Stroke is one of the leading causes of death in modern society, being in third position behind heart disease and cancer in the United States (1) and even in first position in some Eastern European countries (2). Early differentiation between transient ischemic attack (TIA) and stroke remains challenging, and the diagnosis of TIA remains biased by subjective judgment (3). Moreover, the prognosis may vary, although stroke and TIA share pathophysiological mechanisms and their therapeutic consequences have assimilated (4).

Cerebral infarctions may be further itemized to markers associated with infarct size (e.g., astroglial protein S-100B) and with damage to the blood-brain barrier (BBB, for example, c-Fn, matrix metalloproteinase-9 (MMP-9) (5). Recent studies identified stable precursor fragments of the neuropeptides enkephalin (proenkephalin A [PENK-A]) and substance P (protachykinin A [PTA]) as potent markers of BBB integrity (6- 7). Both mature neuropeptides are active as neurotransmitters and are involved in nociception and immune stimulation. They have been implicated in the pathophysiology of certain neuropathologies, including Parkinson's disease, Alzheimer's disease, and severe head injury (8). Elevated plasma levels of methionine-enkephalin in patients after acute cerebral infarction have been reported (9). The role of enkephalins in stroke is, however, incompletely understood as both presence and absence of a neuroprotective effect of opioid receptor stimulation have been described (10- 11).

Early identification of patients with increased risk is of great importance in the process of clinical evaluation of emergency patients. In the era of flowchart-guided decision trees, the identification of novel biomarkers is of growing importance for both diagnostic and prognostic evaluation. Similar to the successful use of biomarkers in the management of acute cardiac events, there are increasing efforts to establish biomarker-based tests for the evaluation of patients admitted to the stroke unit for clinical decision making and risk assessment (12- 13). In contrast to acute cardiac events, however, validated markers for ischemic stroke are lacking (14), and current guidelines make no recommendations for the use of biomarkers for risk stratification in stroke.

The aim of the present study was to evaluate PENK-A and PTA in patients presenting with symptoms of acute cerebrovascular disease in relation to stroke severity and to outcome. We hypothesized that PENK-A and PTA may be associated with a poorer outcome.

Our study population comprised 189 consecutive patients, who presented with symptoms of acute stroke to the emergency department of the University Hospital Tuebingen between January 2008 and May 2009. The median time between symptom onset and hospital admission was 4.5 h (interquartile range [IQR] 1.3 to 19.8 h). Patients with symptom onset >24 h before initial assessment on hospital admission were not considered as acute stroke patients and were therefore excluded. Blood was drawn from the anticubital vein within 1 h of hospital admission. All patients underwent cranial computed tomography (CT [Siemens Somatom Sensation16, Erlangen, Germany]) within 90 min after hospital admission. The scans were performed on a Siemens scanner using 4.5-mm slices. Images were analyzed by an experienced neuroradiologist, as described previously (15). Brain lesion volume was calculated according to the formula for irregular volumes (16).

Exclusion criteria were age <18 years and inability to provide informed consent. The acute cerebral events were grouped according to the type of event as TIA, stroke, or nonischemic event. Transient ischemic attack was defined by the duration of symptoms of stroke <24 h and absence of brain lesion on CT (4). Stroke was considered with symptoms lasting >24 h after onset and with the evidence of an acute clinically relevant brain lesion in patients with rapidly vanishing symptoms in CT. Previously defined etiologic subgroups of stroke were determined according to the TOAST (Trial of ORG10172 in Acute Stroke Treatment) criteria as: 1) large-artery atherosclerosis (LAA); 2) cardioembolism; 3) small-vessel occlusion; 4) stroke of other determined etiology; or 5) stroke of undetermined etiology (17). Stroke severity was assessed according to the National Institutes of Health Stroke Scale (NIHSS) on hospital admission (18). Nonischemic events were considered if none of the previous criteria applied but 1 of the following disorders was identified: epileptic seizure (n = 15), medication interaction (n = 2), transient visual disorder (n = 4), paroxysmal positioning vertigo (n = 5), migraine with aura (n = 11), psychogen (n = 2), brain tumor (n = 4), and alcohol abuse (n = 6). Patients with higher NIHSS score (>12) or intracerebral hemorrhage had been excluded. The study was approved by the local ethics committee.

Clinical outcome was assessed after 3-month follow-up and was performed by hospital chart analysis and by a pre-defined structured telephone interview. For functional outcome assessment, the modified Rankin Scale (mRS) scoring was applied according to previous description (19): 0 = no symptoms at all; 1 = no significant disability despite symptoms; 2 = slight disability; 3 = moderate disability; 4 = moderately severe disability; 5 = severe disability; 6 = dead.

