|Home | About | Journals | Submit | Contact Us | Français|
Serum biomarkers related to the cascade of inflammatory, hemostatic, glial and neuronal perturbations have been identifed to diagnose and characterize intracerebral hemorrhage and cerebral ischemia. Interpretation of most markers is confounded by their latent rise, blood–brain barrier effects, the heterogeneity of etiologies and the wide range of normal values, limiting their application for early diagnosis, lesion size estimation and long-term outcome prediction. Certain hemostatic and inflammatory constituents have been found to predict response to thrombolysis and worsening due to infarct progression and secondary hemorrhage, offering a potential role for improved treatment selection and individualization of therapy. Biomarkers will become increasingly relevant for developing targets for neuroprotective therapies, monitoring response to treatment and as surrogate end points for treatment trials.
Stroke is a sudden loss of neurologic function resulting from focal disturbance of cerebral blood flow due to ischemia or hemorrhage. Transient ischemic attacks (TIAs) are due to the same physiological process but without acute infarction. Stroke is the third leading cause of death in the USA, and 87% of strokes are ischemic . The expanding use of biomarkers in the field of stroke has made a substantial impact in our understanding of the pathophysiology of stroke and the approach to treatment.
A biomarker is any measurable physiological characteristic or substance that marks the risk for or manifestation of a stroke-related process (Box 1). Several categories of biomarkers are studied in stroke – physical markers, imaging markers, electrophysiological markers, histological markers, genetic markers, systemic (serum) markers and neuronal markers. A physical marker associated with stroke may be as simple as the discovery of hypertension. Neuroimaging is widely used to aid in the diagnosis of stroke, to assist in determining a likely etiology for the event, to estimate severity, and to predict functional outcome and risk of recurrence. Cardiac imaging is used to investigate risk factors and to search for a cardioembolic mechanism. Histological markers can be especially useful in evaluating uncommon causes, such as vasculitis and collagen vascular diseases. Genetic markers can identify heritable cerebrovascular conditions that are associated with stroke and may be useful in individualizing treatments. Serum markers are widely used to test for metabolic risk factors, such as diabetes and dyslipidemia. New research is expanding our understanding of the systemic response to stroke (e.g., inflammation) and demonstrates the potential for applying new serum biomarkers to routine clinical use.
Various markers in nearly all of these listed categories have been evaluated for their association with prospective risk for stroke, for diagnostic purposes and for their value in predicting long-term outcome. Markers of risk factors and prevention strategies have been extensively reviewed in treatment guidelines. The American Stroke Association guidelines for stroke prevention contain a thorough review of the markers of prospective stroke risk for primary and secondary prevention [2–4]. The use of biomarkers in diagnosing stroke and assessing prognosis is an emerging and rapidly evolving field. In this article, we will focus on molecular markers in the cerebrospinal fluid (CSF) and serum, and their use in diagnosing stroke, characterizing severity, optimizing treatment and predicting outcome. This article will provide a brief overview of key cellular mechanisms in cerebral ischemia and hemorrhage, will summarize the important recognized molecular markers of those processes and will discuss the application of these markers to assess a patient’s clinical status, and to guide therapy.
Relatively little biomarker research has been conducted with a focus on hemorrhagic stroke compared with the extensive data on ischemic stroke markers. Relevant data are noted when available, but insufficient data exist to draw broad conclusions regarding how to best translate hemorrhagic stroke biomarkers into routine clinical use.
The CNS cellular response to stroke results in characteristic upregulation and release of particular neuronal markers into the CSF and blood-stream. Increased levels of neuronal isoenzymes and other molecules with CNS-specific expression signal damage to the brain parenchyma. The primary focus of molecular biomarkers of stroke has been their application to diagnosis. The model for a diagnostic marker is the cardiac isoenzyme test. Development of high-sensitivity techniques to detect the cardiac isoenzymes of troponin I and troponin T – products of myocardial degradation – led to the eventual revision of the clinical definition of myocardial infarction to elevate the role of biomarkers in diagnosis . The key property of the troponin test that led to its widespread incorporation into clinical care was a nearly absolute specificity for myocardial tissue (unlike creatine kinase-MB), as well as high sensitivity. Although a troponin-like diagnostic biomarker for stroke would be of interest, several other biomarker applications are possible and more likely to be of clinical importance (Box 2).
Biomarkers in stroke provide insight into the pathophysiologic mechanisms of brain injury. Studies of cytokine markers, for example, have enhanced our understanding of the role of inflammation in the infarct core and ischemic penumbra. The observation that higher levels of certain infammatory cytokines have been correlated to progression of infarct size and clinical deterioration emphasizes this relationship. Certain biomarkers may be useful in differentiating between stroke mechanisms, such as cardioembolism or atherothrombosis. Many of the biomarkers reviewed in this paper correlate with the size and clinical severity of the stroke. The only approved pharmacological therapy for acute stroke, thrombolysis, acts on the hemostatic axis. The activity of the hemostatic system has been shown to predict response in terms of likelihood of recanalization and probability for secondary hemorrhage. Finally, selected serum markers have been shown to correlate with long-term outcome measures.
In stroke, brain injury, ultimately due to necrosis, is a result of the deprivation of metabolic substrates and apoptosis owing to the combined effects of inflammatory and oxidative damage. In cerebral infarction, the inciting event in the ischemic process is obstruction of arterial flow. This obstruction may occur as a result of a myriad of pathological processes. In addition, a variety of pathological processes contribute to the occurrence of intracerebral hemorrhage (ICH). In hemorrhagic stroke, the products of hemorrhage are inherently inflammatory and the swelling is due to inflammation, and bulk interposition of blood into the neuropil raises local pressure, inhibiting effective circulation and leading to secondary ischemia. Given the vast array of pathological conditions that can lead to stroke, an exhaustive list of all markers for each of these etiologies would be unhelpful. Although the triggering events are quite different in ischemic and hemorrhagic stroke, the ultimate mechanisms of neuronal injury and death are similar. Familiarity with the pathways of endothelial, glial and neuronal injury that occur in stroke is crucial for understanding the relevance of identified biomarkers. Comprehensive reviews of the cellular mechanisms of stroke injury are available elsewhere, but a brief orientation will provide the context for discussing the relevance of individual biomarkers. The following paragraph provides an abbreviated and simplified overview of the most studied pathways involved in ischemic cellular damage.
