Ischemic Transient Neurological Events Identified by Immune Response to Cerebral Ischemia

Abstract

Background and Purpose

Deciphering whether a transient neurological event (TNE) is of ischemic or nonischemic etiology can be challenging. Ischemia of cerebral tissue elicits an immune response in stroke and transient ischemic attack (TIA). This response, as detected by RNA expressed in immune cells, could potentially distinguish ischemic from nonischemic TNE.

Methods

Analysis of 208 TIAs, ischemic strokes, controls and TNE was performed. RNA from blood was processed on microarrays. TIAs (n=26) and ischemic strokes (n=94) were compared to controls (n=44) to identify differentially expressed genes (FDR<0.05, fold change ≥∣1.2∣). Genes common to TIA and stroke were used predict ischemia in TIA-DWI positive / minor-stroke (n=17), nonischemic TNE (n=13) and TNE of unclear etiology (n=14).

Results

Seventy-four genes expressed in TIA were common to those in ischemic stroke. Functional pathways common to TIA and stroke related to activation of innate and adaptive immune systems, involving granulocytes and B-cells. A prediction model using 26 of the 74 ischemia genes distinguished TIA and stroke subjects from controls with 89% sensitivity and specificity. In the validation cohort, 17/17 TIA-DWI positive / minor-strokes were predicted to be ischemic, and 10/13 nonischemic TNE were predicted to be nonischemic. In TNE of unclear etiology, 71% were predicted to be ischemic. These subjects had higher ABCD² scores.

Conclusions

A common molecular response to ischemia in TIA and stroke was identified, relating to activation of innate and adaptive immune systems. TNE of ischemic etiology were identified based upon gene profiles that may be of clinical utility once validated.

INTRODUCTION

Systemic inflammation is linked to ischemic stroke and transient ischemic attack (TIA). Cerebral ischemia produces many endogenous ligands and cytokines that elicit an immune response ¹. Leukocytes, including neutrophils and monocytes, are activated and recruited to initiate processes of containment, removal, and repair. We have previously demonstrated in a rat model of TIA that a peripheral immune response occurs to transient brain ischemia with similarities to that observed in experimental brain infarction^{2, 3}. This suggests that aspects of the immune response to cerebral ischemia in TIA and stroke are common and may be useful in identifying ischemic events.

In clinical practice, deciphering whether a transient neurological event (TNE) is of ischemic or nonischemic etiology is often difficult. Many common neurological conditions mimic the symptoms of TIA, including migraine, seizure, and syncope. The transient nature of TIA adds to the diagnostic challenge as objective deficits generally resolve by the time of presentation and assessment of symptoms is reliant on patient recall and physician interpretation. Current diagnostic tests including neuroimaging, electrocardiogram, and electroencephalogram are frequently unremarkable, leaving the cause of a TNE unclear. As a result, TIA is estimated to be incorrectly diagnosed in as many as 50% of cases ^4-8. However, correctly identifying ischemic TNE is critical as the risk for stroke is high in TIA and can be reduced by early initiation of stroke prevention therapy ⁹. Additionally, identifying nonischemic TNE (which have a very low risk of stroke) will improve risk stratification and use of health care resources. Thus, improved methods to distinguish ischemic from nonischemic causes of TNE are needed.

We hypothesized that the common immune response to ischemia in TIA and stroke could distinguish ischemic from nonischemic TNE. Previous studies have demonstrated an immune response in patients with stroke by evaluating leukocyte RNA expression using whole genome microarrays ^10-12. However, the distinction between immune response to cerebral ischemia and cerebral infarction remains unclear. We report the pathways associated with the peripheral immune response to cerebral ischemia, which with further study and validation may have clinical utility to identify ischemic causes of TNE.

