Advances in the development of biomarkers for epilepsy

Review

Advances in the development of biomarkers for epilepsy Asla Pitkänen, Wolfgang Löscher, Annamaria Vezzani, Albert J Becker, Michele Simonato, Katarzyna Lukasiuk, Olli Gröhn, Jens P Bankstahl, Alon Friedman, Eleonora Aronica, Jan A Gorter, Teresa Ravizza, Sanjay M Sisodiya, Merab Kokaia, Heinz Beck

Over 50 million people worldwide have epilepsy. In nearly 30% of these cases, epilepsy remains unsatisfactorily controlled despite the availability of over 20 antiepileptic drugs. Moreover, no treatments exist to prevent the development of epilepsy in those at risk, despite an increasing understanding of the underlying molecular and cellular pathways. One of the major factors that have impeded rapid progress in these areas is the complex and multifactorial nature of epilepsy, and its heterogeneity. Therefore, the vision of developing targeted treatments for epilepsy relies upon the development of biomarkers that allow individually tailored treatment. Biomarkers for epilepsy typically fall into two broad categories: diagnostic biomarkers, which provide information on the clinical status of, and potentially the sensitivity to, specific treatments, and prognostic biomarkers, which allow prediction of future clinical features, such as the speed of progression, severity of epilepsy, development of comorbidities, or prediction of remission or cure. Prognostic biomarkers are of particular importance because they could be used to identify which patients will develop epilepsy and which might benefit from preventive treatments. Biomarker research faces several challenges; however, biomarkers could substantially improve the management of people with epilepsy and could lead to prevention in the right person at the right time, rather than just symptomatic treatment.

Introduction More than 50 million people worldwide have epilepsy and an estimated 2·4 million people are newly diagnosed with epilepsy each year.1 However, despite over 20 antiepileptic drugs being available, seizures remain uncontrolled in about 30% of patients. The research community has focused substantial attention on developing antiepileptic drugs with increased efficacy and fewer adverse effects.2 Additionally, awareness of the need to identify individuals at risk of developing epilepsy and to develop preventive or disease-modifying treatments is increasing. One of the major impediments to improving epilepsy treatment options is the heterogeneity of epilepsies. Many different genetic and pathophysiological factors, alone or in combination, can underlie an increased risk of developing a seizure disorder. Among these are different forms of brain injuries such as stroke, traumatic brain injury (TBI), or perinatal and prenatal injury, and CNS malformations or tumours. This broad variety of disease entities implies that a large range of mechanisms can lead to establishment of an epileptogenic focus, and that potentially diverse mechanisms of functional impairment and seizure generation exist. Moreover, even in the comparatively homogeneous groups of patients in whom an initial injury can be identified, such as those with stroke or TBI, on average, less than 20% of patients will develop epilepsy within 1–2 years,3 implying further complexity and potentially intra-syndrome heterogeneity. Accordingly, awareness is increasing that both antiseizure (ie, drugs used to treat epileptic seizures) and antiepileptogenic (ie, drugs used to prevent the development of epilepsy or its progression) treatments need to be developed and prescribed on an individualised basis. This is particularly relevant for antiepileptogenic drugs that would ideally be tailored to the cause for developing epilepsy (eg, a brain injury or genetic mutation). Consequently, identification of biomarkers www.thelancet.com/neurology Vol 15 July 2016

that can help to guide diagnosis and treatment has been at the centre of research efforts in the past decade. A result of these efforts is that the medical assessment of patients with first-time seizures or initial injuries is often comprehensive, and thus detailed clinicopathological data are available. Pitkänen and Engel4 defined a biomarker for epileptogenesis as “an objectively measurable characteristic of a biological process that reliably identifies the development, presence, severity, progression, or localization of an epileptogenic abnormality”. An epileptogenic abnormality is the pathophysiological substrate responsible for the start or continuation of epilepsy, or both. This definition captures the important features of the epilepsies that would benefit from the availability of diagnostic and prognostic biomarkers. An ideal diagnostic biomarker should provide information about clinical status, such as the extent and localisation of the epileptogenic area or the severity of the epilepsy. They might also provide information about sensitivity of clinical symptoms to specific treatments. Prognostic biomarkers, similarly, are of potentially high clinical relevance, but are more difficult to identify than diagnostic markers. They are defined as biomarkers that allow prediction of future clinical features, such as the speed of progression or severity of epilepsy, or the prediction of remission or cure, and could be used to identify features of the development of epilepsy, a process termed epileptogenesis.4–6 This term refers to a process whereby CNS tissue acquires the capability to generate the abnormal and spontaneous electrical activity that underlies seizures. In addition to unprovoked seizures, epilepsy is often associated with somatic, cognitive, psychiatric, and behavioural comorbidities, such as memory impairments.7–9 The development of these comorbidities is an important feature of epileptogenesis. Thus, a further relevant category of biomarkers could be those that can be used to predict the occurrence of comorbidities or sudden

Lancet Neurol 2016; 15: 843–56 See Comment page 782 Department of Neurobiology, A I Virtanen Institute for Molecular Sciences, University of Eastern Finland, Kuopio, Finland (Prof A Pitkänen PhD, Prof O Gröhn PhD); Department of Pharmacology, Toxicology and Pharmacy, University of Veterinary Medicine, Hannover, Germany (Prof W Löscher PhD); Center for Systems Neuroscience, Hannover, Germany (Prof W Löscher); Department of Neuroscience, Experimental Neurology, IRCCS-Istituto di Recerche Farmacologiche “Mario Negri”, Milan, Italy (Prof A Vezzani PhD, T Ravizza PhD); Section for Translational Epilepsy Research, Department of Neuropathology, University of Bonn Medical Center (Prof A J Becker MD) and Laboratory for Experimental Epileptology and Cognition Research, Department of Epileptology (Prof H Beck MD), University of Bonn, Bonn, Germany; Department of Medical Sciences, Section of Pharmacology, University of Ferrara, Ferrara, Italy (Prof M Simonato MD); Unit of Gene Therapy of Neurodegenerative Diseases, Division of Neuroscience, University Vita-Salute San Raffaele, Milan, Italy (Prof M Simonato); The Nencki Institute of Experimental Biology, Polish Academy of Sciences, Warsaw, Poland (Prof K Lukasiuk PhD); Preclinical Molecular Imaging, Department of Nuclear Medicine, Hannover Medical School, Hannover, Germany (J P Bankstahl PhD); Department of Brain and Cognitive Sciences, Zlotowski Center for Neuroscience, Ben-Gurion University of the Negev, Israel (Prof A Friedman PhD); Department of Medical Neuroscience, Dalhousie University, Halifax, NS, Canada (Prof A Friedman); Department of Neuropathology, Academic Medical Center

843

Review

(Prof E Aronica MD), and Swammerdam Institute for Life Sciences, Center for Neuroscience (Prof E Aronica, Prof J A Gorter PhD), University of Amsterdam, Amsterdam, Netherlands; Stichting Epilepsie Instellingen Nederland (SEIN), Heemstede, Netherlands (Prof E Aronica); Department of Clinical and Experimental Epilepsy, UCL Institute of Neurology, London, UK (Prof S M Sisodiya PhD); Epilepsy Society, Chalfont St Peter, Buckinghamshire, UK (Prof S M Sisodiya); Epilepsy Center, Experimental Epilepsy Group, Division of Neurology, Department of Clinical Sciences, Lund University Hospital, Lund, Sweden (Prof M Kokaia PhD); and German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany (Prof H Beck) Correspondence to: Prof Heinz Beck, Laboratory for Experimental Epileptology and Cognition Research, Department of Epileptology, University of Bonn, 53105 Bonn, Germany [email protected] See Online for appendix

unexpected death. In the discovery of prognostic biomarkers, a crucial additional challenge is that these biomarkers are dependent on disease stage. Thus, even in a cross-section of clinically well-defined people with epilepsy, the stage of epileptogenesis (ie, time after the initial precipitating event) is likely to differ. Hence, the expected proportion of the population who are positive for a given biomarker might depend on the timing of sampling relative to disease progression. Perhaps more importantly, as a consequence, the use of combinatorial biomarkers is also likely to vary depending on the stage of the epileptogenic process. Epilepsy specialists might argue that the surrogate endpoints (eg, number of epileptic seizures) often used in clinical trials can be regarded as biomarkers and can be used to predict the effect of the treatment.10 However, in our view, surrogate endpoints do not strictly fall within the definition of biomarker and will not be discussed in this Review. Why is it so important to identify diagnostic and prognostic biomarkers for epilepsy? One of the major objectives in the discovery of diagnostic biomarkers is to individualise and optimise treatment. Over 30 preclinical proof-of-concept studies using animal models of genetic epilepsies, cortical malformations, status epilepticus, or TBI have either shown favourable antiepileptogenic or comorbidity-modifying effects, or both.4 However, none of these experimental studies have led to a clinical trial of an antiepileptogenic drug. Designing appropriately powered clinical trials is not possible even in well-defined groups of patients with epilepsy, because of the heterogeneity of epileptogenesis and the endogenous recovery mechanisms of the brain after injury. As a consequence, adequately powered antiepileptogenic trials will only be possible after the identification of biomarkers that allow stratification of participants on the basis of the predicted risk of epileptogenesis. In this Review, we describe the proof-of-concept preclinical studies and the first clinical studies of