The ethylenediaminetetraacetic acid samples were immediately centrifuged, aliquoted, and stored at −80°C until batch analysis. Routine biochemical parameters were assessed in the standard clinical laboratory.

Plasma samples were blinded as to the underlying diagnosis of the patients for further assessments. Midregional proenkephalin A 119–159 (PENK-A) and N-terminal protachykinin A (PTA) were assessed by a newly developed chemoluminescence sandwich immunoassay, using a chemiluminescence platform and coated tube technique, as described previously (Immunochemical Intelligence GmbH, Berlin, Germany) (6- 7). All assessments were run in duplicates. For the measurement of PENK-A, monoclonal antibodies against the amino acids sequence 121-134 of PENK-A were used. Polyclonal antibodies against the amino acids sequence 139–155 were labeled with methyl-acridinium N-hydroxysuccinimide and used as tracer. The PTA was measured using monoclonal mouse antibodies against the amino acid sequence 3–22 of PTA, and monoclonal antibodies against amino acids 21–36 were labeled for tracing.

Intra-assay and interassay coefficients of variation were <10% and <15%, respectively, for PTA, and <12% and <13% for PENK-A. Lower detection limits were at 13.5 pmol/l and 15.6 pmol/l for PTA and PENK-A, respectively. The normal range of PENK-A is from 41.8 to 131 pmol/l (median 62.3 pmol/l), and for PTA, from 30.8 to 179.2 pmol/l (median 98.7 pmol/l). Normal ranges are not dependent on age or sex.

A 2-tailed probability value of < 0.05 was considered as statistically significant. Variables are presented as median and IQR. Group comparison was performed by analysis of variance, by Fisher's post-hoc test, and by Kruskall-Wallis test as appropriate. PENK-A and PTA were log-transformed for statistical analyses as the Kolmogorov-Smirnov test revealed nonnormal distribution of the variables. Associations between novel biomarkers and clinical parameters were made by Pearson correlation. Logistic regression was applied as appropriate.

Adjustment for possible confounders was performed by the multifactorial analysis of covariance for the decadic logarithms of PENK-A and PTA, which included age, sex, risk factors, and clinical parameter such as arterial hypertension, diabetes mellitus, hyperlipidemia, family history of coronary artery disease (CAD), smoking, concomitant CAD, atrial fibrillation, brain lesion size, laboratory parameters, and medication at the time of admission.

Outcome was assessed for the coprimary endpoint of all-cause mortality and the composite endpoint of all-cause mortality, stroke reoccurrence, and nonfatal myocardial infarction (MACCE) (20). Cox proportional hazard analysis and log-rank test (Mantel-Cox) were used for outcome analyses, and hazard ratios (HR) and 95% confidence intervals (CI) are presented. Odds ratio (OR) for functional outcome is reported. Unadjusted models and adjusted for age, NIHSS, and brain lesion size (PENK-A only) were calculated. For outcome analysis, we assessed biomarkers as continuous variables and grouped for highest quartile versus lower 3 quartiles. Kaplan-Meier survival plots are shown for illustrative purpose.

Time-dependent receiver-operating characteristic (ROC) curves and time-dependent area under the curve (AUC) values are determined from censored survival data using the Kaplan-Meier method (21). To assess differences between the AUCs in ROC analysis, a DeLong test was applied (Analyse-It software, version 2.20, Microsoft, Redmond, Washington).

To evaluate the “added predictive ability” of both PENK-A and PTA, a net reclassification improvement (NRI) was applied (22). Adapted from the Essen stroke risk score (23), our base clinical model for NRI computation comprised age 65 to 74 years (1 point), age ≥75 years (2 points), arterial hypertension (1 point), hyperlipidemia (1 point), diabetes mellitus (1 point), family history of CAD (1 point), smoking (1 point), concomitant CAD (1 point), and atrial fibrillation (1 point), which was categorized into scores of lower (0 to 2 points), intermediate (3 to 6 points), and higher risk (>6 points). The NRI for PENK-A and PTA was determined based on 3 risk categories using Stata statistical software, Stata/IC 10.1 (StataCorp, College Station, Texas).

All other statistical analyses were performed using R version 2.5.1 (Library Design, Hmisc, ROCR), PASW Statistics version 18.0 (IBM SPSS, Chicago, Illinois), and Statview 4.5 (SAS Institute, Cary, North Carolina).

