Transient ischemic attacks characterized by RNA profiles in blood

Abstract

Objective:

Transient ischemic attacks (TIA) are common. Though systemic inflammation and thrombosis are associated with TIA, further study may provide insight into TIA pathophysiology and possibly lead to the development of treatments specifically targeted to TIA. We sought to determine whether gene expression profiles in blood could better characterize the proinflammatory and procoagulant states in TIA patients.

Methods:

RNA expression in blood of TIA patients (n = 26) was compared to vascular risk factor control subjects without symptomatic cardiovascular disease (n = 26) using Affymetrix U133 Plus 2.0 microarrays. Differentially expressed genes in TIA were identified by analysis of covariance and evaluated with cross-validation and functional analyses.

Results:

Patients with TIA had different patterns of gene expression compared to controls. There were 480 probe sets, corresponding to 449 genes, differentially expressed between TIA and controls (false discovery rate correction for multiple comparisons, p ≤ 0.05, absolute fold change ≥1.2). These genes were associated with systemic inflammation, platelet activation, and prothrombin activation. Hierarchical cluster analysis of the identified genes suggested the presence of 2 patterns of RNA expression in patients with TIA. Prediction analysis identified a set of 34 genes that discriminated TIA from controls with 100% sensitivity and 100% specificity.

Conclusion:

Patients with recent TIA have differences of gene expression in blood compared to controls. The 2 gene expression profiles associated with TIA suggests heterogeneous responses between subjects with TIA that may provide insight into cause, risk of stroke, and other TIA pathophysiology.

Transient ischemic attacks (TIAs) are common, affecting over 300,000 people per year in the United States alone. Although TIA symptoms resolve by definition, they are far from benign, with a risk of stroke, myocardial infarction, and death as high as 25% within 90 days.^1–3 Better understanding of TIA pathogenesis is needed to improve our ability to treat this condition and potentially prevent stroke.

We have previously shown blood gene expression changes following experimental TIAs⁴ are associated with microglial activation and macrophage infiltration in brain.⁵ In an experimental rat model of TIA, we demonstrated differences in blood immune cell activation using whole genome expression analysis.⁴ In humans, TIAs have elevated levels of cytokines and leukocytes,^6,7 fibrinogen,⁸ d-dimer,⁹ and platelet collagen receptor glycoprotein VI.¹⁰ Furthermore, levels of high-sensitivity C-reactive protein (CRP) and Lp-PLA2 are associated with recurrent ischemic events in TIA.^11,12 Thus, experimental and human studies of brief cerebral ischemia document evidence of inflammatory and prothrombotic states.

Gene expression profiles derived from blood can distinguish ischemic stroke from controls and identify procoagulant and proinflammatory features important to stroke pathophysiology.^13–15 Whether such profiles exist in patients with acute TIA is unknown. Accordingly, in this study, we sought to determine whether gene expression profiles in blood could characterize the inflammatory and coagulant states in patients with acute TIA.

METHODS

Standard protocol approvals, registration, and patient consents.

Patients with TIA and controls were recruited from the University of California Davis Medical Center, University of California San Francisco Medical Center, and Wake Forest University Health Sciences. The institutional review board at each institution approved the study. Informed consent was obtained from all patients.

The diagnosis of TIA was made by board-certified neurologists. TIA was defined as an acute loss of focal cerebral or ocular function of vascular etiology lasting <24 hours. Neuroimaging was obtained on all TIAs, with 85% having MRI. Controls were persons with vascular risk factors but without symptomatic vascular disease, including no prior stroke, TIA, myocardial infarction, angina, or peripheral arterial disease. They were matched to patients with TIA based on age and gender.

Blood collection and RNA isolation.

Blood was collected from 9.5 to 68.6 hours (mean 35.1 hours) from TIA symptom onset. A venous blood sample (15 mL) was collected into PAXgene Vacutainers (PreAnalytiX, Hilden, Germany) from each subject. PAXgene Vacutainers immediately stabilize blood RNA, thus reducing RNA degradation and changes of gene expression after phlebotomy. After 2 hours at room temperature, PAXgene Vacutainers were frozen at −70°C until processed. Total RNA was isolated according to the manufacturer's protocol (PreAnalytiX) and processed together at the same time and site. The RNA from whole blood is from white blood cells (neutrophils, basophils, eosinophils, lymphocytes, and monocytes), immature platelets, and immature red blood cells.

Microarray hybridization.

RNA was analyzed using an Agilent 2100 bioanalyzer for quality and Nano-drop for concentration. Biotin-labeled cDNA was synthesized from 50 ng of total RNA using the Ovation Whole Blood Solution (Nugen Technologies, San Carlos, CA) kit according to protocol. Each RNA sample was processed on Affymetrix Human Genome U133-Plus-2.0 microarrays (Affymetrix Santa Clara, CA).¹⁵

Statistical analysis.