Biomarker type and indication

Tissue

Diagnostic biomarker for epilepsy

biomarker discovery in epilepsy (table). So far, most studies examining the validity of biomarkers, in particular prognostic biomarkers, have relied on animal models with clearly defined initial injury, such as experimental temporal lobe epilepsy (TLE) models caused by an initial status epilepticus, or stroke or TBI models; in human beings, a large portion of the data have been obtained from patients with TLE. Thus, most of the biomarker data available are from a subset of epilepsies, and might not be generalisable across the whole spectrum of the epilepsies.

Genetic biomarkers Genetic mutations can cause epilepsy, make the brain more vulnerable to develop epilepsy after an acquired brain injury, such as ischaemic stroke or TBI, and influence a patient’s response to a given treatment.3,18 Genetic markers have great potential because DNA is readily available from patients at risk admitted to hospital after potentially epileptogenic injuries such as TBI, new-onset status epilepticus, or stroke. Several genetic variants have been linked to the modulation of seizure threshold after injury or to acquired epilepsy (appendix). Some of these genes are involved in GABAergic neurotransmission, such as the GABAA receptor subunit genes GABRB1,19–21 GABRB2,22 GABRB3,23 and GABRG2;24,25 the GAD1 gene;26 and the ALDH5A1 gene.23 Susceptibility to TLE has also been associated with polymorphisms in genes involved in serotonergic transmission, for example HTR1B and SLC6A4.27–29 Association of genetic variants with epilepsy has also been shown for genes involved in the regulation of neuronal excitability (eg, SCN1A,30 PDYN,31 MTHFR,32 or ASIC1α33), genes that encode pro-inflammatory mediators (eg, CD40,34 interleukin 1β,35,36 interleukin 1α,37 and the interleukin 1 receptor antagonist protein37), and genes involved in protection against oxidative stress (eg, ALDH2,38 NFE2L2,39 or PRNP40). A possible involvement

Method

Biomarker

Serum

Molecular analysis

Increased concentration of hsa-miR-106b-5p

Wang et al (2015)12

Diagnostic biomarker for drug-refractory Serum epilepsy

Molecular analysis

Decreased concentration of hsa-miR-301a-3p

Choy et al (2014)13

Prognostic biomarker for epileptogenesis after hyperthermiainduced status epilepticus in rats on postnatal day 11

Brain

MRI

Reduced amygdala T2 relaxation time

Pollard et al (2012)14

Diagnostic biomarker for focal epilepsy

Plasma

Molecular analysis

Increased TARC (CCL17): telencephalin (sICAM5) ratio

Brain

Intrahippocampal or intracortical (entorhinal cortex) EEG; scalp EEG

Occurrence of high-frequency oscillations

MRI

Reduced temporal lobe white matter fractional anisotropy on diffusion tensor imaging

Wang et al (2015)

11

Bragin et al (1999)15 and Diagnostic biomarker for localisation of Andrade-Valenca et al seizure onset zone (ie, tissue (2011)16 epileptogenicity) Labate et al (2015)17

Diagnostic biomarker for drugBrain refractory medial temporal lobe epilepsy

EEG=electroencephalogram. sICAM5=soluble intercellular adhesion molecule 5. TARC=thymus and activation regulated chemokine.

Table: Studies of diagnostic or prognostic biomarkers in epilepsy

844

www.thelancet.com/neurology Vol 15 July 2016

Review

of glial cells is shown by the association between a single nucleotide polymorphism (SNP) in the ATPsensitive KCNJ10 gene, whose protein product is expressed by macroglia,41 and seizure susceptibility,42 as well as the association between risk of TLE and a SNP in KCNJ10 in combination with SNPs in AQP4, a gene encoding a protein crucial for the conductance of water through the cell membrane.43 A SNP in CALHM1 has been suggested to affect TLE susceptibility because it can alter the protein’s ability to control cytosolic calcium concentrations or modulate amyloid β concentrations, or both.44 Finally, polymorphisms in genes related to cell–cell interactions (eg, PCDH7,30 NRG1,45 and BDNF46) and genes related to key components of the innate immune system and inflammatory response that have a role in neuronal plasticity and excitability (eg, C3)47 have been described. However, most of the association studies of common variants that reportedly increase susceptibility to epilepsy have been small, none have been replicated, and they have been limited to pre-selected candidate polymorphisms or genes. No common variants are in clinical use as predictors of the development of epilepsy. In recent years, a few better-powered genome-wide association studies have been done and their findings have revealed possible associations between genetic variants and epilepsy syndromes or populations of patients.48 In a meta-analysis of 8696 cases and 26 157 controls,30 an association was identified between all types of epilepsy and variation in the SCN1A gene, which had previously been associated with epilepsy syndromes that are typically characterised by febrile seizures. However, this finding cannot be used to predict the individual risk of epilepsy. Association data alone are unlikely to be sufficient for prediction of individual risk without consideration of other predictors.

microRNAs MicroRNAs (miRNAs) are small non-coding RNAs that regulate post-transcriptional gene expression.49 They are differentially expressed in the brain under pathological conditions and might therefore represent both therapeutic targets and diagnostic or prognostic biomarkers for neurological diseases, including epilepsy.12,50 Several studies have assessed the potential of miRNAs as biomarkers for different aspects of the epileptogenic process.51–53 Zucchini and colleagues54 examined the expression of human miRNAs in hippocampal granule cells of surgical specimens from patients with intractable TLE and identified 12 miRNAs that were differentially expressed in patients with granule cell dispersion. One of these, hsa-miR-487a, has been verified in an extended cohort of seven patients with intractable TLE and seven controls.54 Whether the miRNA signature of the resected tissue could be used as a biomarker for any aspects of surgical outcome remains to be assessed. www.thelancet.com/neurology Vol 15 July 2016

miRNAs can be found in plasma and serum as well as brain tissue, where they are associated with proteins or are encapsulated in extracellular vesicles. These circulating miRNAs might represent a form of intercellular communication, because they can act as signalling molecules.55,56 miRNA-containing exosomes can be isolated from brain tissue and probably cross the blood– brain barrier, thus providing a possible peripheral blood source of a biomarker that is representative of a pathophysiological CNS process.57 Analysis of circulating miRNAs could be a non-invasive approach that could be used at any stage of the disease process. In two studies,11,12 circulating miRNAs were assessed in people with epilepsy. Wang and colleagues11 measured serum miRNA concentrations in 30 people with epilepsy and 30 controls, and validated the selected miRNAs in a larger cohort of cases. Six miRNAs were differentially expressed in people with epilepsy compared with controls, and receiver operating characteristic (ROC) analysis revealed that hsa-miR-106b-5p had the highest sensitivity and specificity for a diagnosis of epilepsy of the miRNAs assessed. In another study, Wang and colleagues12 measured miRNA concentrations in serum from 77 drug-resistant and 81 drug-responsive people with epilepsy and 85 healthy controls. Five miRNAs were deregulated in drug-resistant patients compared with the drug-responsive and healthy control groups. Based on ROC analysis, hsa-miR-301a-3p had the best diagnostic value for drug-resistant epilepsy (table). As the investigators of these two studies acknowledged,12 prospective studies are needed to confirm the biomarker value of miRNAs in diagnosis of epilepsy and prediction of drug refractoriness. Can measurement of circulating miRNAs be used to identify individuals who will develop epilepsy after brain injury? Attempts have been made to try and answer this question using animal models of status epilepticus.58,59 Roncon and colleagues59 analysed miRNAs using microarrays of plasma samples collected during epileptogenesis after systemic pilocarpine-induced status epilepticus. Concentrations of 27 miRNAs differed between the control and pilocarpine-treated animals. The miRNAs that had altered concentrations before the appearance of the first epileptic seizure, such as rno-miR-9a-3p, are potential biomarker candidates for epileptogenesis. However, the pilocarpine model is not optimum for biomarker studies because almost all animals will eventually develop epilepsy and the progression of epileptogenesis is fast, occurring within 1–2 weeks after status epilepticus, thus it does not accurately represent the human epilepsies. Therefore, further studies are needed to confirm the role of miR-9a-3p as a biomarker for epileptogenesis, both in animal models and in samples from patients. Even though data are promising, a detailed validation of the analysis of miRNAs will be essential to bring these potential biomarkers to clinical practice. The main issues yet to be fully resolved relate to establishment of 845