We evaluated a total of 189 patients who were consecutively admitted to the emergency department for suspected stroke. In all patients, the diagnosis was determined by the duration of symptoms and by CT scan. Among these patients, 124 (65.6%) presented with stroke, 16 (8.5%) with TIA, and 49 (25.9%) with nonischemic events. During 3-month follow-up, 5 patients were lost, and outcome assessment after 3 months was completed in 184 patients (97.4%). The demographic details are given in (Table 1).

The PENK-A concentration was significantly elevated in patients with stroke (median [IQR]): 123.8 pmol/l [93 to 160.5]) compared to patients with TIA (114.5 pmol/l [85.3 to 138.8]) and to patients with nonischemic events (102.8 pmol/l [76.4 to 137.6]; both groups vs. stroke p < 0.05) (Figure 1A). PTA did not show significant differences between these groups (stroke 56.1 pmol/l [39.8 to 73.6]; TIA 46.2 pmol/l [35 to 58.9]; and nonischemic 53.8 pmol/l [39.8 to 66.4]; all nonsignificant) (Figure 1C).

The PENK-A concentration increased in parallel to increasing severity of stroke as defined by NIHSS classes (r = 0.225; p = 0.002) (Figure 2A). The median CT lesion size was 1.5 ml (IQR: 0.1 to 77.5). Similar to the association with NIHSS, concentration of PENK-A correlated with brain lesion size in stroke patients (r = 0.325; p < 0.001) (Figure 2B). The associations with NIHSS and lesion size were not seen for PTA (data not shown). However, concentration of PENK-A positively correlated with PTA (r = 0.641; p < 0.001) (Figure 2C).

Analysis of patient subgroups according to etiology showed a 40% increase of PENK-A concentrations in patients with large-artery atherosclerosis (TOAST1; p = 0.019) (Figure 1B) as compared to patients without stroke.

Adjustment for possible confounders was performed by the multifactorial analysis of covariance for the decadic logarithms of PENK-A (Table 2) and PTA (6). Apart from CT lesion size with PENK-A (p = 0.019) and in a trend with PTA (p = 0.077), associations of both biomarkers were found to be independent of age, sex, cerebro/cardiovascular risk factors, baseline conventional laboratory parameters such as creatinine, creatine kinase, blood glucose, C-reactive protein, platelet count, and medical treatment at the time of admission.

During 3 months of follow-up, a total of 24 patients died (12.6%), and 34 (17.9%) patients experienced the composite MACCE endpoint (recurrent stroke n = 6; 3.2%; myocardial infarction n = 4; 2.1%). The PENK-A levels on admission in the 24 patients who subsequently died were significantly elevated compared to those of survivors (146.6 pmol/l [101.4 to 191.7] vs. 112.1 pmol/l [84.9 to 139.3]; p = 0.004). In patients who experienced the composite endpoint, PENK-A was also elevated compared to patients without adverse outcome (140.7 pmol/l [89.7 to 191.7] vs. 112.0 pmol/l [86.2 to 137.8]; p = 0.002). No significant differences were observed for PTA for both outcome measures.

In Cox proportional hazard analysis, PENK-A predicted all-cause mortality (all log-10 PENK-A in pmol/l: HR: 8.65; 95% CI: 2.7 to 27.6; p < 0.001) and composite outcome of death, recurrent stroke, and myocardial infarction (HR: 8.19; 95% CI: 2.9 to 23.1; p < 0.0001). After adjustment for age, NIHSS, and brain lesion size, PENK-A remained significantly predictive of total mortality (HR: 4.52; 95% CI: 1.1 to 19.0; p = 0.03) and of the composite endpoint (HR: 6.65; 95% CI: 1.8 to 24.9; p = 0.005). For PTA, no significant association with outcome was observed. Patients in the highest quartile of PENK-A (PENK-A >153 pmol/l) had a threefold higher risk at 3 months for mortality (HR: 3.29; 95% CI: 1.47 to 7.35; p = 0.0037), and more than twofold higher risk for the composite endpoint as compared to the lower 3 quartiles (HR: 2.73; 95% CI: 1.38 to 5.41; p = 0.0040) ((Figure 3)A and Figure 3B). Multivariate adjustment did not change the significant findings (mortality HR: 2.40; 95% CI: 1.02 to 5.40; composite outcome HR: 2.23; 95% CI: 1.10 to 4.54; both p < 0.05). The ROC analysis of PENK-A revealed an AUC of 0.64 (95% CI: 0.50 to 0.77) for all-cause mortality, and 0.62 (95% CI: 0.50 to 0.74) for the composite MACCE endpoint. For PTA, we found AUCs of 0.59 (95% CI: 0.46 to 0.71) for death and 0.59 (95% CI: 0.47 to 0.70) for composite endpoint. Comparison of PENK-A and PTA AUCs did not reach a significant level for all-cause death (p = 0.34) and for the combined endpoint (p = 0.55). Comparing ROC curves of both biomarkers to age and NIHSS was nonsignificant as well for all-cause death (PENK-A vs. age, p = 0.48; PTA vs. age, p = 0.84; PENK-A vs. NIHSS, p = 0.12; PTA vs. NIHSS, p = 0.37) and for the composite endpoint (PENK vs. age, p = 0.46; PTA vs. age, p = 0.69; PENK vs. NIHSS, p = 0.23; PTA vs. NIHSS, p = 0.37).