SUBJECTS AND METHODS

1. Study Patients

Patients with ischemic stroke, TIA, and controls were enrolled from the University of California Davis, and the University of California San Francisco. Study protocols were approved by the institutional review boards at each site and written informed consent was obtained from each patient. All patients received standardized clinical evaluations, including medical history, brain imaging, Doppler, vascular angiography, electrocardiogram, echocardiogram, and 24-48 hour cardiac monitoring. Blood samples were drawn into PAXgene tubes (PreAnalytiX, Hilden, Germany) within 72 hours of symptom onset for gene expression analysis.

The diagnoses of stroke, TIA and TNE of nonischemic etiology required consensus of 3 study neurologists. Subjects where a consensus was not obtained were classified as TNE of unclear etiology. A derivation cohort of TIA (n=26), ischemic stroke (n=94) and matched vascular risk factor controls (n=44) were used to identify genes associated with cerebral ischemia (Figure 1, Table 1). TIA was defined as an episode of neurological dysfunction lasting ≤24 hours resulting from focal cerebral ischemia with no restricted diffusion on MRI. Ischemic stroke was defined by neurological deficits persisting longer than 24 hours with acute cerebral infarction on imaging. Patients with intracerebral hemorrhage or hemorrhagic infarction were excluded from study. Stroke subtype was classified as large-artery atherosclerotic, cardioembolic, small-vessel lacunar, and cryptogenic as previously described ¹³. Controls were subjects with vascular risk factors without symptomatic cardiovascular disease (stroke, myocardial infarction, peripheral vascular disease) who were similar in age, race, and gender to TIA and stroke subjects. Differences between groups were analyzed using Fisher’s exact test, two-tailed t-test, or Wilcoxon-Mann-Whitney test where appropriate (Stata 10.1, College Station, TX, USA).

Figure 1.

Flow diagram of study analyses. A. To derive common ischemia genes, TIA and ischemic stroke subjects were compared to controls. Of the 160 probesets differentially expressed in TIA, 74 (46%) were identical to those in ischemic stroke (FDR<0.05, fold change ≥1.2∣). A subset of 26 probesets were found to optimally distinguish TIA and ischemic stroke from controls using LDA. B. A prediction model using these 26 probesets was developed and evaluated by cross-validation on the original validation cohort, and on a validation cohort of TNE of known ischemic and nonischemic etiology. C. The 26 probesets were subsequently used to predict ischemia in TNE of unclear etiology.

Table1.

Demographic variables for TIA and ischemic stroke patients compared to vascular risk factor controls. p-values represent the comparison of TIA versus control and ischemic stroke versus control.

	Control (n=44)	TIA (n=26)	p	Ischemic Stroke (n=94)	p
Age years (SD)	62.5 (7.0)	65.6 (11.4)	0.17	64.5 (10.9)	0.29
Race Caucasian n(%)	33 (75%)	15 (57.9%)	0.14	57 (60.6%)	0.10
Gender Male n (%)	22 (50%)	13 (50%)	1.00	46 (48.9%)	0.91
Hypertension n(%)	31 (72.1%)	19 (73.1%)	0.93	59 (66.3%)	0.86
Diabetes n(%)	12 (30%)	9 (34.6%)	0.70	27 (30.3%)	0.97
Hyperlipidemia n(%)	18 (51.4%)	16 (61.5%)	0.44	34 (38.2%)	0.18
Atrial fibrillation n(%)	2 (6.3%)	2 (7.7%)	0.83	10 (11.2%)	0.42
Prior Stroke/TIA n(%)	0	8 (30.6%)	0.08	24 (27.0%)	0.09
Stroke/TIA Subtype n(%)
Large Vessel n(%)	N/A	5 (19.2%)		14 (14.9%)
Cardioembolic n(%)	N/A	6 (23.1%)		23 (24.4%)
Small Vessel n(%)	N/A	4 (15.3%)		29 (30.8%)
Cryptogenic n(%)	N/A	11 (42.3%)		28 (29.8%)
NIHSS-24hr (IQR)	N/A	0		10.3 (8.7-12.0)
Hours since TIA/Stroke (SD)	N/A	33.7 (21.0)		34.7 (22.6)

Abbreviations: IQR, Interquartile range; NIHSS, National Institutes of Health Stroke Scale; SD, standard deviation.