Review

standardised sampling (eg, elimination of platelet contamination), RNA extraction, and miRNA analysis protocols.60

Structural biomarkers Structural changes are a notable feature of many epilepsies, particularly in patients with chronic TLE with hippocampal sclerosis, which shows characteristic patterns of damage, with segmental neuronal loss, gliosis, and axonal reorganisation. Whether more subtle forms of structural damage can be detected early in the course of disease and predict the future characteristics of epilepsy is unclear. T2-weighted MRI of the amygdala and thalamus has shown promise in the search for biomarkers for epileptogenesis in an experimental rat model of febrile seizures; reduced amygdala T2 relaxation times predicted development of TLE.13 In another imaging study, structural abnormalities in the thalamus and cerebral cortex, as shown by quantitative relaxation and diffusion MRI, were linked to increased seizure susceptibility in a model of post-traumatic epilepsy induced by lateral fluid-percussion injury.61 In the ongoing FEBSTAT study,62 MRI data from children with febrile convulsions are being collected. Over the next decade, we will learn whether pathological changes, some of which will eventually present as hippocampal sclerosis (eg, T2 hyperintensity), are predictive of the development of epilepsy. Other advanced structural imaging parameters have not yet been assessed for their value as biomarkers of epileptogenesis. For example, in animal models, diffusion tensor imaging with high spatial resolution can be used to visualise the progression of structural changes (eg, mossy fibre sprouting in dentate gyrus) in different hippocampal subfields in the months after status epilepticus induced by systemically administered kainic acid or pilocarpine (figures 1A and B). Diffusion tensor imaging and more advanced diffusion MRI techniques, such as high-angular-resolution diffusion imaging (figure 1C), will probably provide even more specific information than conventional diffusion tensor imaging about microstructural changes related to cell death, reactive gliosis, and structural axonal plasticity. Furthermore, microstructural imaging techniques that assess local magnetic susceptibility differences, such as phase imaging and quantitative susceptibility mapping, will aid sensitive detection of microhaemorrhages, calcification, and white matter damage (figure 1D), which might have promise as biomarkers.63

Functional and electrophysiological biomarkers Electrophysiological parameters have been explored as both diagnostic and prognostic biomarkers. In established epilepsy, recurrent seizures constitute the defining disease symptom.64 An epileptic seizure is a transient occurrence of signs or symptoms, or both, caused by abnormal excessive or synchronous neuronal 846

activity in the brain.64,65 In the context of epileptogenesis, the conceptual questions remaining to be answered are how networks transform from a healthy state to generate unprovoked recurrent seizures, and whether early excitability signatures exist that are less conspicuous than seizures. Such signatures could potentially constitute prognostic biomarkers for epileptogenesis or might represent biomarkers of seizure characteristics such as severity or drug response.

Pathological high-frequency oscillations Arguably one of the strongest candidates for a prognostic biomarker of neuronal hyperexcitability are pathological high-frequency oscillations (HFOs).15,66,67 No uniform definition of pathological HFOs exists. However, most of the work on HFOs in both animal models and human studies has focused on brief events in the spectral range of 80–600 Hz, with different terms being used, depending on the frequency range (ie, ripples 80–250 Hz and fast ripples 250–500 Hz). In people with chronic epilepsy, HFOs constitute an important interictal marker of focal epileptogenic zones15 and have been used clinically to localise such areas.16 More importantly, surgical removal of HFO-generating areas led to improvement in functional outcomes.68–71 Thus, HFOs are probably good indicators of the epileptogenic zone. However, few studies have assessed whether occurrence of HFOs is predictive of later development of epilepsy. In the kainate-induced status epilepticus model of epileptogenesis, two different types of HFOs can be measured before the onset of generalised seizures, and their occurrence is predictive of the development of spontaneous seizures.67 Region-specific changes in HFOs have also been reported in the pilocarpine model of epilepsy, but their potential as predictive markers of the severity of the chronic seizure disorder is not clear.72 In the pilocarpine model of epilepsy, and in people with epilepsy, HFOs have been analysed in conjunction with interictal spikes. Findings from these studies have shown a complex temporal and spatial pattern of interictal spikes with and without superimposed HFO patterns.73,74 Together, these findings suggest that HFOs seem to constitute promising candidates for an excitability-based biomarker for epileptogenesis. One of the most notable features of almost all studies on the relation between HFOs and the epileptogenic zone has been that these studies were done using invasive electroencephalography (EEG) techniques. These data have been extremely useful for delineation of the epileptogenic zone. However, to provide broadly applicable, clinically useful predictors of the disease course, non-invasive recording techniques would need to be established. These techniques would need to be capable of measuring the incidence and spectral properties of pathological HFOs, probably at the cost of anatomical resolution. Studies combining scalp EEG, www.thelancet.com/neurology Vol 15 July 2016

Review

invasive EEG, and magnetoencephalography (MEG) will be needed to validate such non-invasive measures. Highsampling-rate MEG systems constitute a promising candidate method to test the validity of HFOs as biomarkers in human epilepsy.75–77 However, any highresolution technology will have to overcome substantial problems. First, the management and analysis of large amounts of data generated by high sampling rates are challenging. Second, criteria for distinguishing HFO signals reliably, and preferentially automatically, from background noise need to be developed and validated.78,79 This task is far from trivial; however, presurgical assessment with the opportunity to identify intracranial local field potential signals and MEG signals in parallel might be valuable for this purpose. Finally, the use of HFO biomarkers in conjunction with behavioural state or other EEG markers such as interictal spikes, or both, might prove useful to optimise the predictive efficacy of HFOs.79 Solving these issues will pave the way for adequately powered clinical studies that can repeatedly and non-invasively assess cohorts of patients at risk to establish the validity of excitability biomarkers. Future work should establish more precisely the spectrum of pathological HFOs in human beings and whether they constitute several classes of excitability signatures rather than a single entity;71 to identify which excitability signatures can be accurately distinguished from normal high-frequency activity; and to develop and validate methods suitable for non-invasive detection of these forms of activity in human beings.

Seizure susceptibility Biomarkers for diagnosis of epileptogenesis would need to identify the presence of lowered seizure threshold without the occurrence of seizures. Seizure threshold, which is thought to determine the propensity or likelihood of seizure occurrence, is decreased after different types of epileptogenic brain injuries in experimental models, including those of genetic mutations, status epilepticus, stroke, or TBI.80–86 Investigators81,83,86 have assessed the seizure threshold in different animal models of TLE after status epilepticus during the course of epileptogenesis. They found some overlap in seizure susceptibility between the animals that developed and those that did not develop epilepsy, which might have been related to variability in the rate of progression of epileptogenesis within the follow-up time. The investigators86 expanded their findings by combining the data on seizure threshold and findings of behavioural alterations during the first 3 weeks after status epilepticus. By generating a combined seizure threshold and behavioural score, they were able to predict epileptogenesis with high sensitivity and specificity (figure 2A).87 Whether these data can be generalised to models of post-traumatic epilepsy and post-stroke epilepsy, and eventually to human beings, remains to be explored. In patients assessed for epilepsy surgery, the area of ictogenic focus has a lower threshold for induced www.thelancet.com/neurology Vol 15 July 2016

Diffusion tensor imaging

A

B

Control

Kainic acid

r l-m DG

CA3b

C

Diffusion tensor imaging

Diffusion tensor imaging and high-angular-resolution diffusion imaging

High-angularresolution diffusion imaging

CC

CC DG DG

CA3b CA3b

D

Phase imaging

Figure 1: Clinically translatable approaches to identify microstructural imaging biomarkers for epileptogenesis (A and B) Diffusion tensor imaging directionally encoded colour fractional anisotropy maps of (A) a control rat and (B) a rat 79 days after kainic-acid-induced status epilepticus. Arrowheads show microstructural changes in hippocampal subfields. The main diffusion directions are colour coded and shown by the arrows at the bottom of (A): red is medial-lateral, green is dorsal-ventral, and blue is rostral-caudal. (C) High-angular-resolution diffusion imaging directionally encoded colour fractional anisotropy map overlaid with visualisations of fibre orientation distribution functions from Q-ball reconstruction of high-angular-resolution diffusion imaging data from the rat hippocampus (left). Squares show pixels in three representative areas: CC, DG, and CA3b, from which enlarged fibre orientation distribution functions are shown (right). A more complex shape is obtained using high-angular-resolution diffusion imaging than diffusion tensor imaging, representing the heterogeneity of cellular structures. (D) Phase imaging gradient echo phase images of rat brain 5 months after traumatic brain injury induced by lateral fluid percussion. The white arrow shows microbleeds in the external capsule, and the green arrow points at a thalamic calcification. CA3b=CA3b subfield of the hippocampus. CC=corpus callosum. DG=dentate gyrus. l-m=stratum lacunosum-moleculare. r=stratum radiatum.