Neurons are heavily dependent on aerobic metabolism and have a disproportionately high energy demand compared with other cell types. In circumstances of ischemia, oxygen and glucose are rapidly depleted, oxidative phosphorylation in the mitochondria ceases within minutes and cellular function becomes deranged. The onset of energy failure triggers a cascade of events that exacerbate and perpetuate damage. Neuronal membranes depolarize owing to failure of the Na+/K+-ATPase pumps, leading to influx of calcium and sodium, and dysfunctional release of neurotransmitters. Excitatory neurotransmitters, such as glutamate, which are released from depolarized cells overstimulate nearby cells, depolarizing them and, thus, inciting an excitotoxic cycle. The cessation of mitochondrial activity increases oxidative and nitrative stress, promoting apoptosis. An inflammatory response begins to develop in the affected area [6–9]. Certain inflammatory cytokines, most importantly TNF, interact with cellular messaging systems and promote apoptosis . Cell membrane dysfunction leads to a net passive movement of water into cells – cytotoxic edema. In response to cytokine signals and oxidative stress, degradative enzymes, such as matrix metalloproteinases (MMPs), are activated and begin to erode the integrity of the blood–brain barrier (BBB) and vascular structures . Within 24 h, the resulting increased permeability leads to vasogenic edema. Ischemia causes autoregulatory mechanisms to fail and prompts upregulation and release of prothrombotic molecules that lead to intravascular thrombosis [12,13]. The increasing local pressure due to cytotoxic and vasogenic edema, and intravascular thrombosis impair the regional microcirculation, and the decrease in cerebral blood flow leads to further ischemia in regions at risk on the periphery of the infarct core. Damaged cells lose membrane integrity and release their contents into the neuropil.
Table 1 summarizes important properties of molecular markers derived from constituents of damaged neurons and glia, elements of hemostasis, cytokines, inflammatory acute-phase reactants, markers of tissue disruption and apoptosis, measurable alterations in oxidative stress and other miscellaneous molecular signs associated with stroke.
S100-β (S100B) is a calcium-binding peptide secreted by astrocytes in the context of brain injury, neurodegenerative processes and psychiatric disorders. The interaction of S100B with the cerebral immune system can lead to enhanced survival of developing neurons at low concentrations, however, at higher concentrations S100B has been shown to stimulate proinflamamtory cytokine production and apoptosis in vitro. Animal studies have demonstrated a potential neuroprotective effect of S100B by activating expression of cellular processes in neurons that block NMDA-induced excitotoxicity . Increased serum levels of S100B are found in association with ischemic stroke, but are also present in many other acute and chronic conditions, such as traumatic brain injury (TBI), Alzheimer’s disease and schizophrenia .
In ischemic stroke, the serum concentration of S100B reaches a maximum concentration between day 2 and 4, and the peak concentration correlates with higher NIH Stroke Scale (NIHSS) scores, infarct size and the clinical course in both infarcts and subarachnoid hemorrhage [16–18]. Possibly owing to its correlation to infarct size, a higher S100B concentration is also associated with higher risk for hemorrhagic transformation after treatment . S100B levels have also been found to be increased in primary hemorrhagic stroke, associated with initial hematoma volume and predicting a worse clinical course with higher morbidity . S100B levels do not necessarily rise quickly. In one study, for example, serum S100B elevation was only 32% sensitive for acute ischemic stroke on admission .
The interpretation of serum S100B levels has been complicated by the fact that S100B was thought to be detected in cell types other than astrocytes and Schwann cells, and high levels were found in the context of a wide variety of other conditions [22,23]. Early attempts to clarify the source by analyzing dimers was unfruitful . Further investigation into the potential extracerebral sources of S100B led to a more rigorous assessment of the immunoassay that was being used, identifying extracerebral sources as false-positive contaminants . Although S100B has now been clearly established as CNS specific, erroneous references to results obtained by the tainted assay continue to appear in the literature . There is evidence from several studies that serum S100B levels are affected by BBB integrity owing to acute disruption or the effect of chronic microvascular disease [26,27].
Glial fibrillary-associated protein (GFAP) is a monomeric intermediate filament protein present in astrocytes and, to a lesser degree, in ependymal cells of the brain, where it functions as a part of the cytoskeleton. GFAP is widely used in neuropathology as an immunohistochemical marker of glial cell lineage, and increased serum GFAP is specific for neurological injury .
In CSF, both S100B and GFAP levels increase early in proportion to size for medium–large-sized strokes, peaking at day 1–2 and returning to baseline by 3 weeks . The serum concentration of GFAP correlates well with that of S100B, as well as to the volume of infarcted brain and the neurological status at discharge from hospital [29,30]. Serum GFAP was found to be more sensitive than S100B as a marker of brain injury in small lesions and minor strokes, but the delayed rise in concentration limits its application as a diagnostic tool [30,31].
Myelin basic protein (MBP) is a hydrophilic protein important for the correct structure of myelin sheaths. Elevated levels of MBP are found in a variety of neurological conditions. For example, MBP levels in CSF has been used as an index of disease activity in multiple sclerosis . MBP has also been found to be elevated in the CSF of patients after stroke . A more detailed study of CSF MBP in the context of acute ischemic stroke found that concentrations were not significantly elevated until 1 week after the ictus, and returned to low levels after 3 weeks . One study sampling the serum levels in patients with ischemic stroke on admission found that MBP was elevated in 39% of patients, and higher peak levels correlated with higher admission NIHSS scores . In a study using serum from subjects who had enrolled in the National Institutes of Neurological Disease and Stroke (NINDS) Recombinant Tissue Plasminogen Activator (tPA) Stroke Study, peak MBP concentration was significantly associated with higher baseline NIHSS scores and larger lesion volumes . The latent increase in this marker is a limiting factor.