The genes associated with cerebral ischemia were evaluated in a validation cohort comprised of ischemic TNE (n=17) and nonischemic TNE (n=13) (Supplementary Table 1). Ischemic TNE were patients with transient neurological symptoms with DWI-positive MRI and minor strokes with an NIHSS ≤5 at 24 hours. Nonischemic TNE included patients with migraine, seizure or syncope. They had no evidence of infarction on MRI and a clinical presentation inconsistent with transient ischemia.

Genes associated with cerebral ischemia were used predict ischemia in TNE of unclear etiology (n=14) (Figure 1). TNE of unclear etiology were patients with a transient neurological event where diagnostic consensus was not obtained between the three study neurologists, and included two subjects where TIA was thought to be the cause but a brain MRI was not available for review.

2. Sample Processing

A venous blood sample was collected in PAXgene tubes within 72 hours of stroke or TIA onset (PreAnalytiX, Germany). Samples were stored at −80°C and processed at the same time in the same laboratory to reduce batch effect. Total RNA was isolated according to the manufacturer’s protocol (PAXgene blood RNA kit; Pre-AnalytiX). RNA concentration was determined by Nano-Drop (Thermo Fisher) and RNA quality by Agilent 2100 Bioanalyzer. Samples required A₂₆₀/A₂₈₀ absorbance ratios of purified RNA ≥2.0 and 28S/18S rRNA ratios ≥1.8. NuGEN’s Ovation Whole Blood Solution (NuGEN Technologies, San Carlos, CA) was used for reverse transcription, amplification, and sample labeling. Hybridization of each RNA sample was performed according to the manufacturer’s protocol on Affymetrix Human U133 Plus 2.0 GeneChips (Affymetrix Santa Clara, CA). Arrays were washed and processed on a Fluidics Station 450 and scanned on a GeneChip Scanner 3000. Samples were randomly assigned to microarray batch stratified by diagnosis.

3. Microarray Data Analysis

Microarray data files were pre-processed using robust multichip averaging (RMA), mean-centering standardization and log2 transformation (Partek Genomics Suite 6.4, Partek Inc., St. Louis, MO). Nonspecific probesets with an interquartile range <0.5 across the all subjects were filtered as previously described ^{14, 15}. Patients with TIA and ischemic stroke were compared to controls using Analysis of Covariance (ANCOVA) adjusted for age and batch. The analyses included TIA versus controls and ischemic stroke versus controls. After Benjamini-Hochberg false discovery rate (FDR) correction for multiple comparisons, probesets with a corrected p-value <0.05 and fold change ≥∣1.2∣ were considered significant.

4. Gene Functional Analysis

To identify functional pathways associated with the differentially expressed genes, Ingenuity Pathway Analysis (IPA, Ingenuity Systems®, www.ingenuity.com) was used. Pathways with a greater number of genes than expected by chance were considered significant (p<0.05, Fisher’s exact test). To identify genes associated with specific immune cells, the list of differentially expressed genes for TIA and stroke were overlapped with previously published lists of genes shown to be unique to granulocytes, natural killer cells, monocytes, B-cells, CD4 T-cells, and CD8 T-cells ¹⁶.

5. Prediction Analysis

The probesets common to ischemic stroke and TIA were used to develop a prediction model to discriminate ischemia from controls (Figure 1). This control group was used to identify genes due to ischemia and not due to vascular risk factors. From the list of 74 common ischemia genes, the optimal genes to discriminate ischemia from controls were identified using forward selection linear discriminant analysis (LDA). LDA is an analytical method that identifies a linear combination of features to separate two or more classes. For this analysis, the genes expressed in blood were the features, and the classes were ischemia and non-ischemia. A panel of 26 genes was identified and used to develop an LDA prediction model to distinguish ischemia from controls. This model was evaluated using 10-fold leave-one-out cross-validation analysis and in a second cohort of subjects with TIA-DWI positive / minor stroke (n=17) and nonischemic TNE (n=13: migraine n=7; seizure n=3; syncope n=3) (Figure 1). The developed model was then used to predict TNE of unclear etiology.

