Proteomic profiling of plasma biomarkers in acute ischemic stroke due to large vessel occlusion

An overview of the study design is shown in Fig. 1. Briefly, this study is a single-center, retrospective, observational case–control trial based on a prospectively collected stroke data base. Consecutive patients who presented at the Department of Neurology of Tongji Hospital were recruited from July 2016 to April 2018. The clinical diagnosis of acute ischemic stroke was confirmed by cranial magnetic resonance (MR) examinations with diffusion weighted imaging (DWI), and intracranial large vessel status was evaluated by digital subtraction angiography (DSA) within 7 days in every patient. Study-specific inclusion criteria were the presence of intracranial carotid artery terminus or first middle cerebral artery segment (M1) occlusion as confirmed on DSA, and sufficient display of the middle cerebral artery region. Study-specific exclusion criteria were the administration of thrombolysis and thrombectomy, and baseline modified Rankin Scale (mRS) > 1 before stroke onset. Patients with baseline mRS of 0–1 who missed therapeutic time window or refused acute reperfusion therapy were enrolled to rule out the influence from baseline clinical differences and acute reperfusion therapy. Collateral status was not an entry criterion for the trial and was assessed using the American Society of Interventional and Therapeutic Neuroradiology/Society of Interventional Radiology (ASTIN/SIR) grading system by blinded observers. Healthy volunteers with no prior history of stroke were matched with patients with AIS due to LVO for age and sex, and were recruited as a healthy control group, to control for the confounding effects of additional risk factors. Cranial MR with DWI and angiography were performed in healthy volunteers to verify that no silent brain infarcts or artery stenosis were present. Patients with AIS in the middle cerebral artery region but no obstructive large vessels were matched with patients with AIS due to LVO for age and sex, and enrolled as AIS due to non-LVO group.

In the discovery phase, there were 20 volunteers in the healthy control group and 40 patients with stroke due to LVO in the AIS group. 10 individual samples of equal volume from each group were pooled together for the proteomics analysis. In the validation phase, another independent population that included 33 healthy controls, 33 patients with AIS due to non-LVO, and 33 patients with AIS due to LVO were enrolled for Western blots analysis. The patients with AIS due to LVO were further divided into two groups based on their ASTIN/SIR grades: AIS with good collaterals (SGC, ASTIN/SIR 3–4), and AIS with poor collaterals (SPC, ASTIN/SIR 0–2). For prognostic prediction, patients with AIS due to LVO were divided into those with good prognosis (mRS 0–2) and those with poor prognosis (mRS 3–6) at a 3-month follow-up. Table 1 provides demographic details of the study participants.

A detailed description of experimental procedures was provided in Additional file 1.

Statistical analyses were performed using SPSS 19.0 for Windows. The continuous variables were described by mean ± SD or by medians and interquartile ranges (IQR), while categorical variables were described by percentages. Mann–Whitney U test, one-way analysis of variance (ANOVA) with Dunnett’s post hoc test were used to compare differences between groups. Adjusted logistic regression models were used to calculate the odds ratio (OR) and 95% confidence intervals (CI). Covariates included in the adjusted logistic regression models were determined by univariate logistic regression of every factor shown in Table 1. Logistic analysis and receiver operating characteristic (ROC) curves were employed to assess the diagnostic and prognostic accuracy of each biomarker. A combined ROC analysis was performed for panel diagnosis. All tests were considered statistically significant at P < 0.05.

Baseline characteristic details of the patients and controls participating in the study are summarized in Table 1. For the discovery phase study, 40 patients with AIS due to LVO attending our hospital between July 2016 and April 2017 and 20 healthy control subjects were enrolled (Fig. 1). Compared with healthy controls, patients with AIS due to LVO had higher systolic blood pressure and homocysteinemia (HCY) levels. There were no significant differences in age, sex, medical history, or laboratory data between the two groups (Table 1).

For the validation phase study, 33 healthy controls, 33 patients with AIS due to non-LVO and 33 patients with AIS due to LVO were prospectively enrolled from April 2017 to April 2018 (Fig. 1). Higher systolic and diastolic blood pressure, higher HCY levels, higher hypertension and diabetes mellitus rates, as well as more patients with smoking history, were found in the AIS group than in the control group. Patients with AIS were divided into two subgroups: 33 patients with AIS due to LVO and 33 patients with AIS due to non-LVO. Lower low-density lipoprotein level was found in the LVO group. Patients with AIS due to LVO were further divided into two subgroups based on their ASTIN/SIR grades: 17 patients with good collaterals (SGC) and 16 patients with poor collaterals (SPC). There were no significant differences in demographic characteristics or laboratory data between the SGC and SPC groups (Table 1).