Fatty acid-binding proteins (FABPs) are a class of intracellular molecules involved in buffering and transporting long-chain fatty acids. There are nine types of FABPs that are variably expressed in different tissues. These proteins are rapidly released from damaged cells into circulation. Four FABPs are associated with the nervous system, but only two are found in the adult CNS: brain-type (B-FABP) in glia and heart-type (H-FABP) in neurons, with H-FABPs present at more than tenfold higher concentration [34–36]. H-FABP is present in multiple tissue types and has found application as a biomarker for acute myocardial infarction . B-FABP is thought to be present only in central nervous tissue and is undetectable in the serum of healthy individuals [36,38].
Heart-type and brain-type FABPs peak within 2–3 h of stroke onset. The rapid and early peak is attributable to the small molecular size that allows quick passage through the BBB and the short serum half-life of 20 min, owing to renal clearance. B-FABP was found to be most sensitive to small subcortical and lacunar infarcts, but did not correlate to the degree of neurological injury. By contrast, elevation in H-FABP levels was proportionate to infarct size and severity of neurological deficits . A pilot study using H-FABP along with troponin I to exclude a cardiac source found no false positives and 68% sensitivity when sampled on admission, which was a better performance than neuron-specific enolase (NSE) and S100B . Elevated B-FABP concentrations in serum is not specific for stroke, having been detected in a variety of brain injury conditions, including mild TBI and electroconvulsive therapy .
Neuron-specific enolase is one of three recognized forms of enolase, an enzyme in the glycolysis pathway. Although relatively specific for neurons, NSE is also found in neuroendocrine cell lines and their associated cancers. Similar to S100B, NSE concentration rises in the CSF following ischemic stroke as well as other brain tissue injuries, such as subarachnoid hemorrhage, ICH and head injury. Peak concentration after ischemic stroke correlates with baseline NIHSS scores, but not the volume of infarcted tissue . NSE first becomes detectable between 4 and 8 h after the onset of stroke. Although CSF levels of NSE increase in proportion to brain injury with high sensitivity, serum NSE levels are less sensitive and a greater extent of injury is required to produce a consistently measurable change [40–42]. Serum and CSF levels of NSE vary widely across the normal population and no clear relationship has been shown between serum and CSF levels . Although significant trends are detected in various studies, values rise slowly and rarely exceed the reference range of normals (13% within first 24 h, as reported by one study) [40,42,44,45].
Tau protein (TP) is a structural microtubule-associated protein that is well known for its association with a variety of neurodegenerative disorders, including Alzheimer’s disease. The six isoforms of tau are abundant in the brain, facilitating the formation of microtubules by interaction with tubulin . Aside from its pathological aggregation in neurodegenerative disorders, TP has been explored as a biomarker in acute ischemic stroke. Although TP levels sampled at late time points correlate with infarct size and clinical severity, levels rise slowly and only 27% of TP concentrations were elevated over the normal range within the first 24 h . TP levels in CSF have also been studied in the context of acute stroke, showing the same poststroke increase that also correlated to infarct size [47,48]. Further exploration of neurodegenerative-associated biomarkers applied to stroke has been less fruitful. CSF levels of β-amyloid, Apo E and clusterin were not impacted by stroke [47,49].
Neurofilament is a triplet protein that forms part of the structural scaffold of neurons. Neurofilament light protein (NFL) is the subunit that forms the core of the filament. Increased levels of NFL have been observed in a variety of neurodegenerative diseases. Increased CSF concentration of NFL has also been shown in a small series of patients following ischemic stroke .
N-acetylaspartate (NAA) is a free amino acid representing approximately 1% of the dry weight of the brain. It is synthesized in neurons and functions as a molecular water pump, drawing water out of neurons and through the overlying myelinating oligodendrocytes where NAA is recycled. Owing to the neuron’s high metabolic rate compared with other tissue types, metabolic water must be removed at approximately 12-times the rate of that removed in the rest of the body. Therefore, brain NAA is heavily utilized, completely replenishing every 16.7 h . Perturbations in NAA can be identified by magnetic resonance spectroscopy and are useful in characterizing brain lesions. A study investigating NAA in acute ischemic stroke found that serum concentrations were significantly higher in all patients on admission and all points sampled until 48 h. NAA concentration was maximal at time 0 and declined after-ward . This is consistent with prior findings that NAA levels decrease rapidly after damage because the rate of efflux from brain exceeds that of metabolism [52–54].
Combining information about NAA levels with traditional MRI sequences can predict the outcome in patients, where those with lower NAA levels in ischemic tissue do more poorly . Magnetic resonance spectroscopy shows that NAA levels in ischemic injury continues to fall gradually for more than 1 week postinfarction, although any NAA detected beyond 24 h may simply be trapped in cellular debris [54,56,57]. Early reperfusion has been shown to cause a quick, transient increase in serum NAA concentration . This may be due to resumed synthesis from restored metabolic function during a time of persistent BBB dysfunction, or simply from efficient clearance of NAA from the interstitium by restored regional blood flow.
The NMDA glutamine receptor is implicated in mediating the excitotoxic response in cerebral ischemia. Autoantibodies (aAbs) to the NR2 subtype of the NMDA receptor have been demonstrated to be associated with neurotoxicity. A pilot study to examine serum levels of the neurotoxic biomarkers, glutamate, homocysteine and NMDA NR2 aAbs, found that subjects presenting with TIA had significantly elevated levels of NMDA NR2 aAbs compared with controls, and those with ischemic stroke had even higher concentrations . Interestingly, in the pilot study and a follow-up larger trial, patients with ICH demonstrated no significant elevation of NMDA NR2 aAbs whereas patients with both TIA and ischemic stroke demonstrated markedly elevated levels that correlated with the NIHSS score and diffusion-weighted imaging (DWI) lesion volume . In addition to showing high sensitivity, the NMDA NR2 aAb assay may be one of the only biomarkers of stroke that can reliably differentiate between hemorrhagic and ischemic stroke, along with DNA and Apo CI and CIII levels [25,61].