RESULTS

Demographic and clinical characteristics of subjects analyzed in the derivation set are shown in Table 1. There were 26 subjects with TIA, 94 with ischemic stroke, and 44 controls. The mean age was 64.1 years (SD±11.0) and 49.3% were male. The cohort was ethnically diverse: 105 (64%) were Caucasian, 28 (17%) were African American, 16 (10%) were Hispanic, 10 (6%) were Asian, and 5 (3%) were of other race. TIA subjects had a mean ABCD² score of 4.6 (range 2-6). Controls had vascular risk factors but no symptomatic cerebrovascular or cardiovascular disease. Age, gender, race, hypertension, diabetes, and hyperlipidemia were not significantly different between TIA or ischemic stroke and control subjects.

A total of 160 annotated probesets representing 145 genes were significantly different between TIA and control subjects (FDR<0.05 fold change ≥1.2∣). A total of 461 annotated probesets representing 413 genes were significantly different between ischemic stroke and control subjects (FDR<0.05 fold change ≥1.2∣). There were 74 probesets (46%) of the 160 probesets differentially expressed in TIA that were common to the 461 probesets differentially expressed in ischemic stroke (Figure 1). Analysis of the common functional pathways between TIA and ischemic stroke are shown in Table 2. Of the common genes and pathways in TIA and stroke, most were associated with activation and development of the immune cells, including TLR5, TLR6, TLR10, CASP1, CASP6, CARD16, IL8, IL10, CD36 and CD86. TREM-1 signaling involving TLR was the top canonical pathway common in TIA and stroke (Table 2). There were 86 probesets unique to TIA (Supplementary Table 3), the functional analyses of which are shown in Supplementary Table 4.

Table 2.

Functional analysis of pathways common to TIA and ischemic stroke. The canonical and functional pathways represented greater than expected by chance are shown (p<0.05, Fisher’s exact test), along with the genes expressed in the listed pathways. Pathways represent alterations in the blood that occur in patients with TIA and ischemic stroke. Genes that are underlined represent those common to TIA and ischemic stroke.

	Pathway	Genes	TIA p- value	Stroke p- value
Canonical Pathways	TREM1 Signaling	CASP1,CASP5,CD86,FCGR2B,IL8,IL10, TLR5,TLR6,TLR10	1.2×10−4	4.2×10−6
	Communication between Innate and Adaptive Immune Cells	CD86,IL8,IL10,TLR5,TLR6,TRL10	1.4×10−2	7.5×10−3
	Altered B Cell Signaling	CD86,IL10,LTB,TLR5,TLR6,TLR10	3.2×10−2	9.6×10−3
	Role of Patterns Recognition Receptors	C1QC,CASP1,IL10,IRF7,TLR5,TLR6	2.8×10−2	2.9×10−2
Molecular Functions	Leukocyte development	ADM,BATF,BCL6,BCL11A,BCL11B,C1QC, CASP1,CAV1,CCND3,CCR2,CD36,CD59, CD86,ETS1,F5,FCGR2B,FN1,FUS,HGF, IFNAR1,IL8,IL10,IL12RB2,IL2RB,IL5RA, JAG1,JMJD6,KIT,LIFR,LTB,LTBR,MEIS1, MYB,NRBP2L2,NQO2,PLSCR1,PRL, RNASE1,RNASE2,SMAD7,SNRK,SOCS1, TAPBP,TLR5,TLR6,TOP2A,TPP2,WT1, XRCCR	1.9×10−4	1.6×10−6
	Phagocyte development	C1QC,CCR2,CD36,CD86,FCGR2B,FN1, HGF,IFNAR1,IL10,IRF7,KIT,LIFR,LTB, LTBR,MYB,N4BP2L2,RNASE1,RNAS32, SOCS1,TLR5,TLR6	7.4×10−3	8.9×10−6
	Inflammatory response	ANXA3,AIM2,BCL6,CASP1,CASP4,CASP5, CAV1,CCR2,CD36,CD59,CD86,CD164, COL1A1,CXCR7,EDNRB,ETS1,F5,FCGR1A, FCGR2B,FN1,GM2A,GNA12,HGF,IFNAR1, IL8,IL10,IL12RB2,IL18BP,IL2RB,JMJD6, KIT,LTB,LTBR,MARCO,MEIS1,MYLK3, NMV,NQO2,PFDN6,PLSCR1,PRL,RAB27A, RNASE2,S100A12,S1PR3,SERPINF1,SLPI, SIGLEC8,SMAD7,SOCS1,TAPBP,TLR5, TLR6	4.7×10−4	5.0×10−5