Microarray data were normalized using Robust Multichip Average and log2 transformed. Individual probes were summarized into probe sets as described by Affymetrix, and used for probe set and gene level analysis. Analysis of covariance was conducted in Partek Genomics Suite 6.5 (Partek Inc., St. Louis, MO) to identify probe sets significantly different between patients with TIA and controls adjusting for age and microarray batch effect. Multiple comparison correction was performed using Benjamini-Hochberg false discovery rate (FDR) of 5%. Probe sets were considered significant with an absolute fold change ≥1.2 and a p value ≤0.05 after the FDR correction for multiple comparisons.

Differences in demographic and clinical data between groups were analyzed using Fisher exact test or t test. Prediction analysis was performed using linear discriminant analysis (LDA). Cross-validation was performed by randomly dividing subjects into 10 equal groups, developing a prediction model using 90% of the patients, and evaluating the model on the remaining 10%. This process was repeated 10 times to estimate the overall prediction ability of the developed model. Function and pathway analyses were performed using ingenuity pathways analysis.

RESULTS

Patients.

A total of 52 patients were analyzed: 26 with TIA and 26 vascular risk factor controls. The average age was 62.9 years. There were 26 men and 26 women. Patients were of mixed race and ethnicity, with Caucasian race accounting for 77% of the patients. A summary of the TIA and control subjects is shown in table 1. Since there was a significant difference in race between TIA and controls, we performed a secondary analysis restricted to Caucasians with TIA (n = 15) and controls (n = 25).

Table 1.

Demographic and clinical features of patients with transient ischemic attack and control subjects^a

Abbreviation: TIA = transient ischemic attack.

^aDifferences in demographic data between groups were analyzed using Fisher exact test or t test as appropriate. There was no significant difference in age, gender, vascular risk factors, and medications between the 2 groups analyzed using t test or χ² test. Race is significantly different between the two groups analyzed using χ² test. Therefore, we performed a secondary analysis restricted to Caucasian TIA (n=25) and Caucasian controls (n = 15).

TIA genomic profiles.

There were 480 probe sets corresponding to 449 genes that were differentially expressed between TIA and control subjects (FDR ≤0.05; fold change ≥1.2) (table e-1 on the Neurology® Web site at www.neurology.org). Of the 480 probe sets, 129 probe sets were downregulated (table e-1A) and 351 probe sets were upregulated (table e-1B). Each gene was represented by 1 to 6 probe sets. Hierarchical cluster analysis of the 480 probe sets is shown in figure 1. This visually demonstrates how TIAs separate from controls based on pattern of gene expression. A secondary analysis restricted to Caucasian patients with TIA and controls revealed similar results. This comparison of Caucasians demonstrated that >90% of the differentially expressed genes were common to the non-race-stratified analysis (453 out of 480 probe sets; data not shown).

Figure 1. Hierarchical cluster analysis of identified genes in transient ischemic attack (TIA) vs matched vascular risk factor controls.

Hierarchical cluster analysis of 480 gene probe sets differentially expressed in blood between patients with TIA and control subjects (false discovery rate ≤0.05, absolute fold change >1.2). Each column on the x-axis represents 1 patient, with 26 patients with TIA (blue) and 26 controls (orange). Each row on the y-axis represents individual probe sets (usually for individual genes). TIAs cluster separately from controls as indicated by the green arrow (top). Within subjects with TIA, at least 2 clusters are apparent as indicated by the red arrow. These 2 TIA clusters are labeled TIA1 and TIA2. One patient with TIA (ID: IT-062) clustered with controls. Two controls (ID: IT-155 and IT-177) clustered with TIAs. Diagnosis = blue (TIA) and orange (controls). Green = downregulation; ID = subject ID; Red = upregulation.

The presence of 2 TIA groups, labeled as TIA1 and TIA2, was identified on the cluster plot (figure 1). To determine whether a clinical feature was associated with the 2 TIA groups based on RNA patterns, we compared the following features between TIA1 and TIA2: age, gender, race, time after TIA, hypertension, diabetes, hyperlipidemia, smoking, history of stroke, cardiovascular disorder, atrial fibrillation, large vessel disease, ABCD2 score, DWI lesion, recurrent ischemic event within 90 days, and medications. None of these features were statistically different between the TIA1 and TIA2 groups. However, the power to detect a difference was limited due to the small sample size. Notably, the 3 patients with DWI lesions were in the TIA1 group, as was the one patient who had a subsequent stroke by 90 days (table 2). The genes associated with these TIA subtypes are shown in table e-1, C and D.

