Proteomic and Demographic Comparisons of Recurrent Ischemic Stroke Patients

Abstract

Rates of recurrent strokes have remained relatively unchanged over the past couple decades, highlighting a need for advancements in secondary prevention of stroke recurrence. This study utilizes the Blood And Clot Thrombectomy Registry And Collaboration (BACTRAC) tissue bank to identify proteomic and demographic differences in recurrent ischemic stroke patients. Blood samples were collected during mechanical thrombectomy of large-vessel occlusion ischemic strokes. Plasma levels for 184 inflammatory and cardiometabolic proteins were measured in systemic blood and intracranial blood from the infarction area. Differences between recurrent and first-stroke patients were analyzed using Fisher’s Exact Test for categorical variables and Student’s independent samples t tests or Welch’s t tests for continuous variables. Proteins were divided into systemic and intracranial proteins, and independent samples t tests were performed with a False Discovery Rate of 5.0%. Significant variables were used in multiple logistic regression. There were 20 patients in the prior stroke group and 121 in the first stroke group. The prior stroke group had a significantly higher percentage of females (80.0% vs 50.4%, p = 0.016) and lower rate of hyperlipidemia comorbidity (10.5% vs 35.5%, p = 0.034). Two systemic proteins were significantly higher in those with a prior stroke: CCL14 and FGF-19. Multiple logistic regression found higher levels of CCL14 and FGF-19 to be predictive of a stroke being recurrent. Along with other demographics, these proteins could provide a predictive model to identify patients with risk of recurrent ischemic strokes. Serum CCL14 and FGF-19 levels are easily accessible biomarkers, making them possible therapeutic targets for recurrent stroke prevention.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12975-025-01353-1.

Introduction

Ischemic stroke is the third leading cause of combined death and disability worldwide. Over the last couple of decades, the prevalence of stroke has increased by 85%, and 62% of all new strokes were ischemic strokes in 2019 [1]. Emergent large vessel occlusion (ELVO) strokes are an important subset of ischemic strokes defined as a cerebral vascular occlusion that results in significant clinical deficit and can be accessed for endovascular mechanical thrombectomy [2]. Mechanical thrombectomy (MT), which involves the retrieval of the clot endovascularly, is the preferred management of ELVOs within 6 h of stroke onset. The introduction of this therapy in the last couple of decades has resulted in improved outcomes and mortality rates of ischemic stroke patients [3, 4]. Due to improved acute stroke interventions and diagnostic neuroimaging, more patients are surviving ischemic stroke—age-standardized mortality decreased by 36.0% from 1990 to 2019 [1]. This improved survival increases the population at risk of a recurrent stroke. According to data from 2016, 23% of the annual strokes in the USA are recurrent events [5]. Despite the improvements in stroke management and decreases in mortality, the rates of recurrence have remained relatively unchanged over the past couple of decades [6]. This population and high rates of recurrence highlight a need for secondary prevention of stroke recurrence, an area in which there have been limited advancements.

There are several established models which can be used to estimate a patient’s risk of stroke for secondary prevention, including the ABCD2, CHA2DS2-VASc Score, and Framingham Stroke Risk Profile. While the ABCD scoring system provides some predictive value after a transient ischemic attack (TIA), there are not well-established risk factors or models which can reliably predict stroke recurrence [7–9]. Despite the lack of a predictive model, several individual demographics and comorbidities have been associated with recurrent stroke, such as hypertension, diabetes mellitus, atrial fibrillation, and coronary artery disease [10]. Another effort for prevention has been on identifying serum biomarkers predictive of recurrent strokes; these markers would serve to both identify patients at higher risk and could also provide a potential therapeutic target to decrease the risk. Inflammation, thrombosis, and oxidative stress are key processes in ischemic stroke and have been the focus of many such studies. However, the studied biomarkers from systemic blood samples have had weak predictive value and little to no clinical significance, including IL- 6, CRP, fibrinogen, and d-dimer [11–13].

The Blood And Clot Thrombectomy Registry And Collaboration (BACTRAC) tissue bank collects systemic and intracranial blood samples from only patients undergoing MT for ELVO. Plasma from these blood samples are analyzed for inflammatory and cardiometabolic proteins. BACTRAC has been used in previous studies to investigate changes in proteomics and other biomolecules during stroke to determine their association with clinical outcomes [14–25]. This study aims to utilize the BACTRAC data bank to identify proteomic and demographic differences in patients suffering from a recurrent ischemic stroke compared to a first stroke. These differences can lead to the discovery of predictors to identify patients at risk of recurrent stroke.