We first compared plasma protein profiles from 40 consecutive patients with AIS due to LVO and 20 healthy controls. Pooled plasma samples were analyzed in replicates using 6-plex iTRAQ with LC–MS/MS. Patients with AIS due to LVO had a specific proteomic signature. The molecular features that were detected are depicted in the heatmap (Additional file 1: Figures S1 and S2). Seven differentially expressed proteins, thrombospondin 1 (THBS1), secreted phosphoprotein 2 (SPP2), lymphatic vessel endothelial hyaluronic acid receptor 1 (LYVE1), cadherin 1 (CDH1), insulin-like growth factor 2 (IGF2), pro-platelet basic protein (PPBP), and apolipoprotein C 4-2 (APOC4-APOC2), were finally acquired with a 1.2/0.83-fold or greater difference in abundance in AIS samples compared to healthy controls (Additional file 1: Table S1). Of these seven proteins, THBS1, LYVE1, IGF2, and PPBP were more abundant in AIS due to LVO samples, while SPP2, CDH1, and APOC4-APOC2 were downregulated compared to control samples.

Interestingly, among these differentially expressed proteins identified in plasma samples of AIS due to LVO, all the upregulated proteins, IGF2, LYVE1, PPBP, and THBS1, are known to be involved in regulating blood hemostasis and endothelial function [6‐11]. We then verified the differential abundance of four selected proteins (IGF2, LYVE1, PPBP, and THBS1) using Western blots analysis. Consistent with the tendency found in proteomics, Western blots results showed similar alterations in an independent population, as the levels of those four identified proteins were significantly higher in the AIS due to LVO group compared to the healthy control group (Fig. 2, Tables 2 and 3). To describe the specificity for AIS due to LVO, we then detected the levels of the four identified proteins in patients with AIS due to non-LVO. Interestingly, expression of IGF2, LYVE1 and PPBP showed no significant difference between AIS due to non-LVO and healthy control, except that THBS1 was higher in AIS due to non-LVO group. Further, comparisons between the non-LVO and LVO groups presented higher expression of IGF2, LYVE1, PPBP and THBS1 in the LVO group (Fig. 2, Table 2). Altered levels of these proteins might indicate endothelial dysfunction and clinical implications of large vessel occlusion.

Patients with AIS due to LVO were further divided into SGC and SPC subgroups, according to the ASTIN/SIR grades assigned to their collaterals. Comparisons between the SGC and SPC group indicated higher expression of LYVE1 and IGF2 in the SGC group and higher level of THBS1 in the SPC group, but no significant difference in PPBP (Fig. 2, Table 4).

In order to define the capacity of these potential biomarkers for diagnosis, logistic regression and receiver operating characteristic (ROC) curves were employed. Logistic analysis results supported the role of plasma levels of IGF2, LYVE1, PPBP, and THBS1 as potential biomarkers that indicate a significantly increased risk of AIS due to LVO (Table 3).

We then assessed the diagnostic efficiency of the potential biomarkers using ROC analysis. The areas under the ROC curve (AUC) were 0.731 (0.609–0.853) for IGF2, 0.813 (0.707–0.919) for LYVE1, 0.754 (0.638–0.870) for PPBP, and 0.769 (0.652–0.885) for THBS1, respectively (Fig. 3a, Table 3). Furthermore, the diagnostic potential of the 4-protein panel was also evaluated. The AUC value of the panel reached 0.947 for distinguishing patients with AIS due to LVO from healthy individuals, which was interpreted as excellent classification performance (Fig. 3b).

We further investigated the association between plasma levels of these identified potential biomarkers and the prognosis of AIS due to LVO. Patients were divided into those with good prognosis (modified Rankin scale, mRS 0–2) and those with poor prognosis (mRS 3–6) at a 3-month follow-up. Logistic analysis results showed that higher IGF2 and LYVE1 levels predicted favorable outcomes, while higher THBS1 levels were related to a poor prognosis. The results implicated that the three collateral-related proteins (IGF2, LYVE1, and THBS1) were prognostic for patients with AIS due to LVO, while PPBP showed no potential for outcome prediction (Table 4).