The added value for higher PENK-A and PTA values as compared to the clinical model with 3 risk categories provided for all-cause death an NRI = 20.2% (PENK-A, p = 0.026) and an NRI = 6.0% (PTA, p = 0.439), and for composite endpoint an NRI = 29.2% (PENK-A, p = 0.006) and an NRI = 22.5% (PTA, p = 0.002) ((6) and 6).

Functional capacity was assessed after 3 months, patients have been assessed by mRS. A high mRS score (score ≥3) was found in patients with an enhanced PENK-A concentration compared to patients with a lower concentration (mRS score 3 to 6 vs. 0 to 2, 135.1 pmol/l [IQR: 99.2 to 174.1] vs. 108.9 pmol/l [IQR: 88.6 to 139.5]; p = 0.014) (Figure 3C). Thus, patients with the highest quartile of PENK-A also had a significantly increased risk for the development of moderate to higher full functional disability (mRS score ≥3) at 3 months (unadjusted OR: 2.5; 95% CI: 1.3 to 5.1; p = 0.009). After adjustment for age and brain lesion size, the highest quartile remained a significant predictor of impaired functional outcome (OR: 2.67; 95% CI: 1.14 to 6.25; p = 0.02). When NIHSS was added to the multivariate model, however, only a trend for association remained (OR: 2.4; 95% CI: 0.9 to 6.44; p = 0.08).

Analysis of stroke patients according to TOAST subgroups (TOAST1, n = 47; TOAST2, n = 43; TOAST3, n = 3; TOAST4, n = 5; TOAST5, n = 26) showed that patients with large-artery atherosclerosis (TOAST1) had a significantly increased PENK-A (122.9 pmol/l [95 to 201.8]) expression compared to patients without stroke (Figure 1B). Elevated PENK-A levels in patients with cardioembolism (TOAST2, 127.3 pmol/l [91.6 to 167.9]) did not, however, reach the level of significance (Figure 1B). These results were paralleled for PTA: elevated PTA levels were observed in the TOAST1 subgroup (64.2 pmol/l [48.9 to 91.1] vs. nonischemic; p < 0.05), but not in TOAST2 (50.6 pmol/l [35.3 to 68.5] vs. nonischemic; nonsignificant) (Figure 1D).

Comparing TOAST1 patients to patients with non-LAA revealed a nonsignificant difference for PENK-A levels (TOAST1 vs. non-LAA, 114.8 pmol/l [88.3 to 142]; p = 0.057) and a significant difference for PTA levels (TOAST1 vs. non-LAA, 52.4 pmol/l [38.6 to 66]; p = 0.002).

For both biomarkers, the patient number in other subgroups of stroke etiology (TOAST3 to TOAST5) was too small to reach statistically significant levels between groups (all p > 0.05).

The main findings of the present study are that plasma concentrations of PENK-A are elevated in acute stroke patients compared to patients with TIA and with nonischemic events. The elevation of PENK-A correlated with stroke severity as assessed by NIHSS score and with CT brain lesion size. Further, elevated PENK-A concentrations predicted 3-month outcome for all-cause mortality and for the composite endpoint of death, stroke reoccurrence, and myocardial infarction. Elevated PENK-A concentrations were also predictive of more severe functional disability of patients as assessed by mRS score. In contrast, PTA did not show any discriminative power. This is the first study to evaluate the impact of the precursor neuropeptides PENK-A and PTA on outcome in patients with acute cerebrovascular events.

In subgroup analysis for stroke etiology, a 40% increase of PENK-A concentrations in patients with large-artery atherosclerosis (TOAST1) was observed in comparison to patients without stroke. Increase of PTA levels in patients with large-artery atherosclerosis were less pronounced (+20%) and not different from nonischemic patients and from patients with cardioembolic events (Figure 1D).