Methods

Sample Collection

The study was approved by the University of Kentucky IRB. Tissue samples for this study were collected for the BACTRAC tissue bank (ClinicalTrials.gov ID NCT03153683) using methods previously described [26]. Briefly, tissue was collected during MT procedures on patients suffering from large-vessel occlusion ischemic strokes, and consent was then obtained from the patient or a representative. During the procedure, a microwire and microcatheter were navigated through the thrombus to collect 1 mL of intracranial arterial blood distal to the thrombus; this intracranial blood sample provides proteomic data from within the area of infarction. Additionally, 10 mL of systemic arterial blood is collected peripherally from the arterial access site, providing proteomic data from the systemic circulation. Blood samples are collected from both systemic and intracranial circulation distal to the clot in order to provide data systemically as well as proteomics specific to the area of infarction during the hyperacute phase of a stroke. These samples were aliquoted into 500 μL samples in BD Microtainer tubes with K2E (K2EDTA; Becton, Dickinson and Company), inverted 10 times, and then placed in a centrifuge for 15 min at 2000 g. Following centrifugation, a 50 µL plasma sample was removed from both the intracranial and systemic blood samples. The remaining plasma was removed and stored in Wheaton CryoELITE cryogenic vials (DWK Life Sciences; Millville, New Jersey) in a − 80 °C freezer. The 50 µL samples of plasma were shipped on dry ice to Olink Proteomics (Olink Proteomics, Boston, MA, USA), where the samples were analyzed for inflammatory and cardiometabolic proteins in both the systemic and intracranial samples.

Population

All subjects collected by BACTRAC to date at the start of the study were included in the present study. Males and females ≥ 18 years old who met the criteria for thrombectomy and for whom appropriate consent was obtained were enrolled in BACTRAC; females of reproductive age must have a negative pregnancy test to be included. This data bank enrolled subjects from a single, comprehensive stroke center. Subjects were categorized into two groups based on a history of stroke. The first stroke group were patients who had no history of stroke. The prior stroke group had a history of at least one stroke, and the data captured by BACTRAC was from the recurrent stroke. BACTRAC also collects data on control patients; these are patients undergoing a diagnostic cerebral angiogram for cerebrovascular disease which is not stroke.

Measures

The variables in this analysis include proteomic, demographic, and clinical measures, all collected as part of the BACTRAC protocol. Plasma levels of 184 proteins in systemic blood and intracranial blood were measured for each patient through Olink Proteomics’ inflammatory and cardiometabolic panels. Demographic variables included basic patient information and comorbidity risk factors for ischemic stroke, including age, sex, BMI, hypertension, hyperlipidemia (HLD), diabetes, history of myocardial infarction, and smoking status. Clinical variables included were lab values at the time of admission (A1c, POC glucose, Glucose/A1c, thyroid-stimulating hormone (TSH), low-density lipoprotein (LDL), high-density lipoprotein (HDL), triglycerides (TG), and total cholesterol), stroke etiology, and functional stroke scores at the time of admission and discharge (modified Rankin Scale (mRS), NIH Stroke Scale (NIHSS), and Montreal Cognitive Assessment (MoCA)).

Statistical Analysis

Differences in categorical demographic and clinical variables between the prior stroke and first stroke groups were analyzed using Fisher’s Exact Test. Differences in continuous demographic and clinical variables were analyzed using Student’s independent samples t tests or Welch’s t tests when variances were unequal, as indicated by a significant F test. Non-parametric Mann–Whitney U tests were performed for variables with a non-normal distribution. Proteins were divided into systemic and intracranial proteins, and independent samples t tests were performed with a False Discovery Rate (FDR) for each group. The FDR was set to 5.0% using the two-stage step-up method of Benjamini, Krieger, and Yekutieli [27]. Significant proteins and demographics identified by these methods were then used in a multiple logistic regression to predict patients with a prior stroke, using α = 0.05 for statistical significance. Missing data was handled using listwise deletion. Finally, significant proteins identified by these methods were then compared to controls by one-way analysis of variance (ANOVA) and Tukey’s multiple comparison’s test. All data was analyzed, and figures were created using GraphPad Prism 10.2.3.

Results

Demographics

There were 20 patients in the prior stroke group and 121 patients in the first stroke group. The demographic and clinical variables are summarized in Table 1. The prior stroke group had a significantly higher percentage of females than the first stroke group (80.0% vs 50.4%, odds ratio (95% CI): 3.93 (1.30–11.26), p = 0.016). The prior stroke group also had a significantly lower rate of HLD comorbidity (10.5% vs 35.5%, odds ratio (95% CI): 0.21 (0.05–0.84), p = 0.034).

Table 1.