Table 2.

Demographic and clinical features of TIA1 and TIA2 subgroups^a

Abbreviations: AF = atrial fibrillation; CVD = cardiovascular disease; DWI = diffusion-weighted imaging; LVD = large vessel disease; TIA = transient ischemic attack.

^aThere was no significant difference between the 2 groups analyzed using t test or χ² test.

Prediction analysis for TIA.

A list of 34 probe sets that optimally distinguished TIA from control patients was identified using forward selection linear discriminate analysis on the 480 probe sets (table 3). A LDA prediction model was developed using these 34 probe sets and evaluated using 10-fold leave-one-out cross-validation. TIA could be distinguished from controls with 100% sensitivity (26 out of 26 TIAs correctly classified) and 100% specificity (26 out of 26 controls correctly classified) (figure 2).

Table 3.

The 34 genes derived using linear discriminant analysis that distinguished TIA from control with 100% specificity and 100% sensitivity

Abbreviation: TIA = transient ischemic attack.

Figure 2. Predicted probability of transient ischemic attack (TIA) or control diagnosis based on a linear discriminant analysis (LDA) model.

The LDA model was used to derive the 34 genes that optimally distinguished TIA from controls. Probabilities are based on 10-fold leave-one-out cross-validation analysis. The probability of predicted diagnosis is shown on the y-axis, and subjects are shown on the x-axis. Patients with TIA are shown on the right, and control subjects on the left. The predicted probability of TIA is shown in red, and the predicted probability of control is shown in blue. Patients with TIA could be distinguished from controls with 100% sensitivity and 100% specificity.

Function analysis of TIA-associated genes.

Function analysis of the 480 probe sets that distinguished TIA from controls demonstrated that they were significantly associated with immune functions. The genes associated with TIA demonstrated similar immune patterns as previously described in autoimmune diseases such as rheumatoid arthritis and Crohn disease. Function pathways for TIAs were also associated with immune patterns previously described in atherosclerosis and coronary artery disease (table e-2A). The pathways and associated differentially expressed genes are shown in table e-3A.

Function analysis of TIA subtype-associated genes.

Patients with TIA clustered into 2 groups based upon their RNA expression profiles (figure 1). To better characterize the significance of this finding, we performed a second function analysis of the TIA1 and TIA2 subtypes. The subgroup labeled as TIA1 had statistically more genes associated with atherosclerosis, stroke, coronary artery disease, hypertension, and diabetes mellitus as compared to the TIA2 subgroup (table e-2B). The pathways and associated genes for the TIA RNA subgroups are shown in table e-3, B and C.

DISCUSSION

Specific genes were expressed in blood following TIA associated with inflammation and coagulation as compared to controls. Whether these genes represent potential targets for TIA-specific therapies or biomarkers of TIA requires further study. Two patterns of RNA expression were observed in patients with TIA. The clinical significance of these RNA subtypes remains unclear, though these data strongly suggest that different patterns of immune activation may exist in TIA.

Patients with TIA may have unique patterns of inflammation that are associated with the development of subsequent ischemic vascular events. Previous studies have demonstrated a relationship between TIA and increased leukocyte activation and systemic inflammatory response,^6,7 presence of infection,^16–18 and other inflammatory conditions.^19–22 A number of genes differentially expressed in patients with TIA were similar to those previously associated with inflammatory bowel disease, rheumatoid arthritis, and cardiovascular disease. Whether this pattern of inflammation existed prior to TIA onset or was a result of acute TIA remains unclear. If the inflammation associated with TIA is an acute process, one would expect it to change over time. When we compared gene expression in TIAs with blood drawn within 24 hours (n = 11) to TIAs with blood drawn between 24 and 72 hours (n = 15), >90% of the genes expressed by 24 hours were similar to those expressed between 24 and 72 hours (data not shown). Though further studies of gene expression over time are required, this suggests that there may be a chronic inflammatory state in TIA that contributes to the development of symptomatic cerebrovascular disease.

Patients with TIA were also found to have increased expression of genes associated with coagulation. Specifically, genes involved in primary hemostasis, including platelet activation and aggregation as well as secondary hemostasis including intrinsic and extrinsic pathways, were identified. Of interest, tissue factor pathway inhibitor, an inhibitor of thrombin and factor Xa, was upregulated in TIA, suggesting differential regulation of the coagulation cascade and clot formation may be present in TIA. Genes encoding collagens and peptidases that break down von Willebrand factor (vWF)²³ were also overexpressed in patients with TIA compared to controls. Though peptidases that break down vWF have been associated with stroke^24,25 and thrombotic thrombocytopenic purpura,²⁶ this is the first report in TIA. These data suggest that procoagulant features are present in TIA that might provide treatment targets.