Visinin-like protein (VLP)-1 is the human homolog of a protein that was identified by screening mice with gene-array analyses for biomarkers that are preferentially and abundantly produced in brain. VLP-1 was then detected in CSF using a rat model of ischemic stroke, and in the serum of humans after stroke. VLP-1 is a neuronal Ca2+ sensor protein expressed primarily in brain but also found in other tissues . VLP-1 was detected in 8% of controls with no stroke and in 44% of subjects with stroke in samples taken within the first 24 h after onset of stroke . Although the approach to screening potential brain biomarkers is novel, these early results are not compelling.
Thrombomodulin (TM) is an endothelial membrane protein that plays an important role in hemostasis by binding thrombin and activating protein C, thereby exerting an antithrombotic mechanism. TM expression in the brain is controlled through astrocyte regulation of TM mRNA expression . TM is present in serum and urine in a soluble form. Various TM gene polymorphisms have been associated with coronary events and stroke, although serum TM levels are not correlated . A study investigating 492 stroke patients and matched controls found that serum TM did not correlate to stroke exposure, however, among case subjects, high TM levels were associated with a higher death rate .
A small pilot study found that TM was elevated on admission in 43% of patients, and peak levels correlated with discharge modified Rankin scale scores but was contradicted by a larger study that found no significant difference between subjects with stroke and controls on admission or at 1 month [21,67]. Both this study and another by the same group found elevated serum levels of TM only in the subacute phase [67,68]. Likewise, serum TM concentration from NINDS recombinant tPA Stroke Study subjects did not correlate to NIHSS score or stroke lesion volume at any time point or functional outcome . Soluble TM in serum does not appear to be a product of endothelial degradation, but rather a regulated component in the body’s complex hemostasis axis. At present, the significance of serum TM concentration is not clear.
Abnormalities in the hemostasis axis are an uncommon cause of ischemic stroke [69–71], nevertheless, perturbations are seen following stroke onset that may aid in diagnosis, be informative about the severity of the event and predict response to therapies. Some coagulation markers have been used extensively for diagnostic purposes for other conditions. The d-dimer test in cases of suspected pulmonary embolism is a classic example . The limitation to the use of coagulation markers is that they are nonspecifc in nature.
An increase in fibrinogen levels is seen after ischemic stroke and is associated with infarct size and outcome [73–75]. High levels of d-dimer, a small protein fragment produced by fibrin degradation, predicts progression in ischemic stroke and poor outcome in ICH [76–79]. The case is similar for other bio-markers of hemostasis: higher concentrations of fibrinopeptide A, b-thromboglobulin, prothrombin fragments 1 and 2, thrombin–anti-thrombin complexes, platelet factor 4 and von Willebrand factor have all been associated with a worse clinical course in ischemic stroke and increased mortality [76,80,81]. Fibrinopeptide A, thrombin–antithrombin complexes, d-dimer and, possibly, β-thromboglobulin and platelet factor 4, are elevated in patients presenting with TIA compared with matched controls, suggesting a dysfunction in coagulation and platelet activation [82–84]. Several studies that have looked concurrently at multiple hemostatic biomarkers have concluded that the d-dimer level is very sensitive to stroke, it may help differentiate between stroke etiologies and is the strongest prognostic indicator for worsening or recurrence within the subacute period [79,85,86].
Constituents of the hemostasis axis may provide information beyond identifying stroke occurrence and severity. The activity of many agents in the coagulation and fibrinolysis cascade increase in the context of stroke. Measuring the activity of various components may be useful in predicting response to therapies or for individualizing treatments. Patients presenting with ischemic stroke whose international normalized ratio from warfarin treatment is therapeutic have smaller infarcts and perfusion deficits, lower NIHSS score on admission and lower modified Rankin score at discharge . Plasma levels of thrombin-activable fibrinolysis inhibitor (TAFI) are elevated in the acute phase of ischemic stroke, and plasminogen activator inhibitor (PAI)-1 is elevated in patients that show progression of intracranial atherosclerosis, along with other proinflamamtory markers [88,89]. Patients with high levels of PAI-1 are less likely to achieve recanalization after receiving tPA and experience poorer outcomes . Baseline levels of TAFI and PAI-1 also predict symptomatic hemorrhage after tPA treatment . Similarly, polymorphisms of both the TAFI and PAI-1 genes influence the probability of recanalization after tPA treatment . A point-of-care assay to measure the activity of these fibrinolysis inhibitors could allow for adjustments to thrombolytic dosing, which may increase efficacy in patients with high inhibitor levels and reduce the likelihood of hemorrhagic transformation in patients with a concerning profile.
Cytokines are extracellular proteins and glyco-proteins that function as short-range molecular messages, exerting a paracrine or autocrine effect on local tissue. Triggered by a variety of stimuli, cytokines are critical mediators of the tissue inflammatory and immune responses . Inflammatory reactions are intricately mediated by proinflamamtory and anti-inflammatory pathways that often operate in tandem to regulate the extent and intensity of the response. There has been significant interest in delineating the role of proinflamamtory cytokines in the perpetuation and propagation of damage postinfarction. Since inflammation generates the formation of edema and can promote apoptosis, inflammatory biomarkers are potentially useful for prognosis and as prospective targets for neuroprotective therapies. Although none of the cytokines that activate in the brain following stroke are CNS specific, careful investigation has confirmed that the elevation in plasma levels of these signaling molecules is owing to their production in the CNS and not from peripheral blood cells .