Demographics of included patients. Continuous variables are reported as mean (SD) where a parametric test was used and median (interquartile range) where a non-parametric test was used. Categorical variables are reported as n (%)

Characteristic	Prior stroke (n = 20)	First stroke (n = 121)	p value
Age	67.0 (59.3–77.3)	69.0 (58.5–76.0)	0.804
Female sex	16 (80.0%)	61 (50.4%)	0.016
Body mass index	27.2 (4.9)	28.6 (6.3)	0.334
Hypertension	17 (85.0%)	93 (76.9%)	0.565
Hyperlipidemia	2 (10.5%)	43 (35.5%)	0.034
Atrial fibrillation	7 (38.9%)	47 (39.5%)	> 0.999
Diabetes mellitus type 1	0 (0.0%)	3 (2.5%)	> 0.999
Diabetes mellitus type 2	8 (42.1%)	35 (29.2%)	0.290
Previous myocardial infarction	0 (0.0%)	9 (7.4%)	0.610
Smoking status			0.059
Current	5 (25.0%)	34 (28.0%)
Previous	6 (30.0%)	12 (9.9%)
Never	7 (35.0%)	60 (49.6%)
A1c	5.7 (5.3–6.1)	5.8 (5.4–6.7)	0.417
POC glucose	113.0 (101.0–133.0)	122.0 (102.0–147.0)	0.448
Glucose/A1c	21.1 (17.7–23.3)	20.9 (18.4–24.4)	0.783
Thyroid stimulating hormone	2.0 (0.7–2.9)	1.6 (0.9–2.7)	0.767
Low density lipoprotein	64.6 (21.7)	69.4 (28.2)	0.518
High density lipoprotein	35.5 (28.3–41.0)	38.5 (32.0–50.8)	0.272
Triglyceride	116.0 (92.0–182.0)	115.0 (84.0–172.0)	0.416
Total cholesterol	129.4 (27.4)	136.4 (35.9)	0.455
Stroke etiology			0.845
Cardioembolic	11 (55.0%)	67 (55.3%)
Atheroembolic	2 (10.0%)	21 (17.3%)
Intracranial stenosis	1 (5.0%)	4 (3.3%)
Dissection	0 (0.0%)	3 (2.4%)
Carotid occlusion	0 (0.0%)	1 (0.8%)
Infection	1 (5.0%)	4 (3.3%)
Unknown	5 (25.0%)	24 (19.8%)
mRS at discharge			0.749
0	0 (0.0%)	11 (11.0%)
1	1 (5.6%)	10 (10.0%)
2	1 (5.6%)	10 (10.0%)
3	1 (5.6%)	5 (5.0%)
4	7 (38.9%)	28 (28.0%)
5	5 (27.8%)	25 (25.0%)
6	3 (16.7%)	11 (11.0%)
NIHSS at admission	17.0 (11.0–20.0)	16.5 (10.0–22.0)	0.696
NIHSS at discharge	12.0 (2.5–19.5)	6.0 (1.0–14.0)	0.135
MoCA at discharge	20.0 (15.0–22.6)	22.5 (15.0–25.8)	0.577

Significant p values are bolded

Proteomic Differences

Protein levels were analyzed for 184 proteins in systemic and intracranial blood and are reported within this study as a Normalized Protein eXpression (NPX). Mean and standard deviation for all proteins are reported in the spreadsheet (Supplement 1). Two proteins were found to be significantly different in those with a previous versus first stroke after correction for FDR: systemic CCL14 and systemic FGF- 19 (Figs. 1–2). The difference in CCL14 level means was 0.505 (0.134) (difference (SE); q = 0.022). The difference in FGF- 19 level means was 0.863 (0.227) (q = 0.022). For both proteins, serum levels were higher in the previous stroke group. Multiple logistic regression performed with age, sex, HLD comorbidity, CCL14 levels, and FGF- 19 levels found CCL14 and FGF- 19 levels to be significant. Higher levels of CCL14 and FGF- 19 were associated with a greater likelihood of the stroke being a recurrent stroke. These results are summarized in Table 2.

Fig. 1.

Systemic plasma levels of chemokine ligand 14 (CCL14) in prior versus first stroke patients (* indicates a q value of < 0.05)

Fig. 2.

Systemic plasma levels of fibroblast growth factor 19 (FGF- 19) in prior versus first stroke patients (* indicates a q value of < 0.05)

Table 2.