Hierarchical clustering demonstrated 2 TIA groups based upon genes differentially expressed between TIAs and controls. Despite this, we were unable to identify a clinical factor that was significantly different between these 2 groups. However, the TIA1 group contained all 3 of the patients with restricted diffusion identified on MRI. Such patients are known to be at high risk for subsequent stroke. In addition, one patient in the TIA1 group had a stroke within 3 months of the TIA. These patients suggest that the TIA1 group may represent patients at increased risk for recurrent ischemic events, though much larger studies will be required to ultimately address this hypothesis.

The genes that were different between TIA1 and TIA2 groups also provide insight into the differences between these 2 TIA groups. The pattern of gene expression in TIA2 was more similar to controls as compared to the TIA1 group. This may suggest that the TIA2 group includes nonvascular TIA mimics, which would be consistent with previously reported challenges in making the diagnosis of TIA.²⁷ The TIA2 group may have less inflammation compared to TIA1. The genes that were overexpressed in TIA1 compared to the TIA2 group were mostly related to inflammation and extracellular matrix remodeling. Although these TIA subgroups are of interest, further evaluation in larger samples is required before definitive conclusions can be drawn.

There were several limitations to the present study. Sample size relative to the number of genes analyzed was small. Although correction for multiple testing was performed, replication in an independent larger cohort is required. PCR validation is needed before genes are claimed to be TIA-specific markers. Our control group had more Caucasians compared to the TIA group. Although a secondary analysis restricted to Caucasians revealed very similar results, the influence of racial differences contributing to identified genes requires further study. The identification of relevant activated biological pathways in individual patients with TIA might be potentially useful to personalize approaches to treatment.

Patients with acute TIAs have different patterns of gene expression compared to asymptomatic controls with vascular risk factors. Genes expressed in TIA were associated with systemic inflammation, platelet activation, and prothrombin activation. Further study of the 2 patterns of RNA expression in patients with TIA is warranted to ascertain any relevance of these genes to cause of TIA or risk for subsequent stroke.

GLOSSARY

CRP

C-reactive protein

FDR

false discovery rate

LDA

linear discriminant analysis

TIA

transient ischemic attack

vWF

von Willebrand factor

AUTHOR CONTRIBUTIONS

Dr. Zhan: designed the studies, acquired the data, analyzed and interpreted the data, performed statistical analysis, drafted the manuscript, and revised the manuscript. Dr. C. Jickling: acquired the data, analyzed and interpreted the data, performed statistical analysis, and revised the manuscript. Dr. Tian: acquired the data and revised the manuscript. Dr. Stamova: revised the manuscript. Dr. Xu: acquired the data and revised the manuscript. Dr. Ander: acquired the data and revised the manuscript. Dr. Turner: acquired the data and revised the manuscript. M. Mesias: acquired the data. Dr. Verro: acquired the data and revised the manuscript. Dr. Bushnell: acquired the data and revised the manuscript. Dr. Johnston: designed the studies and acquired the data. Dr. Sharp: conceived and designed the studies, acquired the data, analyzed and interpreted the data, obtained funding, helped draft the manuscript, and revised the manuscript.

DISCLOSURE

Dr. Jickling is a fellow of the Canadian Institutes of Health Research (CIHR). Drs. Xu, Ander, and Tian are AHA Bugher Fellows. Dr. Turner is supported by a National Health and Medical Research Council (Australia) Postdoctoral Fellowship. Dr. Zhan is co-holder of a patent re: The use of blood RNA in TIA. Dr. Jickling, Dr. Tian, and Dr. Stamova report no disclosures. Dr. Xu receives research support from the American Heart Association Bugher Foundation. Dr. Ander serves as a consultant for Ischemia Care and receives research support from the American Heart Association. Dr. Turner and M. Mesias report no disclosures. Dr. Verro has received speaker honoraria from Boehringer Ingelheim and receives research support from the American Heart Association Bugher Foundation. Dr. Bushnell serves on a scientific advisory board for Boehringer Ingelheim and has received research support from Bristol-Myers Squibb/Sanofi Pharmaceuticals Partnership, the NIH/NINDS, the American Heart Association/Bugher Foundation, and the Hazel K Goddess Fund for Research on Stroke in Women. Dr. Johnston is co-holder of patent re: The RNA panel to identify TIA and risk stratify and receives research support from sanofi-aventis, Strkyer Neurovascular, Boston Scientific, the NIH (NCRR, NINDS), Kaiser-Permanente, and AHA/ASA, Bugher Award. Dr. Sharp serves as Senior Editor for Brain Research; holds a patent re: Gene expression profiles for the diagnosis of ischemic stroke; and receives research support from the NIH/NINDS and the American Heart Association.