Serum sampling of patients presenting with ischemic stroke has shown elevation in many cytokine markers, including IL-1, IL-6, IL-8, IL-10, IL-13, TGF-β, ICAM-1, VCAM-1, E-selectin, L-selectin, P-selectin, TNF-α, chitotriosidase and monocyte chemoattractant protein-1 [73,89,94–101]. Levels of soluble selectin proteins are increased following stroke, indicating the activation of leukocyte trafficking . Elevated levels of proinflamamtory cytokines and low levels of anti-inflammatory markers are associated with severity, infarct volume, clinical worsening and long-term outcome [102–104]. Elevated levels anti-inflammatory mediators, such as IL-6, IL-10 and IL-1-receptor antagonist, concurrent with elevated proinflamamtory markers suggest a complex interaction of pro-and anti-inflammatory mechanisms at work in the hours and days following ischemic and hemorrhagic stroke [73,96,100]. The appearance of TNF-α and IL-6 has also been correlated to the volume of perfusion-impaired tissue, suggesting the role of inflammation in the ischemic penumbra . Although systemic anti-inflammatory medications, such as corticosteroids, have not proven to be effective treatments for stroke, inflammatory biomarkers have been used as surrogate markers of efficacy for neuroprotective medications [98,105].
C-reactive protein (CRP) is an acute-phase protein associated with inflammation. Produced in the liver in response to IL-6, CRP binds and aggregates a variety of soluble ligands and activates the classical complement pathway . More recent research has also advanced a hypothesis that CRP may directly cause disruption of the BBB . Owing to the inflammatory nature of atherosclerosis, modest elevations in CRP levels have been correlated to increased risk for stroke in large population studies, independent of other traditional cardiovascular risk factors . CRP and fibrinogen levels rapidly increase as a result of stroke, and CRP on admission has been shown to predict mortality in tPA-treated patients [73,99,109]. As in the case of many other inflammatory biomarkers, levels of CRP are correlated with infarct volume, stroke severity and long-term outcome . Levels of CRP and other acute-phase markers, such as the erythrocyte sedimentation rate, remain elevated up to 3 months following stroke . Higher CRP levels measured within the first day of TIA or stroke onset and 3 months later are both associated with an increased risk for recurrent stroke following the initial event [110,111].
Astrocytes and microglia that are involved in an inflammatory response are stimulated to release extracellular matrix degrading proteases. The most well recognized of these proteolytic enzymes are serine proteases and MMPs. MMPs are a family of 14 enzymes, a subset of which have been found to be active in brain tissue. The various MMPs function as stromelysines, gelatinases, colagenases, a membrane-type proteinase and a matrilysin. Serine proteases include elastase and plasminogen activators. Owing to their toxic potential, these enzymes are produced in latent form and require activation. Once active, MMPs are implicated in opening the BBB, damaging capillaries, perpetuating inflammation and, thereby, promoting secondary progression of ischemia and hemorrhagic transformation . Inflammation and associated alteration of the vascular wall structures due to MMP activity have been implicated in the chronic, progressive damage of intracranial atherosclerosis, as well as acute-phase injury . Various cytokines act on astrocytes and microglia in order to promote the production of MMPs and serine protease. The activity of these matrix-degrading systems are regulated by the opposing effects of tissue inhibitors of metalloproteinases (TIMPs).
Matrix metalloproteinase levels are elevated after both ischemic and hemorrhagic stroke, and serial measurements of MMP-2 and MMP-9 levels demonstrate a correlation with proinflamamtory cytokine expression [95,112,113]. In ischemic stroke, serum concentration of MMP-9 correlates to NIHSS scores and initial and final DWI infarct volumes [97,100,114], whilst peak concentration is associated with late hemorrhagic transformation and post-tPA hemorrhage [115–118]. MMP-9 and MMP-13 are correlated with DWI expansion . In hemorrhagic stroke, higher MMP-9 is associated with perihematomal edema, this is probably owing to its role in opening the BBB. The MMP inhibitor TIMP-1 is negatively correlated to the extent of edema. MMP-2, MMP-3 and TIMP-2 all show characteristic rises after ICH .
TNF is an inflammatory cytokine with a myriad of effects. It can be expressed by endothelial cells, microglia and astrocytes in response to injury, where it exerts a local proinflamamtory and procoagulant effect. There is also animal-based experimental evidence implicating TNF in the pathophysiologic processes that trigger stroke. Once a stroke has occurred, TNF activates MMP and, thereby, increases vascular permeability, promoting edema formation and entrance of proinflamamtory plasma constituents through the impaired BBB. Importantly, TNF stimulates the production of reactive oxygen species (ROS), potentiates glutamate-mediated excitotoxicity and directly signals cells to induce apoptosis [10,100]. The combined effects on the microvasculature impairing local microcirculation, promoting excitotoxicity, facilitating edema formation, inflammation and apoptosis stimulation all contribute to expand the zone of ischemic tissue from the infarct core out into the penumbra.
Neuronal apoptosis is the pathway by which damage extends in a delayed fashion beyond the infarct core that undergoes necrosis. The importance of apoptosis pathways in stroke has been demonstrated in multiple studies, including neuroprotective trials in which blockade of important components of the apoptosis cascade substantially reduced infarct volumes in animal models . Caspases are a family of cysteine proteases that constitute the main molecular cascade in the apoptosis pathway. Initiator caspases activate precursor forms of effector caspases, and those effector caspases, in turn, degrade specific molecular targets within the cell. Caspase-3 is an effector that functions as the final step in the chain of reactions that results in DNA fragmentation, and has been explored as a potential biomarker that would estimate the degree of active apoptosis [122–124]. The basis for interest in caspase-3 is due in part to animal studies demonstrating a potential therapeutic role in decreasing the activity of this enzyme. In a study investigating human stroke patients, plasma levels of caspase-3 were elevated in acute stroke, and correlated to infarct growth and short- and long-term neurological outcome . The Fas system is a parallel pathway that induces apoptosis through a receptor in the TNF-receptor family . Decreased levels of soluble Fas-variant molecules in the serum that act as apoptosis inhibitors, along with increased stimulatory Fas ligands in the perihematomal region measured in deceased patients with ICH shows a correlation between perihematomal edema growth and these markers of active apoptosis .