Multiple logistic regression of selected variables. “Sex(Male)” indicates male sex is the reference condition. CCL14 and FGF- 19 were significant predictors of patients with a recurrent stroke

Variable	Odds ratio	95% confidence interval	Variance inflation factor (VIF)	p value	Significance
Age	0.994	0.959–1.032	1.080	0.733	ns
Sex (male)	2.773	0.809–11.47	1.079	0.123	ns
Hyperlipidemia	0.208	0.030–0.879	1.038	0.058	ns
CCL14	4.102	1.448–12.780	1.123	0.010	*
FGF- 19	1.923	1.097–3.617	1.080	0.030	*

There were also significant differences by ANOVA for both proteins on comparison to controls (p < 0.05). For CCL14, the mean expression of prior stroke patients was 0.505 higher than first stroke patients (p = 0.003) and 0.672 higher than controls (p < 0.0001). There were no significant differences between first stroke patients and healthy controls (p = 0.137). For FGF- 19, the mean expression of prior stroke patients was 0.863 higher than first stroke patients (p = 0.006) and 0.704 higher than controls (p < 0.035). There were no significant differences between first stroke patients and controls (p = 0.569).

Discussion

The primary objective of this study was to identify proteomic and demographic differences in patients experiencing their first versus recurrent stroke. Recurrent strokes constitute a substantial portion of ischemic strokes. Currently, there are no reliable predictors to identify which patients are at risk of a recurrent stroke, nor are there effective interventions or therapeutics to prevent these events. In this study, sex, HLD comorbidity, serum CCL14 levels, and serum FGF- 19 levels were significantly different in patients with recurrent stroke compared to those experiencing a first stroke. These proteins and demographics are potential predictors of recurrent stroke and warrant further investigation.

A primary risk factor for ischemic stroke is the presence of atherosclerosis. In a meta-analysis of recurrent strokes, large artery atherosclerosis and cardioembolic stroke etiologies had the highest rate of recurrent strokes, implicating atherosclerosis as an important process in the risk of recurrent strokes [6]. The threat of atherosclerosis for ischemic stroke occurs when the plaque becomes unstable and is considered vulnerable, meaning it is at risk of rupturing and forming an embolus. Intraplaque neovascularization and hemorrhages play a major role in the progression of plaques to this vulnerability [28, 29]. Chemokine ligand 14 (CCL14) is a chemokine involved in the activation of immune cells and may have a role in plaque neovascularization, making the finding of significantly elevated CCL14 in recurrent stroke patients important in this study. Analysis of carotid plaque samples from vulnerable and stable plaques has shown significantly increased levels of CCL14 and VEGF-A in vulnerable plaques [30]. Additionally, CCR5, the receptor for CCL14, has also been shown to have a role in the progression of atherosclerotic plaques [31]. CCL14/CCR5 may have an influence on neovascularization through the action of vascular endothelial growth factor A (VEGF-A). The Janus kinase 2 (JAK2)-VEGF signaling pathway is well established as a regulator of new blood vessel growth, and evidence exists that CCL14 promotes JAK2 phosphorylation and the activation of this pathway [30]. Of note, VEGF-A was also included in this study but had no significant findings (data not shown).

The elevated FGF- 19 in the prior stroke group may also be related to the process of atherosclerosis. FGF- 19 is associated with the development of type 2 diabetes mellitus (T2DM), and higher levels of FGF- 19 have further been associated with the development of subclinical atherosclerosis, the cause of diabetic macroangiopathy, in patients with T2DM [32]. However, there is no direct evidence of a relationship between FGF- 19 and stroke. Fibroblast growth factors (FGF) are important regulators of cell growth and differentiation; FGF- 19 is part of the endocrine subfamily and regulates the synthesis and metabolism of bile acid. The finding of elevated serum FGF- 19 levels in the recurrent stroke group may, therefore, be related to the difference in proportions of HLD, a known risk factor for ischemic stroke [33]. Hyperlipidemia refers to high levels of LDL or triglycerides and is a key contributor to the development of atherosclerosis. The exact relationship between hyperlipidemia and FGF- 19 is unclear due to inconsistent evidence. In obese mice, FGF- 19 has demonstrated suppression of fatty acid synthesis, suggesting it could reverse hyperlipidemia and associated diseases [34]. In contrast, other evidence suggests that FGF- 19 can increase triglyceride and cholesterol concentrations, which may be due to actions on different receptors [35]. Despite the possible tie between FGF- 19 and HLD, the lower rates of HLD in the prior stroke group in this study contradict other evidence in the literature, which have found no association between HLD and risk, either increased or decreased, of recurrent stroke [10].