Using genes previously identified as unique to each immune cell type ¹⁶, the proportion of cell types expressing RNA in TIA and ischemic stroke were estimated. Genes differentially expressed in TIA were associated with granulocytes and B-cells (Figure 2). Genes differentially expressed in ischemic stroke were associated with granulocytes, monocytes, natural killer cells (NK) and B-cells, and, to a lesser extent, with megakaryocytes (MK) and CD4 T-cells (Figure 2).

Figure 2.

Proportional representation of the immune cells that contributed to RNA expression in blood in TIA and ischemic stroke patients. Previous studies have identified genes unique to each cell type in blood ¹⁶. These profiles were used to identify cells contributing to the differentially expressed genes in TIA and stroke. The genes differentially expressed in TIA were associated mostly with granulocytes and B-cells. The genes differentially expressed in ischemic stroke were associated mostly with granulocytes, monocytes, natural killer cells and B-cells, with some contribution from megakaryocytes and CD4 T-cells.

Of the 74 probesets (63 genes) common to TIA and stroke, a 26 gene panel was identified that optimally distinguished subjects with cerebral ischemia from controls (Figure 1). An LDA model based on this gene panel correctly predicted 26 out of 26 TIA subjects, 85 out of 94 ischemic stroke subjects, and 41 out of 44 control subjects. On cross-validation analysis, the 26 gene LDA model correctly predicted 89% of cerebral ischemia subjects and 89% of control subjects (Figure 3). The probability of predicted diagnosis for the majority of subjects with and without ischemia was >90%.

Figure3.

Probability plots of the predicted diagnosis of cerebral ischemia in TIA and ischemic stroke versus controls. The predicted probabilities are from cross-validation of the LDA 26 common ischemic gene model on the derivation cohort. The predicted probability of ischemia versus nonischemia in shown for (3A) the 26 patients clinically diagnosed as TIA; (3B) the 94 with ischemic strokes; (3C) the 44 vascular risk factor controls. The probability of a subject being ischemia is shown in red, and the probability of a subject being nonischemia is shown in blue. Cerebral ischemia was correctly predicted in 89% of subjects, and nonischemia was correctly predicted in 89% of controls. The probability of predicted diagnosis for the majority of subjects was >90%.

In a validation cohort of TIAs with DWI positive lesions and minor strokes, ischemia was the predicted diagnosis in 17 out of 17 subjects (100%). In a cohort of nonischemic TNE that included patients with migraine, seizure, and syncope, nonischemia was predicted in 10 of 13 subjects (77%). The 26 gene LDA model was then used to predict ischemia in subjects with TNE of unclear etiology. Cerebral ischemia was predicted in 71% (n=10) of subjects, and 29% (n=4) were predicted to be nonischemic events. Though the sample size was small, TNE of unclear etiology predicted to be ischemic had higher ABCD² scores compared to those predicted to be non-ischemic (Table 3).

Table 3.

