (Figure 2C), calculated an area under the curve at 0.941 for the combination of 3 circRNAs. Furthermore, stroke outcome was predicted by the changing rates of circRNAs that further increased the predictive significance of 3 circRNAs with area under the curve at 0.960 (Figure 2D).

No significant increase of circFUNDC1 levels was found in platelets, lymphocytes, or granulocytes for AIS patients compared with HCs (Figure 3A). There was no significant change in platelets between stroke patients and controls, whereas the levels of circPDS5B in lymphocytes and granulocytes were significantly higher in stroke patients than in controls and showed consistent tendencies with the levels in plasma (Figure 3B). For the level of circCDC14A (Figure 3C), there is significant increase in granulocytes but not in platelets and lymphocytes.

Pathophysiological processes that might be controlled by circRNAs after stroke according to bioinformatics analysis were listed in Figure III in the online-only Data Supplement.