847

Review

electrographic epileptiform or imaging changes after administration of convulsants (eg, pentetrazol or bicuculline) or electrical stimulation of the ictogenic focus, when compared with non-focal areas. These findings provide some support to the idea of using seizure susceptibility testing to diagnose epileptogenesis,89–92 even though electrical stimulation was of little use at identifying focus lateralisation.92

Metabolic and activity biomarkers In addition to electrophysiological measurements of function, several imaging methods can be used to establish the functional state of brain tissue by assessing metabolic parameters non-invasively. The findings from a preclinical longitudinal analysis88 of ¹H-magnetic resonance spectroscopy of the hippocampus of rats after drug-induced status epilepticus showed a progressive increase in myo-inositol and glutathione concentrations before the development of spontaneous seizures. A negative association was found between the concentrations of these metabolites during epileptogenesis and the extent of neurodegeneration in chronically epileptic rats. Moreover, glutathione concentrations during epileptogenesis were inversely associated with the frequency of spontaneous seizures. ROC analyses showed that the concentrations of these two metabolites during epileptogenesis are good discriminators for differentiating control rats from rats that will develop epilepsy (figures 2B and 2C). ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET has been used in two preclinical studies93,94 to assess the association between glucose metabolism and epileptogenesis after TBI or status epilepticus. In both studies, abnormalities in ¹⁸F-FDG PET were identified during epileptogenesis,

A

B

Seizure threshold and behavioural hyperexcitability

but the sensitivity and specificity of the changes remain to be established. Furthermore, radioisotope imaging studies have assessed binding of GABAA and NMDA receptors in experimental and human epilepsy.95,96 However, the potential of receptor labelling imaging as epileptogenesis biomarkers remains to be shown. Finally, large-scale network alterations have gained momentum as a key concept in epilepsy, mainly as a result of advances in non-invasive imaging techniques, which provide maps of functional and structural connectivity in the brain by exploiting resting-state functional MRI and diffusion-MRI-based tractography. Although there is evidence of functional connectivity changes in people with epilepsy, few studies have assessed connectivity changes during epileptogenesis, before the appearance of the first unprovoked seizure. In a preclinical study, resting-state connectivity was measured in rats 4 months after lateral-fluid-percussioninduced TBI and compared with hyperexcitability of the brain induced by pentetrazol in a seizure-susceptibility test. No significant association was found between functional connectivity and seizure susceptibility; however, the number of animals used in this study was small (12 injured vs eight controls).85 Nevertheless, adequately powered network analyses in animal models might produce clinically testable hypotheses in the search for biomarkers.

Biomarkers of neuroinflammation Neuroinflammation is a prominent feature of most types of epilepsies. Neuroinflammation is characterised by the synthesis of cytokines, chemokines, danger signals mediated by endogenous molecules that are released upon tissue damage, and downstream effector molecules

Myo-inositol and creatine concentrations

C

Glutathione and creatine concentrations

True positive rate (sensitivity; %)

100

80

60

40

20 AUC 1·0 p=0·003

AUC 0·96 p=0·004

AUC 0·90 p=0·01

0 0

20 40 60 80 100 False positive rate (100 – specificity; %)

0

20 40 60 80 100 False positive rate (100 – specificity; %)

0

20 40 60 80 False positive rate (100 – specificity; %)

100

Figure 2: Examples of prognostic biomarkers for epileptogenesis using metabolic and excitability measures (A) Combination of change in seizure threshold and behavioural hyperexcitability measured over 3 weeks after pilocarpine-induced status epilepticus in rats. Diagnosis of epilepsy was based on continuous video-electroencephalogram monitoring over the subsequent weeks and months (data from Bröer and Löscher87). ROC analysis using a combined parameter derived from the seizure threshold and behavioural score predicted the development of epilepsy. (B and C) ROC curves using concentrations of (B) myo-inositol and creatine as an indirect measure of astrogliosis and (C) glutathione and creatine as an indirect measure of oxidative stress, measured by ¹H-magnetic resonance spectroscopy in the hippocampus differentiate control rats from rats that will develop epilepsy after pilocarpine-induced status epilepticus (data were re-analysed from Filibian and colleagues88). AUC=area under the ROC curve. ROC=receiver operating characteristic.

848

www.thelancet.com/neurology Vol 15 July 2016

Review

www.thelancet.com/neurology Vol 15 July 2016

Suitable blood biomarkers of neuroinflammation are molecules that are stable in biological fluids and can be rapidly measured in an affordable way. The short half-life of some neuroinflammatory molecules implicated in epilepsy, such as interleukin 1β, has hampered investigations of their biomarker value and produced discordant results.98 Circulating molecules that reflect neuroinflammation can originate by CSF-to-blood diffusion; however, they can also be derived from activated peripheral leucocytes, and thus do not always accurately represent neuroinflammation. Ideally, the inflammatory molecules measurable in blood should be CNS specific and their analysis should not be confounded by non-specific release from peripheral sources (eg, complement factors that are also released by the liver). In a proof-of-principle study,54 plasma concentrations of soluble intercellular adhesion molecule 5 (sICAM5, also A

B

MRI T1 signal

Blood vessel Dysfunctional BBB Intact BBB Slope of increased signal

ΔT1

Time after injection of contrast agent

C

D 100 True positive rate (sensitivity; %)

from microglia, astrocytes, neurons, and the microvessel endothelium, as a result of activation of the innate immune system.97 Clinical evidence has shown that neuroinflammation is a hallmark of the epileptic focus in drug-resistant epilepsies. Moreover, neuroinflammatory mechanisms contribute to seizure generation (ie, ictogenesis) in animal models of epilepsy.98,99 Evidence also exists of rapid onset and persistent neuroinflammation during epileptogenesis in rodents.100 These findings, together with evidence of disease modifications mediated by drugs targeting the inflammatory cascade,101,102 suggest that neuroinflammation plays a part in the development of epilepsy. Even though separation of the changes in inflammatory markers that relate to injury from those that suggest the development of epilepsy will be crucial, neuroinflammation is not only a promising target for novel treatments, but also a potential mechanistic biomarker of epileptogenesis and ictogenesis. Molecular imaging techniques that allow visualisation of inflammatory cells (figure 3A) and measurement of inflammatory mediators in the blood hold promise for monitoring neuroinflammation in people with epilepsy.104 Imaging techniques with high translational potential include PET and MRI.104,105 In a preclinical study, Choy and colleagues13 attempted to validate an imaging biomarker of epileptogenesis in a rat model of prolonged febrile seizures, in which only a subpopulation of animals developed epilepsy. ROC analysis showed that a decrease in amygdaloid and medial thalamic T2 relaxation times measured 2 h after febrile seizures predicted epileptogenesis. The reduction was associated with enhanced oxygen use, an increase in venous blood deoxyhaemoglobin concentration, and release of the neuroinflammatory protein HMGB1 from amygdaloid neurons.98 HMGB1 is bound to nuclei and its release from cells requires a nucleus-to-cytosol translocation, an energy-demanding process that increases oxygen demand. Moreover, the temporal evolution of HMGB1 translocation during epileptogenesis was consistent with the decrease in T2 signal. An early decrease in T2 signal in the amygdala and the dorsal hippocampus was also associated with the development of learning disabilities 2–3 months after febrile seizures, although the sensitivity and specificity of T2 as a biomarker for cognitive decline were not assessed.106 So far, findings from only one proof-of-concept clinical study107 have shown that brain imaging can be used to detect neuroinflammation in people with epilepsy. Binding of TSPO was increased in the region of seizure origin and generalisation in patients with pharmacoresistant focal-onset epilepsy.107 Further studies are needed to assess whether TSPO-PET, other PET ligands, or MRI methods designed to detect neuroinflammation can be used to stratify patients at high risk of developing epilepsy, surgical candidates, or patients who will benefit from anti-inflammatory treatments.88