Circulating levels of DNA measured in plasma increase in response to a variety of injurious conditions, including stroke. Plasma concentration of DNA is elevated within 3 h of ischemic and hemorrhagic stroke onset, and patients who experience a poor outcome or die as a result of their stroke have higher levels than those who do not. The level of DNA also correlates to the volume of hematoma in ICH . DNA concentration is significantly higher among patients with hemorrhagic versus ischemic stroke, and the combination of DNA and S100B levels has been proposed as a way to differentiate those conditions using molecular biomarkers .
Metabolic perturbations caused by ischemia lead to an increased production of ROS, such as free radicals, superoxide radical anions and hydrogen peroxide. At the same time, mitochondrial function fails and the brain’s limited stores of antioxidant substances are rapidly exhausted. Reperfusion may actually exacerbate the degree of oxidative stress by repleting oxygen in ischemic tissue. ROS activate MMPs, whose function in broadening the zone of ischemic damage was described previously. In addition to the secondary ischemia that occurs in hemorrhagic strokes, iron released from hemoglobin degradation also catalyzes hydrogen peroxide formation. As many ROS are inherently unstable, biomarkers are necessary as a surrogate method of measuring oxidative stress load. Candidate molecules are those whose chemical structure is altered by oxidation into a new stable entity in a predictable pattern. Oxidative damage occurs via lipid peroxidation, DNA oxidation and protein oxidation. Each of these processes produce markers that can be measured in the serum, along with circulating levels of antioxidants [8,128–130].
Measuring the plasma concentration of various antioxidants is confounded by a few factors. First, an increase in oxidative stress can upregulate the production of antioxidant enzymes, such as members of the superoxide dysmutase family. Second, delivery of peripheral antioxidant molecules is limited owing to impaired circulation and, therefore, the depletion of body stores is unpredictable. Finally, there is wide individual variation in baseline levels of antioxidant activity that may obscure the effects of oxidative stress in an individual case. Stroke patients have been found to have lower levels of retinol, ascorbic acid, uric acid, a-tocopherol and carotenoids, but many studies have demonstrated inconsistent or negative results. Markers of lipid peroxidation include malondialdehyde, thiobarbituric acid-reactive substances and F2-isoprostanes. Of these three groups, F2-isoprostanes have the most favorable chemical characteristics for use as a reliable marker, but clinical studies thus far have not produced definitive results . For DNA oxidation, the measurement of 8-hydroxy-2-deoxy-guanosine demonstrates promise . There is limited literature on biomarkers of protein oxidation, but measuring the carbonyl content of IgG has been proposed as feasible .
Albumin is the most abundant protein found in plasma, functioning as a nonspecific carrier molecule and maintaining oncotic pressure. In the context of ischemia, structural changes occur at the N-terminus of the protein, possibly as a result of exposure to ROS. This change in the albumin protein is detected by a cobalt-binding test. Over the last decade, increased ischemia-modified albumin concentration has been linked to acute myocardial ischemia, limb ischemia, mesenteric ischemia and deep venous thrombosis [133–136]. Serum levels of ischemia-modified albumin are increased in subjects with both ischemic and hemorrhagic stroke compared with controls at baseline .
Nucleoside diphosphate kinase A (NDKA) and PARK7 (also called DJ-1) were selected as potential biomarkers of stroke by screening samples of postmortem CSF for markers that are not found in the CSF of healthy controls. PARK7 is a redox-sensitive molecular chaperone activated in the context of oxidative stress. NDKA is a kinase that catalyzes the transfer of the terminal phosphate from ATP to nucleotides. Using plasma samples from several cohorts, one team of investigators demonstrated that concentrations of NDKA and PARK7 were both increased within 3 h of stroke, with higher sensitivities and specificities than those typically observed using other individual stroke biomarkers . These findings await validation. Furthermore, neither of these markers are specific to the CNS and both have been implicated in the pathogenesis of neurodegenerative disorders, which may complicate their application.
Polyamines complex with RNA and facilitate translation. The polyamine spermine has been evaluated in acute stroke, together with spermine oxidase (SMO) and acetylpolyamine oxidase (AcPAO), enzymes that catalyze its degradation, and protein-conjugated acrolein, one of its metabolites. Following acute stroke, plasma levels of spermine decrease in tandem with an increase in SMO, AcPAO and acrolein. Researchers have hypothesized that SMO and AcPAO are released from CNS tissue following stroke and catalyze the degradation of spermine into acrolein in the periphery. Unfortunately, the sensitivity and specificity of all these markers is poor compared with controls, and levels of these markers are altered by other conditions, such as renal dysfunction .
Ubiquitin fusion degradation protein (UFDP) is an enzyme in the ubiquitin degradation pathway that is expressed in many body tissue types. UFDP has been tested as a stroke biomarker in several cohorts along with glutathione S-transferase P, a molecule involved in detoxifying ROS. The sensitivity of UFDP ranged from 62 to 95%, but only when the cutoff levels were adjusted by a factor of 8 to optimize the outcome [140,141]. Tested in a single cohort, glutathione S-transferase P1 was 72% sensitive .
The introduction of a new biomarker has the potential to profoundly revolutionize the field of stroke if it targets an appropriate need. The use of the neuroimaging biomarkers applied in TIA is a case in point. The purpose of defining TIA was to identify cases of cerebral ischemia that do not involve tissue infarction. The early definition of TIA included cases whose deficits persisted up to 24 h . Over the last several decades, work using imaging biomarkers identified DWI evidence suggestive of infarction in up to half of patients who were clinically diagnosed with TIA, mostly those with deficits lasting from 1 to 24 h [143,144]. These biomarker data led to a widespread reconsideration of the definition of TIA with a move to employ neuro-imaging evidence for infarction as indicative of stroke, regardless of the resolution of clinical symptoms .