Summary of transient neurological events of unclear etiology predicted to be due to ischemia or nonischemia. Prediction was based on the 26 gene model developed from common ischemia genes in patients with TIA, ischemic stroke and controls (Fisher’s exact test, Wilcoxon-Mann-Whitney test)

	Predicted Ischemia (n=10)	Predicted Nonischemia (n=4)	p
Age years (SD)	71.7 (19.2)	70.3 (13.5)	0.90
Race Caucasian n(%)	7 (70%)	3 (75%)	1.00
Gender Male n(%)	4 (40%)	1 (25%)	1.00
Hypertension n(%)	8 (80%)	2 (50%)	0.52
Diabetes n(%)	4 (40%)	0 (0%)	0.25
Hyperlipidemia n(%)	2 (20%)	2 (50%)	0.31
Atrial fibrillation n(%)	4 (40%)	0 (0%)	0.25
ABCD2 Score (IQR)	5 (4-6)	3 (2-4)	0.04
Duration >60minutes n(%)	7 (70%)	2 (50%)	0.58
Weakness n(%)	6 (60%)	1 (25%)	0.56
Language/Speech n(%)	4 (40%)	1 (25%)	1.00
Sensory symptoms n(%)	4 (40%)	3 (75%)	0.28
Headache n(%)	2 (20%)	1 (25%)	1.00

DISCUSSION

TIA and stroke were demonstrated to share a common immune response to cerebral ischemia. One half of the genes expressed in TIA were also expressed in stroke, representing activation of innate and adaptive immune responses. This common immune response to ischemia in TIA and stroke was able to distinguish subjects with and without cerebral ischemia. Given the difficulty in determining etiology of transient neurological events, and the significance of this distinction to treatment, a reliable marker that identifies ischemic TNE would be clinically useful.

Immune response to ischemia to identify TIA

TIA is a challenging diagnosis because transient neurological symptoms can be mimicked by several common conditions including migraine, seizure, and syncope ^{4, 6-8}. The genes associated with cerebral ischemia, as identified by those common to TIA and ischemic stroke, might identify cerebral ischemia in patients with transient neurological symptoms. We identified a profile of 26 genes able to distinguish TIA and minor stroke subjects from nonischemic TNE. When the profile was used to predict subjects with transient neurological symptoms of unclear etiology, 71% were predicted to be ischemic. This is important as improved recognition of ischemic TNE could improve the delivery of urgent neurovascular evaluation and treatment shown to prevent stroke in TIA. Furthermore, identifying nonischemic TNE that mimic TIA could reduce the costs of hospital observation and extensive evaluation typical of patients diagnosed with TIA.

TNE predicted to be ischemic tended to have higher ABCD² scores. This is consistent with other studies demonstrating the diagnostic potential of the ABCD² score to identify transient events due to ischemia ^{17, 18}. An ABCD² score ≥4 has been shown to identify 60% of TIAs and 82% of minor strokes ¹⁸. Additionally, an ABCD² score <4 identified 65% of nonischemic TNE. However, this leaves 40% of TIAs, 18% of minor strokes, and 35% of nonischemic TNE incorrectly identified, indicating that additional methods to distinguish ischemic from nonischemic TNE are required. Diagnosis is critical as early stroke risk is low in nonischemic TNE and is as high as 10-25% in TIA ¹⁹. The genes identified in this study were able to identify all TIAs and minor ischemic strokes, including those subjects with an ABCD² score <4. Further study is required to determine if an RNA profile can add to the ABCD2 and/or MRI-DWI. The ABCD2 score and/or RNA may be of use in patients where MRI cannot be performed due to contraindications, cost or accessibility. Additionally, RNA may provide additional information that complements the ABCD2 score in TIA. For example patients with a low ABCD2 score with an RNA profile suggesting ischemia may benefit from urgent evaluation. However, this requires additional evaluation.

Immune Response Common to TIA and Stroke

The common peripheral immune response to cerebral ischemia identified was associated with communication between innate and adaptive immune cells, including activation of granulocytes and B cells. This response reflects features shared by TIA and stroke, including immune cell interaction with ischemic but not infarcted tissue, acute thrombosis, and possibly a stress response to acute injury.