–0·03

Slope

0·03

80 60 40 20

AUC 0·85 p=0·032

0 0

20 40 60 80 100 False positive rate (100–specificity; %)

Figure 3: Non-invasive detection of inflammation and neurovascular biomarkers (A) Translocator protein PET with the novel ligand ¹⁸F-flutriciclamide can be used for longitudinal assessment of neuroinflammation and has revealed microglia activation (arrows) 7 days after pilocarpine-induced status epilepticus in a rat. (B) Severity of BBB leakage after TBI as a prognostic biomarker for post-traumatic epileptogenesis. Within anatomically defined blood vessels, injection of gadolinium-based contrast agent resulted in a rapid increase in MRI T1 signal due to arterial input followed by a signal decrease (washout; blue line). When signal dynamics within voxels of brain tissue with an intact BBB were analysed, the T1 signal followed the arterial input, but to a smaller magnitude (green line). After TBI, when the BBB became dysfunctional, the leakage of contrast agent resulted in a slow increase in T1 signal (red line). The signal change over time can be measured as the slope of increased signal (dashed line) using a linear regression analysis or by measuring the total change in signal intensity compared with the pre-injection scan (ΔT1). (C) Dynamic contrast-enhanced MRIs from a patient after mild TBI. The slope of the T1 signal change is increased in the injured right temporal lobe. (D) ROC analysis revealed that BBB dysfunction in the injury area was a prognostic biomarker for epileptogenesis after TBI with an AUC of 0·85 (data from Tomkins and colleagues103). AUC=area under the ROC curve. BBB=blood–brain barrier. ROC=receiver operating characteristic. TBI=traumatic brain injury.

849

Review

For SANAD II see http://www. sanad2.org.uk/ For the Human Epilepsy Project see http://humanepilepsyproject. org/

known as telencephalin), which originates in the CNS, was assessed. sICAM5 is expressed specifically by glutamatergic neurons; it plays an active part in synapse formation and dendritic spine plasticity and has antiinflammatory effects when it binds to receptors on microglia and lymphocytes. sICAM5, together with an array of inflammatory mediators, was measured in blood samples from people with drug-resistant focal epilepsy during the interictal phase, disregarding timing of the last seizure, and compared with healthy controls. A five times reduction of sICAM5 was detected in patients, with a concomitant significant increase in concentrations of several pro-inflammatory cytokines, the highest of which was the pro-ictogenic cytokine interleukin 1β.98 Pollard and colleagues14 showed that none of the tested analytes, including interleukin 1, discriminated between people with epilepsy and controls. However, the ROC analysis showed that the plasma ratio between CCL17 (thymus and activation regulated chemokine) and sICAM5 was predictive of seizure activity. Together with TSPO-PET, these data support the idea that markers of neuroinflammatory changes can serve as biomarkers of the epileptogenic process. Prospective studies in large cohorts of well-phenotyped patients with diverse causes of epilepsy are needed to validate the ability of inflammatory and other markers to predict outcomes of the epileptogenic process arising because of genetic abnormalities or brain injuries. Three ongoing studies, FEBSTAT,108 SANAD II, and the Human Epilepsy Project, might soon provide valuable information on this subject.

Microvascular injury biomarkers Another functional system that is interlinked with neuronal function and is altered in many forms of epilepsy is the brain microvasculature. Microvascular injury can mediate delayed and long-lasting changes in the local neurovascular network (figure 3B–D). Specifically, blood– brain barrier dysfunction, and the associated leakage of serum proteins (eg, albumin), that occurs after epileptogenic injuries initiates glial activation and an inflammatory response, which are hallmark pathological abnormalities of the epileptogenic brain. Serum albumin activates specific signalling pathways in astrocytes, leading to a transcriptional response that mediates a reduction in buffering of extracellular potassium and glutamate, an increase in local neuroinflammatory responses, and selective excitatory synaptogenesis.109,110 These astrocytemediated responses lead to hyperexcitability and hyperconnectivity within the neuronal network. Microvasculopathy and damage to the blood–brain barrier are prominent features of many animal models of epilepsy induced by TBI, stroke, or status epilepticus. Findings from animal studies suggest that the presence of microvascular injury in particular brain regions is suggestive of vascular-mediated epileptogenesis. In MRI studies of people with TBI or stroke,111,112 microvascular

850

pathological abnormalities have been consistently noted, and these findings have been confirmed neuropathologically.113 Histological analysis of resected epileptic tissue from patients with drug-resistant epilepsy has also shown evidence of vascular damage and leakage of serum albumin into the neuropil.114–116 Moreover, cerebrovascular dysfunction is associated with comorbidities, such as behavioural and cognitive deficits, many of which occur in people with epilepsy.117 Finally, established and emerging neuroimaging technologies enable quantitative measurement of microvascular functional integrity (including blood–brain barrier permeability, cerebral blood flow, cerebral blood volume, angiogenesis, and cerebrovascular reactivity) in animals and human beings,118,119 and underscores the potential of microvasculopathy-linked biomarkers of epileptogenesis after injury. However, so far, no prospective studies have tested the predictive value of microvascular dysfunction either in animal models or in the clinical setting. In a retrospective study, Tomkins and colleagues103 used contrast-enhanced MRI and quantitative EEG analysis in 37 patients after mild-to-moderate TBI, of whom 19 had post-traumatic epilepsy. The 18 patients without epilepsy had headache, cognitive impairment, mild motor dysphasia, or acute stress reaction. All patients with TBI showed some EEG slowing compared with controls (ie, higher delta and theta, and lower alpha bands), and no differences were noted between patients with and those without posttraumatic epilepsy. By contrast, people with posttraumatic epilepsy were more likely to have a blood–brain barrier lesion than those without post-traumatic epilepsy. Moreover, in most patients with post-traumatic epilepsy, the blood–brain barrier lesion was co-localised with the suspected epileptic region (ROC analysis resulted in an area under the ROC curve of 0·85; p=0·032; figure 3D). The volume of brain tissue with abnormal permeability was about six times higher in patients with than in those without post-traumatic epilepsy, suggesting that sensitivity and specificity would have been higher if a volume threshold had been applied. Although this study confirmed that quantification of vascular permeability is feasible in patients with TBI, it is limited by the small number of patients included and selection bias. Therefore, based on these findings, the conclusion that vascular functions can predict epilepsy, or in other words serve as a biomarker for epileptogenesis, cannot be made. However, findings from several reports120,121 have shown that blood–brain barrier dysfunction is colocalised to the site of the epileptic focus, further supporting the notion that microvasculopathy is likely to represent a relevant aspect of the epileptogenic network.

Current challenges In the past 10 years, promising developments have been made in the biomarker specialty, but key challenges remain in the design of epilepsy biomarker studies. www.thelancet.com/neurology Vol 15 July 2016

Review

Key challenges related to animal models Animal models, in particular those in which an initial injury causes epilepsy, have been used successfully to identify and validate biomarkers in the preclinical setting. However, although stroke and TBI models might be expected to closely match the disease course in human beings, this is not always the case. Models of epilepsy caused by status epilepticus, for example, are phenotypically similar to TLE with hippocampal sclerosis, but to what extent they mimic processes of epileptogenesis in people with TLE without hippocampal sclerosis, or other type of epilepsy, is not clear. Research into biomarkers has mainly focused on the above-mentioned models. Which biomarkers are generally applicable across models, and which are specific to one model, is unclear. A further challenge in models of disease after status epilepticus is the proportion of animals that ultimately develop chronic epilepsy. In rodent models of poststroke epilepsy or post-traumatic epilepsy, only a subpopulation of animals develops epilepsy. However, in status epilepticus models, which are the most commonly used models of epileptogenesis, almost all animals develop epilepsy. In the past decade, efforts to refine status epilepticus models have been made. By using a drug cocktail that irreversibly terminates pilocarpine-induced status epilepticus after 60, 90, or 120 min, Brandt and colleagues86 were able to modify the outcome in such a way that only a subpopulation of animals developed epilepsy. Similar findings have been obtained by terminating chemically or electrically induced status epilepticus with diazepam122,123 or by using young animals.124 Another aspect of model development is that of high-throughput in-vitro models to predict the efficacy of new antiepileptogenic treatments.125