First, nearly all of the serum biomarkers research that has been reported thus far in the field of stroke is exploratory in nature. As previously noted, some markers have been explored because of their association with known mechanisms of stroke, and others have been selected empirically by screening postmortem CSF or performing comparative proteomic analysis [61,140,146]. A consortium is in place (the Human Brain Proteome Project) to facilitate the identification of potential brain markers with proteomic techniques . At present, serum biomarkers have performed poorly when tested prospectively for use in acute diagnosis . There are several reasons. Although a difference in mean concentration of these markers can be demonstrated between a group of stroke patients versus controls, most demonstrate considerable overlap in their distribution. Significant heterogeneity exists in the reference cutoffs used and between the populations that these markers are tested in, even between different cohorts evaluated by the same research group [41,140,141]. In some studies, a cutoff was selected to set specificity at 100% based on a small set of controls since the normal distribution of the marker had not yet been established . When a marker is used as a diagnostic test, a diagnosis is made only when the level is less than or greater than the range that encompasses 95% of normal individuals. Therefore, a clear statistical difference does not translate into a high sensitivity test when the mean levels of the exposed and unexposed populations both lie within the normal range . Many biomarkers of stroke rise slowly after the event. Owing to the delayed change, they are not useful in detecting a stroke in an emergency department setting, and furthermore, only subacute levels correlate well with other outcome measures [18,21].
The heterogeneity of the causes of stroke further complicates interpretation. As with the diagnosis of stroke, other correlations between biomarker levels and severity of outcomes are generated by demonstrating a difference in mean levels between groups with differing clinical status. What is true for a population may not be accurate on an individual level. A patient with infectious endocarditis may have very high elevations in inflammatory and cytokine markers despite having a small infarction, for instance, which would confound the normal correlation with infarct size. Further confounding evidence exists owing to the many other comorbid processes that can alter CNS biomarker levels, including migraine, seizures, neurometabolic disorders, TBI, chronic microvascular white matter disease and, possibly, dementia – a list that encompasses many of the important stroke mimics to be excluded by a diagnostic test [27,148–150].
The biomarker characteristics of the molecules reviewed here cannot be interpreted without recognizing the complex role they play in their highly regulated systems. The biology of IL-6 is a good example. Elevated levels of IL-6 have been correlated with both an increased and decreased risk for worse outcome in various studies [91,92,94–96]. This is because the signaling outcome of IL-6 depends on the milieu of other cytokines and the cellular state, exerting either a proinflamamtory or anti-inflammatory effect. Changes that take place in the acute to subacute period can also affect biomarker levels but otherwise may not be apparent, making it difficiult to impossible to control for those effects. NAA levels, for example, can quickly spike owing to reperfusion, but it is frequently not possible to know whether, or at what point, reperfusion has occurred with respect to when a sample was drawn. Perhaps the most critical confounder of biomarker concentrations is the effect on the BBB.
Biomarker concentration shifts are generally more pronounced in the CSF, which receives cellular degradation products directly, than the serum. CSF sampling avoids the confounding effect of the BBB, but is not practical for routine use when reliable clinical and imaging markers exist for diagnosis and prognosis. Even in the CSF, many of the described biomarkers demonstrate age dependency for the range of normal reference values [150,151]. The BBB plays an important role in mediating the release of neuronal and glial cellular constituents to the systemic circulation. In tissues outside the CNS, products of cellular degradation are either swept directly into circulation or are spilt into the interstitium and pass into circulation via lymphatic channels. By contrast, the unique interaction of astrocytic foot processes (or glia limitans) and endothelial cells seal the CNS from systemic circulation exposure by a series of tight junctions. This prevents the passive movement of cellular constituents from the CNS into circulation. A similar barrier exists between the CSF and circulation in the choroid plexus . Alterations in the integrity of the BBB can alter the ability of CNS molecules to enter circulation. In instances such as stroke, the same pathological processes that damage neurons and glia disrupt the BBB, causing release of cellular constituents and their passage into systemic circulation. A recent study of BBB function after stroke demonstrated that lacunar stroke patients had widespread BBB disruption, indicating that the relationship between BBB and infarct size is not proportional . The BBB dysfunction may be due to the stroke or may be baseline dysfunction, as patients with white matter disease (leukoaraiosis) are known to have disrupted BBB function . Since the extent of BBB integrity and extent of brain tissue damage are not strictly correlated, serum concentrations of CNS biomarkers may not correlate well to the rate of biomarker release within the CSF or the absolute CSF concentration .
The dependency of serum biomarker levels on BBB function was directly demonstrated in a study of patients who underwent BBB disruption with intra-arterial mannitol . A mathematical model to account for potential BBB effects has been proposed for S100B, but the model was based on one sample set of patients and is not applicable to other markers . Biomarkers can pass into circulation even when the BBB is fully intact. All neuronal biomarkers except B-FABP have been detected in the serum of normal individuals. The ability of various molecules to pass through an uncompromised BBB may be mediated, in part, by molecular weight and solubility, and the movement of some markers into serum appears to be unrelated to BBB function [155,156]. Interpreting serum biomarker concentrations is challenging as perturbations may represent BBB dysfunction alone, concurrent BBB and tissue damage or substantial tissue damage with good preservation of BBB integrity.
Attempts at using molecular markers to estimate the integrity of the BBB have typically required the sampling of the CSF, although the serum level of the monomeric form of transthyretin, a protein produced by cells in the choroid plexus, may be useful [155,156]. The benefit of transthyretin is that it is not expressed in neurons or astrocytes and, therefore, may be independent of parenchymal brain damage.
To consider the potential applications of biomarkers to stroke, refer to the various potential biomarker applications that are summarized in Box 2. Several well-studied biomarkers are nonspecifc CNS constituents and others, such as NDKA and PARK7, have been selected without regard to pathways of ischemic damage. By contrast, some of the markers that have been studied were selected from knowledge of their cellular function. Research that has clearly demonstrated associations between these molecules and the occurrence of stroke has clarified and validated hypotheses regarding the pathways of cellular damage in stroke. Monitoring the activity of molecules known or suspected to be involved in cellular injury has been shown to be a valid method of exploring pathological mechanisms. The utility of biomarkers in delineating and confirming pathological molecular pathways will ensure they continue to be a resource in stroke research.