Leakage of the BBB integrity is a relevant feature of stroke pathophysiology. Measurement of the mature neuropeptides in plasma is, however, complicated by rapid physiologic degradation of the bioactive peptides and by low in vitro stability. By contrast, assessment of stable fragments of biological inactive precursor peptides has been shown to circumvent this problem as they adequately reflect neuropeptide production and show high stability in human plasma and for >48 h (6- 7). For PTA and PENK, a 90 and 100 times higher concentration has been observed in cerebrospinal fluid compared to plasma. These concentrations gradients are among the highest recognized for a cerebrospinal fluid protein in human physiology.

Earlier studies found that determination of molecular compounds derived from the BBB such as c-Fn and MMP-9 may be useful tools in predicting severe cell injury (24- 25). However, these disruption markers of the BBB are expressed differently, as an increase of c-Fn expression seems to be more adequate for the prediction of hemorrhagic transformation (25). The field of prognostic biomarkers in the ischemic stroke setting is increasingly in the focus of ongoing research, and several candidates are being evaluated such as natriuretic peptides, copeptin, MMP9, Lp-PlA2, and others. Lipoprotein-associated phospholipase A₂ (Lp-PlA1) is involved in the development of arteriosclerosis and plaque rupture and predicts cardiovascular events including stroke (26); Lp-PlA2 has been approved by the Food and Drug Administration for assessing the risk of ischemic stroke. Further work is warranted, however, to establish novel biomarkers in the setting of ischemic stroke for a similar successful use as in the management of cardiac events.

Enkephalins have recently been used for the assessment of neuroinflammation and vascular dementia (27,8). In the present study, an association between high PENK-A expression and prediction of ischemic stroke has been shown, in particular for patients with large-artery atherosclerosis (TOAST1). Intriguingly, PENK-A concentration was significantly lower in patients with TIA compared to patients with ischemic stroke, reflecting the definition of TIA “without evidence of infarction” (4).

An elevated PENK-A concentration showed a poor clinical outcome for cumulative event-free survival and composite endpoint at 3-month follow-up as well as for mRS score. Our study observed a single-center cohort, and the results need to be confirmed in larger and multicenter populations. Three-month mortality was 12.6% in our study, which is in the line with reports from other stroke cohorts in comparable settings (28).

In contrast to the associations of PENK-A with stroke severity and outcome, the results of PTA are more inconsistent, although we found a positive correlation between these 2 BBB disruption markers. This discrepancy cannot be fully explained. It should be taken into account, however, that both markers are not organo-specific, but particularly, PTA is also expressed in inflammatory cells (29) and in the gastrointestinal tract (30). However, multifactorial analysis of covariance revealed that PENK-A and PTA plasma concentrations were independent of possible confounders, in particular, assessing laboratory markers such as creatinine, creatine kinase, blood glucose, C-reactive protein, and platelet count.

Over the last years, several single biomarkers have been shown to help risk assessment in coronary artery disease and ACS, but a multimarker strategy proved to be the most efficient method for early risk stratification and for future clinical practice (31). In comparison to cardiac diseases, there is an unmet need for development of reliable biomarkers in stroke, and a framework of criteria for evaluation of the value of novel biomarkers has been proposed (32). In the line with this consensus statement, our study addresses the early phases of evaluation of biomarkers in a proof-of-concept trial and demonstrating outcome prediction in a prospective cohort.

Pursuing the pathophysiologic concept that acute BBB changes are regulated by mast cells through release of inflammatory and vasoactive mediators, BBB markers such as PENK-A may have a potential role in monitoring therapeutic effects to stabilize BBB after acute stroke (33). Future studies should also evaluate the time course with serial measurements of both biomarkers.

A major limitation to this study is that patients with higher NIHSS score (>12) were not included in the study. The upper limit of NIHSS 12 was a prospectively defined enrollment criterion due to the inability to provide informed consent with more advanced stroke severity. Although the predictive value of PENK-A for the MACCE composite endpoint is interesting, it may seem difficult to explain at this stage on pathophysiological grounds the association with stroke recurrence at the same time.

In our study, we did not pursue a multimarker approach to identify a promising candidate out of a range of variables. We rather measured PENK-A and PTA on the basis of pathophysiologic considerations and on the prospective hypothesis that these markers are promising candidates to indicated BBB damage. Moreover, the sample size of our study cohort was rather small, and the confirmation of our results in larger and independent samples is warranted. Of note, the results of the analysis of covariance with 23 numerators may be blurred by overfitting and possible multicollinearity.

An elevated PENK-A expression was associated with the clinical severity of ischemic stroke, with brain lesion size, and with a poor clinical outcome. PENK-A and other disruption markers of the BBB may be promising targets for biomarker analysis in stroke. Future studies with larger collectives should substantiate PENK-A as a disease marker in ischemic stroke.

For supplemental tables, please see the online version of this article.