Key challenges related to heterogeneity of the epilepsies Genetic heterogeneity, ethnicity, and sex probably have large effects on biomarker expression. Moreover, that drugs (anti-seizure or otherwise) or comorbidities can affect biomarker profiles cannot be excluded. One further important source of variability is the age of patients. So far, studies have focused on biomarker identification in adults. However, the onset of epilepsy is often in childhood, which will need to be taken into account in epilepsy biomarker studies in the future. Large-scale and more thorough genetic association studies (eg, using next-generation sequencing) will enable identification of mechanisms that are shared across different types of epilepsy and might help predict risk after brain injury. Such studies are underway.30,48

Key challenges related to accessibility of biomarkers Although basic research has led to identification of several potential biomarkers, for translation into clinical practice, these biomarkers need to be accessible to www.thelancet.com/neurology Vol 15 July 2016

physicians. That is, specimens to be analysed need to be accessible; brain tissue is accessible only from a small patient population at a single timepoint, and usually only in patients with severe or chronic disease or both. Biological fluids, including blood and, to a lesser extent, CSF, are more readily available than brain tissue, and can be sampled at several timepoints. Identifying the usefulness of biomarkers measured from accessible specimens will be crucial, as will be the validation of technologies to assess brain function non-invasively (ie, MEG or functional imaging techniques). The cost of biomarker analyses will be a factor in how extensively these biomarkers could be used in future trials and clinical practice.

Key challenges related to identification of patterns An underestimated challenge in the assessment of biomarkers is the development of biomathematical analysis strategies for management of the complex datasets that are increasingly available from people with epilepsy. These datasets are likely to include many of the categories of data mentioned earlier (eg, from imaging studies). There will be a high demand on resources for mathematical modelling and pattern mining, as well as the development of novel analysis methods to define patterns of relevant biomarkers from large datasets.

Conclusions and future directions What, then, is the way forward? Valuable lessons can be learned from other specialties. Probably the specialty most advanced in terms of biomarker discovery and their clinical use is oncology. This specialty has been driven by both technological advances in so-called omics technologies, which enable identification of molecular pathways, and by the wide availability of tumour tissue after resective surgery. As a consequence, molecular platforms for stratification of treatment of people with glioma are already a clinical reality.126 From the lessons learned in oncology, and the key challenges we have identified, we suggest that future research should focus on the following areas.

Animal models Epilepsy is one of the few neurological disorders in which tissue is available from patients for validation of biomarker expression. Moreover, various animal models of epilepsy are available and have led to successful development of anti-seizure treatments. Still, many of the clinically prevalent epilepsy syndromes have no or few incompletely validated animal models (eg, poststroke epilepsy, cortical dysplasia, and paediatric epilepsies). Development of animal models that reproduce the key features of human epilepsy syndromes requires accurate phenotyping of a human epilepsy syndrome. Examples of studies that will provide such information are the large studies on TBI in Europe and the USA (TRACK-TBI and CENTER-TBI).

For TRACK-TBI see http:// tracktbi.ucsf.edu/ For CENTER-TBI see https:// www.center-tbi.eu/

851

Review

Multimodal biomarker panels Findings in oncology have shown the need to combine biomarkers to achieve higher clinical usefulness. First, pathological changes in cancer are diverse, and assessment of a panel of biomarkers of different pathogenic mechanisms progressing in parallel at any specific timepoint is likely to allow a clearer separation of different disease entities than assessment of individual biomarkers. Second, multimodal markers might offer an opportunity for staging of disease progression. Third, multimodal markers might enable better delineation of specific comorbidities (figure 4). Understanding the timing of biomarker expression during the disease process is important for the choice of a biomarker as well as the analysis platform (eg, blood vs imaging). Thus, we suggest that combinatorial molecular approaches will be needed both to be sensitive to a specific type of epilepsy and to stage the evolution of the disease.

Sensitivity of analysis platforms In addition to model refinement, the sensitivity of analysis platforms needs to be improved. Progress in automated detection of various soluble biomarkers in body fluids, simultaneous pattern recognition of molecular

Early epileptogenesis

Late epileptogenesis

combinations, and multiplexed analysis of single cell surface and intracellular macromolecules with nextgeneration flow cytometry techniques will undoubtedly advance biomarker identification, not only in preclinical studies, but also in patient samples. Animal studies provide only small amounts of blood, CSF, saliva, eye secretion fluid, microbiota, or urine samples for analysis, which is a factor to be considered when designing preclinical biomarker studies. High-field clinical MRI (≥3 T) with advanced parallel imaging is already commonly available for both human beings and animals, making translation of findings from preclinical MRI to clinical settings feasible. Technical improvements in gradient hardware and pulse sequences will aid more precise and faster mapping of brain networks.

Statistical approaches The availability of datasets from high-throughput technologies, as well as increasingly complex clinicopathological datasets, provides great potential for the identification of biomarkers. However, substantial biomathematical challenges exist in assessing these complex, multidimensional datasets and identifying combinatorial biomarker signatures. Further development of systems

Epilepsy diagnosis

Recurrent seizures with or without further evolution of epileptogenesis

Magnitude of abnormality

MRI diffusion trace

* Biomarker 6

Biomarker 3

T2 relaxation time Biomarker 7

Biomarker 4 Biomarker 2 Biomarker 8

Biomarker 1

Biomarker 5 Injury Development of epileptogenesis

Hours to days

Weeks

Months

Years

Analysis timepoint after injury

Figure 4: Time dependence of epileptogenesis related to a set of biomarkers Depending on the analysis timepoint after injury (x axis), the list of candidate biomarkers varies. Also, the magnitude of abnormality (y axis) can differ between people who will eventually develop epilepsy and those who will not. For example, at early timepoints, markers are probably released by damaged neurons, glial cells, or vasculature, whereas markers of cellular plasticity will probably become apparent later on. In MRI, the values of diffusion traces and T2 relaxation time also depend on the analysis timepoint (eg, early decrease in MRI diffusion trace followed by a long-lasting increase), and can be affected by recurrent seizures (asterisk). The purple triangle represents a gradual increase in seizure susceptibility before epilepsy diagnosis; purple bars show the fluctuation in number of seizures.

852

www.thelancet.com/neurology Vol 15 July 2016

Review

biology approaches that incorporate data-driven assessment of complex datasets, but also integrate expert knowledge about the relevant underlying biological function, will be necessary.127

Collaborative research approaches Addressing the heterogeneity of epilepsies in human beings and developing and analysing animal models are huge tasks that need collaborative efforts. A start has been made by several large consortia that will be able to implement appropriately powered analyses in many animal models and large patient groups (eg, EPITARGET).

Validation and clinical translation The key final goal of biomarker identification is the validation of candidate biomarkers in larger prospective preclinical and clinical studies across different cohorts and underlying pathological abnormalities.128–132 So far, study cohorts have been small, challenging the statistical power of the analyses, and phenotyping of epilepsy and comorbidities has been scarce. Finally, validation cohorts for candidate biomarkers have been used in only two studies of people with epilepsy.11,12 An additional challenge that will need to be faced is the absence of standardised preclinical and clinical designs for epilepsy biomarker studies. However, novel, multistage, adaptive clinical study designs are being developed for other indications.133 A cornerstone of any validation of biomarkers will be the recruitment of large numbers of patients and careful clinicopathological assessment. This issue has been recognised, and Common Data Elements for harmonisation of epilepsy studies between centres are available for clinical studies. The first Common Data Elements specifically designed for harmonisation of biomarker and proof-of-concept preclinical studies across laboratories are available through EPITARGET. The work ongoing in the International League Against Epilepsy and American Epilepsy Society translational task force for harmonisation of preclinical methods will extend the EPITARGET Common Data Elements to a wider spectrum of models and methods. The coming 10 years will show if these and related efforts will form a basis for the consistent collection of genetic, pathological, functional, and behavioural epilepsy phenotypes that can be used in clinical studies. The identified phenotypes might then, in turn, be used to develop models of epilepsy and identify preclinical biomarkers of the disease.