Several potential clinical applications have been considered – facilitating diagnosis, identifying stroke mechanisms, characterizing size and clinical severity, identifying an ischemic penumbra and estimating risk of progression or worsening. Many molecular markers have been associated with brain ischemia, but at present they offer little additional information on an individual basis beyond that gained from standard clinical evaluation and neuroimaging. From a diagnostic perspective, a major current challenge is to identify a way in which these markers can add value to existing routinely obtained data. Some neuroimaging markers in stroke face similar obstacles in gaining widespread application. Many correlations have been drawn between imaging parameters (e.g., diffusion- and perfusion-weighted imaging volumes, degree of leukoaraiosis and location) and outcomes (e.g., length of hospital stay, long-term disability, odds of infarct progression and clinical severity) [157–162]. Despite these many detailed associations, the clinical evaluation of the patient is still more informative [163–165]. There have been several attempts to address these limitations by developing more complex models that measure up to 50 markers simultaneously. Sets of four to five of the most predictive markers from these multimarker studies have achieved sensitivity and specificity up to just over 90% [166–168]. The addition of a few basic demographic data (age, sex and presence of atrial fibrillation) or initial computed tomography results slightly improves performance . The strategy in these studies was to combine sensitive, but nontissue-specific, markers with at least one marker specific to the CNS. The use of nonspecific biomarkers may be necessary to achieve useful sensitivity. Furthermore, biomarker testing in one of the studies was performed on a point-of-care platform that is feasible in an emergency department setting .
It is unlikely that a single biomarker or panel of markers will supplant the current method of diagnosing stroke by a clinical and radiographic approach. Established imaging techniques also continue to be superior at characterizing infarct and penumbra size. MRI equipment is increasingly available at community hospitals. The most promising application for stroke biomarkers is in realms where current clinical capabilities are limited – in accurately determining stroke etiology, in predicting response to treatment, in individualizing treatment and in monitoring treatment effects. One study utilizing a combination of inflammatory, hemostatic, neural and apoptotic biomarkers (CRP, d-dimer, soluble receptor for advanced glycation end products, MMP-9, S100B, brain natriuretic peptide, neurotrophin-3, caspase-3, chimerin and secretagogin) found that it was useful in differentiating cardioembolic from other types of strokes . Given the correlation between the serum levels and genetic polymorphisms of hemostatic modulators, such as TAFI and PAI-1, and tPA recanalization, including the association between several biomarkers (NSE, TAFI, PAI-1, d-dimer and MMP-9) and the risk for secondary hemorrhagic transformation, better selection of thrombolysis candidates and individualization of thrombolytic doses may be possible [170,171].
Finally, if the correlations between serum biomarkers of stroke and patient outcome become more firmly established, those markers should be used as surrogate outcome measures. Serum stroke biomarkers have already been used to monitor efficacy in acute several stroke drug trials. Two trials studying the effects of simvastatin in acute ischemic stroke and vasospasm in aneurysmal subarachnoid hemorrhage used panels of biomarkers as surrogate markers of ischemic injury, and a trial of arundic acid, a modulator of astrocyte activation, used the astrocyte-associated marker S100B as a surrogate marker of neurological deficit [98,172,173]. New US FDA rules allow for drug approval based on one large Phase III trial with supportive evidence from a Phase IIB trial demonstrating an effect on a relevant biomarker . As a result of the enormous expense and complexity of conducting a clinical trial in acute stroke, the use of well-established biomarkers could have a significant impact in drug development and approval. Aside from the need for more firmly established correlations between biomarkers and outcomes, the crucial impediment to using biomarkers in clinical trials is adjusting for the effects of the confounders, as discussed in the limitations section previously – systemic inflammation, natural variability in levels and comorbid processes that activate glia or damage the CNS. Furthermore, biomarker levels generally correlate to the volume of CNS damage, whereas stroke-related disability is a function of many factors – volume, localization, age, comorbidities and psychosocial determinants. To date, since most biomarker research has been exploratory in nature, the lack of rigor in study designs has impeded the development of biomarkers as outcome measures.
In a recent commentary on the use of biomarkers in stroke, Hill posed two “key clinical questions in the acute phase:
Perhaps more important is the question, what value does this test add beyond the clinical examination and neuroimaging? Serum biomarkers related to the cascade of inflammatory, hemostatic, glial and neuronal perturbations have been identified to diagnose and characterize ICH and cerebral ischemia. Interpretation of most markers is confounded by their latent rise, BBB effects, the heterogeneity of conditions that cause stroke and the wide range of normal values. In the realm of stroke diagnosis, no serum biomarker or panel of markers has been proven to add to the data obtained through the history, physical examination and imaging. Associations have been identified between biomarker concentrations and lesion size, clinical severity, worsening due to infarct progression and secondary hemorrhage and long-term outcomes. Animal-based research and human trials that have used biomarkers of stroke to identify therapeutic targets and monitor response to treatment demonstrate promising potential applications.
Given the widespread availability of advanced neuroimaging technology and the many factors that confound interpretation of serum biomarker levels, stroke diagnosis may not be the most relevant application for molecular markers at present. Intriguing correlations between hemostatic and inflammatory markers and clinical outcomes, such as response to thrombolysis, infarct progression and secondary hemorrhage, offer an opportunity to use biomarkers in the acute setting to individualize treatment selection and dosing. Biomarkers of inflammation and tissue disruption have become useful in developing a more complete understanding of the pathophysiology of stroke. New therapies have already been tested in animals to modulate the tissue response to injury. Relevant markers will continue to play an important role in developing targets for neuroprotective therapies. Finally, biomarkers may play an important role in monitoring response to treatment and as surrogate end points for treatment trials. A transition from the current exploratory phase to achieve further integration of biomarkers into clinical care will require large-scale, well-designed, prospective research to yield well-controlled outcome associations.
For reprint orders, please contact: moc.enicidemerutuf@stnirper
Financial & competing interests disclosure
Mathew B Maas and Karen L Furie are supported by NIH (grant P50NS051343). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.
No writing assistance was utilized in the production of this manuscript.
Papers of special note have been highlighted as:
of considerable interest