A significantly higher proportion of females in the prior stroke group was also found by this study. Evidence in the literature conflicts this finding. Multiple larger studies have demonstrated there is no association between sex and the risk of a recurrent ischemic stroke [36–38]. Although, women do have a higher lifetime incidence of stroke, with menopause being a known contributor to the risk of stroke, which may contribute to the sex differences in the current study [39]. The differences in sex may be dependent on the study population and stroke etiologies included as well; for patients with high-grade carotid stenosis, male sex has been demonstrated as a significant predictor [40]. The present study, however, found no differences in stroke etiology for the prior stroke group (Table 1). Sex differences in the risks of recurrent stroke may also be partly attributable to risk factors such as hypertension, diabetes, and obesity, which varies between males and females [41]. Further, there are sex-specific risk factors for ischemic stroke including pregnancy hypertensive disorders, oophorectomy, orchiectomy, and androgen deprivation therapy [42]. The potential differences in rates of recurrent strokes based on sex is a complex question with many confounding variables. With only four males in the prior stroke group, analysis to explore some of these potential confounding factors could not be performed.

The current study identified serum CCL14 and FGF- 19 levels as significantly elevated in and predictive of patients with a prior stroke. Both proteins may play a role in atherosclerosis, as discussed, and inflammation, as both are known immune regulators. Importantly, scores of low-grade inflammation have been associated with a higher risk of stroke recurrence [43]. Therefore, targeting these processes through these proteins is a potential option for future therapeutics to reduce the risk of a recurrent stroke. Additionally, CCL14 and FGF- 19 may be used individually or in combination with other demographics to predict which patients are at risk of a recurrent stroke. Since there is no current method to identify these patients, this study helps build a framework for a model which can reliably predict patients at risk. Successful identification of these patients can reduce the incidence of recurrent strokes through identifying which patients need prevention. Further research is warranted before CCL14 and FGF- 19 are used for therapeutic targets and before these variables can be used to predict patients who will have a recurrent stroke.

The most notable limitation of this study is the small prior stroke group. Five variables were included for the logistic regression from a small population of 20 patients with a recurrent stroke, making the model underpowered and overfitted. This underpowering is evident by the wide confidence intervals, representing low precision of the results. Use of an underpowered logistic regression still allows identification of significant variables, but the ROC curve becomes inaccurate with an inflated area under the curve (AUC). For this reason, the ROC curve and AUC were excluded from the results. The small prior stroke group also prevented further analysis, such as differences between males and females. Additionally, the external validity of this study may be limited by generalizability: the current study population has rates of stroke and comorbidities (diabetes, hypertension, hyperlipidemia) higher than the US national rates. Future studies should address these limitations with larger, prospective studies.

Conclusion

In summary, our study used the BACTRAC study to identify demographic and proteomic differences in patients with a recurrent stroke receiving MT. Proportions of female sex and the absence of HLD comorbidity were significantly higher in patients experiencing a recurrent stroke, although these were not significant in the multiple logistic regression model. Previous studies have failed to discover a strong, reliable predictor of recurrent stroke, making this study useful in its exploratory design to evaluate a large number of plasma proteins. The expression of CCL14 and FGF- 19 in systemic blood obtained during thrombectomy are easily accessible and quantifiable biomarkers and are possible therapeutic targets for recurrent stroke. These findings provide two new proteins that, with other demographic variables, could provide a predictive model to identify patients with a high probability of recurrent ischemic strokes. However, due to the limitations and novel aspect of this study, further research is needed to confirm these findings and explore the pathophysiological mechanisms.

Supplementary Information

Below is the link to the electronic supplementary material.

Nonstandard Abbreviations

BACTRAC

Blood And Clot Thrombectomy Registry And Collaboration

ELVO

Emergent large vessel occlusion

Author Contribution

J.H., J.F.F., C.J.M., and K.R.P. performed clinical study design. M.N.A.-K., S.P., D.L.D. III, J.F.F., J.A.F., A.L.T., and A.M.S. collected data. N.M., J.A.F., C.J.M., and W.C. analyzed data. N.M. created figures and wrote the original draft. J.H., C.J.M., J.A.F., and K.R.P. edited the manuscript. All authors approved the final manuscript.

Funding

This work was supported by a research project grant from the National Institute of Neurological Disorders and Stroke (R01 NS127974). The funding organization had no role in the design or conduction of this study.

Data Availability

Data is available by request to the corresponding author, KRP. Protein expression data is available as a supplement to the manuscript.

Declarations

Ethics Approval

This study was approved by the University of Kentucky Institutional Review Board prior to the study.

Consent to Participate

Written consent was obtained from all individual participants, or a representative when applicable, for inclusion in the BACTRAC database.

Consent for Publication

Not applicable.

Conflict of Interest

Authors AMS, JFF, and KRP are co-founders/equity holders in Cerelux, LLC. The other authors declare no competing interests.