Conclusions The development of epilepsy biomarkers, especially those capable of reliably predicting clinically significant epileptogenesis, would be of huge value for testing antiepileptogenic treatments. Whatever final common pathways might be shared across epilepsy syndromes in human beings in the clinical manifestation of seizures, www.thelancet.com/neurology Vol 15 July 2016

Search strategy and selection criteria We searched PubMed for articles published between Jan 1, 2011, and Dec 31, 2015, with the terms “biomarker”, “epilepsy”, “seizure”, “seizure threshold”, “epileptogenesis”, “antiepileptogenesis”, “imaging biomarker”, “co-morbidity”, and their combinations. Data from experimental and clinical studies were included. For genetics, we did searches using the terms “polymorphism and epilepsy” and “polymorphism and epilepsy and outcome”. We also identified articles through searches of the authors’ own files. Only articles published in English were reviewed. For EPITARGET see http://www. epitarget.eu

the process of getting to those final common pathways, that is, the processes of epileptogenesis, is likely to be cause and disease specific. As our knowledge, especially from genetics, of the different disorders that cause epilepsy in human beings increases, we will need to consider how biomarkers will best reveal such diversity. Considering the cost and the efforts of developing disease-specific biomarkers for each individual epilepsy syndrome, data from animal models will be crucial in ensuring that properly designed and adequately powered clinical validation studies can be done. Combinatorial platforms would seem most likely to be successful in this context. Although the challenges ahead should not be underestimated, biomarkers would be a major advance in the management of people with epilepsy. Biomarker discovery could be regarded as an advance in the personalisation of medicine, and could allow prevention in the right person at the right time, rather than just symptomatic treatment. The promising proof-of-concept data on putative epilepsy biomarkers summarised in the table will undoubtedly lead to the design of adequately powered studies to maximise the accuracy of the biomarkers assessed. Furthermore, inclusion of biomarker sample collection and analysis into ongoing and planned trials will enable possible prediction of treatment efficacy. These tasks need regulatory standards for biomarker development and use. The final challenge will be to compare the biomarker platforms between the clinical study cohorts and corresponding animal models, which has not been done thoroughly as yet in any of the brain diseases. In this regard, epilepsy research, with its various animal models to recapitulate epilepsy syndromes, provides a unique opportunity to assess preclinical biomarkers in the clinical setting and vice versa.

For Common Data Elements see https://commondataelements. ninds.nih.gov/

Contributors This Review was envisioned and planned as a collaborative activity within the EU consortium EPITARGET, of which all authors are members. The general framework of the article was conceived by AP and HB. Individual sections of the article were drafted by AP, WL, AV, AJB, MS, KL, OG, JPB, AF, EA, JAG, TR, SMS, MK, and HB. A comprehensive manuscript draft was assembled and written by AP and HB. Revision of the manuscript was done by all authors. All authors approved the final version of the Review.

853

Review

Declaration of interests SMS has received grants from UCB. HB has received personal fees from BIAL and UCB. All other authors declare no competing interests.

22

Acknowledgments The authors are primarily supported by the FP7-HEALTH project 602102 (EPITARGET, all authors); the Academy of Finland (AP); Polish Ministry of Science and Education grant W19/7.PR/2014 (KL); Fondazione Monzino and Citizens United for Research in Epilepsy (AV and TR); the Niedersachsen-Research Network on Neuroinfectiology of the Ministry of Science and Culture of Lower Saxony, Germany (WL); the Israel Science Foundation (AF); FP7-PEOPLE-2011-IAPP project 285827 (EPIXCHANGE, MS and MK); and SFB1089 of the Deutsche Forschungsgemeinschaft (AJB and HB).

23

References 1 WHO. Epilepsy fact sheet. http://www.who.int/mediacentre/ factsheets/fs999/en/ (accessed Oct 15, 2015). 2 Perucca E, French J, Bialer M. Development of new antiepileptic drugs: challenges, incentives, and recent advances. Lancet Neurol 2007; 6: 793–804. 3 Pitkänen A, Immonen R. Epilepsy related to traumatic brain injury. Neurotherapeutics 2014; 11: 286–96. 4 Pitkänen A, Engel J Jr. Past and present definitions of epileptogenesis and its biomarkers. Neurotherapeutics 2014; 11: 231–41. 5 Nieh SE, Sherr EH. Epileptic encephalopathies: new genes and new pathways. Neurotherapeutics 2014; 11: 796–806. 6 Schmidt D, Sillanpaa M. Evidence-based review on the natural history of the epilepsies. Curr Opin Neurol 2012; 25: 159–63. 7 Gaitatzis A, Trimble MR, Sander JW. The psychiatric comorbidity of epilepsy. Acta Neurol Scand 2004; 110: 207–20. 8 Gaitatzis A, Carroll K, Majeed A, Sander W. The epidemiology of the comorbidity of epilepsy in the general population. Epilepsia 2004; 45: 1613–22. 9 Kandratavicius L, Lopes-Aguiar C, Bueno-Junior LS, Romcy-Pereira RN, Hallak JE, Leite JP. Psychiatric comorbidities in temporal lobe epilepsy: possible relationships between psychotic disorders and involvement of limbic circuits. Rev Bras Psiquiatr 2012; 34: 454–66. 10 Simonato M, Brooks-Kayal AR, Engel J Jr, et al. The challenge and promise of anti-epileptic therapy development in animal models. Lancet Neurol 2014; 13: 949–60. 11 Wang J, Yu JT, Tan L, et al. Genome-wide circulating microRNA expression profiling indicates biomarkers for epilepsy. Sci Rep 2015; 5: 9522. 12 Wang J, Tan L, Tan L, et al. Circulating microRNAs are promising novel biomarkers for drug-resistant epilepsy. Sci Rep 2015; 5: 10201. 13 Choy M, Dube CM, Patterson K, et al. A novel, noninvasive, predictive epilepsy biomarker with clinical potential. J Neurosci 2014; 34: 8672–84. 14 Pollard JR, Eidelman O, Mueller GP, et al. The TARC/sICAM5 ratio in patient plasma is a candidate biomarker for drug resistant epilepsy. Front Neurol 2012; 3: 181. 15 Bragin A, Engel J Jr, Wilson CL, Fried I, Mathern GW. Hippocampal and entorhinal cortex high-frequency oscillations (100–500 Hz) in human epileptic brain and in kainic acid-treated rats with chronic seizures. Epilepsia 1999; 40: 127–37. 16 Andrade-Valenca LP, Dubeau F, Mari F, Zelmann R, Gotman J. Interictal scalp fast oscillations as a marker of the seizure onset zone. Neurology 2011; 77: 524–31. 17 Labate A, Cherubini A, Tripepi G, et al. White matter abnormalities differentiate severe from benign temporal lobe epilepsy. Epilepsia 2015; 56: 1109–16. 18 Noebels J. Pathway-driven discovery of epilepsy genes. Nat Neurosci 2015; 18: 344–50. 19 Kauffman MA, Levy EM, Consalvo D, Mordoh J, Kochen S. GABABR1 (G1465A) gene variation and temporal lobe epilepsy controversy: new evidence. Seizure 2008; 17: 567–71. 20 Xi B, Chen J, Yang L, Wang W, Fu M, Wang C. GABBR1 gene polymorphism(G1465A) is associated with temporal lobe epilepsy. Epilepsy Res 2011; 96: 58–63. 21 Gambardella A, Manna I, Labate A, et al. GABA(B) receptor 1 polymorphism (G1465A) is associated with temporal lobe epilepsy. Neurology 2003; 60: 560–63.