Aspects of the common immune response have previously been demonstrated in stroke ²⁰. Toll like receptor (TLR) signaling was present in both TIA and stroke, with expression of TLR5, TLR6 and TLR10 identified. TLRs play critical roles in initiating the innate immune response and shaping the adaptive immune response . In mice, knockout of TLR4 results in smaller infarct volumes, better outcomes, and decreased inflammatory markers including IRF-1, iNOS, COX2 and MMP9 ^{21, 22}. Likewise, knockout of TLR2 can reduce CNS inflammatory markers ICAM-1, IL-6, MCP, and ELAM-1 ^{23, 24}, and modulation of TLR-9 can reduce infarct size in mice and non-human primates ²⁵ . In human stroke, TLR expression is less well characterized. The expression of TLR2 and TLR4 on monocytes at the time of stroke is associated with larger infarct volumes, increased protein levels of IL1B, TNF-α, IL6 and VCAM1 and worse outcomes at 3 months ^{26, 27}. TLR7 and TLR8 may also be associated with poor outcome and greater inflammatory response ²⁸. Our study suggests that TLR 5, 6 and 10 play a role in the peripheral inflammatory response in both stroke and TIA, suggesting that further study of these receptors is required.

TREM-1 signaling was the top common pathway identified in both TIA and ischemic stroke. TREM (Triggering receptor expressed on myeloid cells) is a receptor that can modulate the innate immune response to prevent excessive inflammation and tissue damage ^{29, 30}. TREM regulation of TLR responses is mediated through modification of the caspase-recruitment domain protein (CARD) complex ^{31, 32}. In our study, expression of TLR and CARD genes was identified, with CARD16 being expressed more in TIA than stroke. Further study of TREM-1 signaling in cerebral ischemia may provide a means to modulate inflammation in ischemic brain injury.

The expression of genes associated with B-cells occurred in both stroke and TIA. This suggests that B-cells may play an important role in cerebral ischemia. B-cells have prominent effects on inflammatory responses ^{21, 33}. Depletion of B cells in multiple sclerosis worsens CNS injury ^{34, 35}. In experimental stroke, infarct size can be reduced with B cell transfusion ³⁶. The ability of B-cells to limit CNS injury is dependent on the anti-inflammatory effects of IL-10 ^36-38. Our study suggests that B-cells are important in the immune response to cerebral ischemia in stroke and TIA.

Limitations

Sample size was small in this preliminary study. Larger studies are required to ensure TIA and nonischemic TNE populations are comprehensively evaluated. Microarray studies evaluating many genes have an inherent risk of false discovery. Though we adjusted for this risk using FDR corrected p-values and assessed the developed model in a validation cohort, evaluation of identified genes and pathways in a second independent cohort is required. Reliability of findings is increased because identified genes and pathways associated with ischemia were present in both TIA and stroke subjects. Since blood samples were obtained within 72 hours of TIA or stroke, differences in immune response to ischemia that may exist beyond this time remains unclear. Analysis of individual immune cells was performed based on prior studies on genes unique to each cell type. Further evaluation of individual immune cell response to ischemia in stroke and TIA is required. A clinical diagnosis of TIA has limitations in its accuracy, and this may have been introduced into the derived genes to predict ischemia in TNE. Though TIAs were selected based on criteria to make ischemia the most likely etiology, it is possible that some TIA patients were misclassified. Ischemic stroke patients were used to ensure the identified genes were present in patients with definitive cerebral ischemia. Future studies could follow patients with TIA and TNE over time to refine clinical diagnoses. Since short-term risk of stroke confirms that an initial event was a TIA rather than a mimic, the profile may also be valuable as a predictor of stroke risk.

In conclusion, a common peripheral response to ischemia in stroke and TIA was identified. The genes and pathways provide insight into the peripheral inflammatory response to cerebral ischemia. With further study, the common immune response may demonstrate clinical utility to discriminate ischemic from nonischemic causes of TNE, and thus facilitate evaluation and treatment of TIA.