854

24

25

26

27

28

29

30

31

32

33 34

35

36

37

38

39

40

41

42

43

Wang X, Sun W, Zhu X, et al. Association between the gamma-aminobutyric acid type B receptor 1 and 2 gene polymorphisms and mesial temporal lobe epilepsy in a Han Chinese population. Epilepsy Res 2008; 81: 198–203. Pernhorst K, Raabe A, Niehusmann P, et al. Promoter variants determine gamma-aminobutyric acid homeostasis-related gene transcription in human epileptic hippocampi. J Neuropathol Exp Neurol 2011; 70: 1080–88. Haerian BS, Baum L, Kwan P, et al. Contribution of GABRG2 polymorphisms to risk of epilepsy and febrile seizure: a multicenter cohort study and meta-analysis. Mol Neurobiol 2015; published online Oct 9. DOI:10.1007/s12035-015-9457-y. Balan S, Sathyan S, Radha SK, Joseph V, Radhakrishnan K, Banerjee M. GABRG2, rs211037 is associated with epilepsy susceptibility, but not with antiepileptic drug resistance and febrile seizures. Pharmacogenet Genomics 2013; 23: 605–10. Darrah SD, Miller MA, Ren D, et al. Genetic variability in glutamic acid decarboxylase genes: associations with post-traumatic seizures after severe TBI. Epilepsy Res 2013; 103: 180–94. Stefulj J, Bordukalo-Niksic T, Hecimovic H, Demarin V, Jernej B. Epilepsy and serotonin (5HT): variations of 5HT-related genes in temporal lobe epilepsy. Neurosci Lett 2010; 478: 29–31. Manna I, Labate A, Gambardella A, et al. Serotonin transporter gene (5-Htt): association analysis with temporal lobe epilepsy. Neurosci Lett 2007; 421: 52–56. Li J, Lin H, Zhu X, et al. Association study of functional polymorphisms in serotonin transporter gene with temporal lobe epilepsy in Han Chinese population. Eur J Neurol 2012; 19: 351–53. International League Against Epilepsy Consortium on Complex Epilepsies. Genetic determinants of common epilepsies: a meta-analysis of genome-wide association studies. Lancet Neurol 2014; 13: 893–903. Stogmann E, Zimprich A, Baumgartner C, Aull-Watschinger S, Hollt V, Zimprich F. A functional polymorphism in the prodynorphin gene promotor is associated with temporal lobe epilepsy. Ann Neurol 2002; 51: 260–63. Scher AI, Wu H, Tsao JW, et al. MTHFR C677T genotype as a risk factor for epilepsy including post-traumatic epilepsy in a representative military cohort. J Neurotrauma 2011; 28: 1739–45. Lv RJ, He JS, Fu YH, et al. ASIC1a polymorphism is associated with temporal lobe epilepsy. Epilepsy Res 2011; 96: 74–80. Zhang B, Chen M, Yang H, Wu T, Song C, Guo R. Evidence for involvement of the CD40/CD40L system in post-stroke epilepsy. Neurosci Lett 2014; 567: 6–10. Kanemoto K, Kawasaki J, Yuasa S, et al. Increased frequency of interleukin-1beta-511T allele in patients with temporal lobe epilepsy, hippocampal sclerosis, and prolonged febrile convulsion. Epilepsia 2003; 44: 796–99. Kanemoto K, Kawasaki J, Miyamoto T, Obayashi H, Nishimura M. Interleukin (IL)1beta, IL-1alpha, and IL-1 receptor antagonist gene polymorphisms in patients with temporal lobe epilepsy. Ann Neurol 2000; 47: 571–74. Salzmann A, Perroud N, Crespel A, Lambercy C, Malafosse A. Candidate genes for temporal lobe epilepsy: a replication study. Neurol Sci 2008; 29: 397–403. Yang H, Song Z, Yang GP, et al. The ALDH2 rs671 polymorphism affects post-stroke epilepsy susceptibility and plasma 4-HNE levels. PLoS One 2014; 9: e109634. Liu Z, Yin X, Liu L, et al. Association of KEAP1 and NFE2L2 polymorphisms with temporal lobe epilepsy and drug resistant epilepsy. Gene 2015; 571: 231–36. Labate A, Manna I, Gambardella A, et al. Association between the M129V variant allele of PRNP gene and mild temporal lobe epilepsy in women. Neurosci Lett 2007; 421: 1–4. Nwaobi SE, Cuddapah VA, Patterson KC, Randolph AC, Olsen ML. The role of glial-specific Kir4.1 in normal and pathological states of the CNS. Acta Neuropathol 2016; published online March 9. DOI:10.1007/s00401-016-1553-1. Buono RJ, Lohoff FW, Sander T, et al. Association between variation in the human KCNJ10 potassium ion channel gene and seizure susceptibility. Epilepsy Res 2004; 58: 175–83. Heuser K, Nagelhus EA, Tauboll E, et al. Variants of the genes encoding AQP4 and Kir4.1 are associated with subgroups of patients with temporal lobe epilepsy. Epilepsy Res 2010; 88: 55–64.

www.thelancet.com/neurology Vol 15 July 2016

Review

44 45

46

47

48

49 50 51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66 67

Lv RJ, He JS, Fu YH, et al. A polymorphism in CALHM1 is associated with temporal lobe epilepsy. Epilepsy Behav 2011; 20: 681–85. Zhu WY, Jiang P, He X, et al. Contribution of NRG1 gene polymorphisms in temporal lobe epilepsy. J Child Neurol 2016; 31: 271–76. Shen N, Zhu X, Lin H, et al. Role of BDNF Val66Met functional polymorphism in temporal lobe epilepsy. Int J Neurosci 2016; 126: 436–41. Jamali S, Salzmann A, Perroud N, et al. Functional variant in complement C3 gene promoter and genetic susceptibility to temporal lobe epilepsy and febrile seizures. PLoS One 2010; 5: e12740. Leu C, Coppola A, Sisodiya SM. Progress from genome-wide association studies and copy number variant studies in epilepsy. Curr Opin Neurol 2016; 29: 158–67. Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 2004; 116: 281–97. Jimenez-Mateos EM, Henshall DC. Epilepsy and microRNA. Neuroscience 2013; 238: 218–29. Aronica E, Fluiter K, Iyer A, et al. Expression pattern of miR-146a, an inflammation-associated microRNA, in experimental and human temporal lobe epilepsy. Eur J Neurosci 2010; 31: 1100–07. Kaalund SS, Veno MT, Bak M, et al. Aberrant expression of miR-218 and miR-204 in human mesial temporal lobe epilepsy and hippocampal sclerosis-convergence on axonal guidance. Epilepsia 2014; 55: 2017–27. Miller-Delaney SF, Bryan K, Das S, et al. Differential DNA methylation profiles of coding and non-coding genes define hippocampal sclerosis in human temporal lobe epilepsy. Brain 2015; 138: 616–31. Zucchini S, Marucci G, Paradiso B, et al. Identification of miRNAs differentially expressed in human epilepsy with or without granule cell pathology. PLoS One 2014; 9: e105521. Alvarez-Erviti L, Seow Y, Yin H, Betts C, Lakhal S, Wood MJ. Delivery of siRNA to the mouse brain by systemic injection of targeted exosomes. Nat Biotechnol 2011; 29: 341–45. Haqqani AS, Delaney CE, Tremblay TL, Sodja C, Sandhu JK, Stanimirovic DB. Method for isolation and molecular characterization of extracellular microvesicles released from brain endothelial cells. Fluids Barriers CNS 2013; 10: 4. Jin XF, Wu N, Wang L, Li J. Circulating microRNAs: a novel class of potential biomarkers for diagnosing and prognosing central nervous system diseases. Cell Mol Neurobiol 2013; 33: 601–13. Gorter JA, Iyer A, White I, et al. Hippocampal subregion-specific microRNA expression during epileptogenesis in experimental temporal lobe epilepsy. Neurobiol Dis 2014; 62: 508–20. Roncon P, Soukupova M, Binaschi A, et al. MicroRNA profiles in hippocampal granule cells and plasma of rats with pilocarpine-induced epilepsy—comparison with human epileptic samples. Sci Rep 2015; 5: 14143. McDonald JS, Milosevic D, Reddi HV, Grebe SK, Algeciras-Schimnich A. Analysis of circulating microRNA: preanalytical and analytical challenges. Clin Chem 2011; 57: 833–40. Immonen R, Kharatishvili I, Gröhn O, Pitkänen A. MRI biomarkers for post-traumatic epileptogenesis. J Neurotrauma 2013; 30: 1305–09. Lewis DV, Shinnar S, Hesdorffer DC, et al. Hippocampal sclerosis after febrile status epilepticus: the FEBSTAT study. Ann Neurol 2014; 75: 178–85. Lehto LJ, Sierra A, Corum CA, et al. Detection of calcifications in vivo and ex vivo after brain injury in rat using SWIFT. Neuroimage 2012; 61: 761–72. Fisher RS, van Emde BW, Blume W, et al. Epileptic seizures and epilepsy: definitions proposed by the International League Against Epilepsy (ILAE) and the International Bureau for Epilepsy (IBE). Epilepsia 2005; 46: 470–72. Fisher RS, Acevedo C, Arzimanoglou A, et al. ILAE official report: a practical clinical definition of epilepsy. Epilepsia 2014; 55: 475–82. Jacobs J, Staba R, Asano E, et al. High-frequency oscillations (HFOs) in clinical epilepsy. Prog Neurobiol 2012; 98: 302–15. Bragin A, Wilson CL, Almajano J, Mody I, Engel J Jr. High-frequency oscillations after status epilepticus: epileptogenesis and seizure genesis. Epilepsia 2004; 45: 1017–23.

www.thelancet.com/neurology Vol 15 July 2016

68

69

70

71

72

73

74

75

76

77

78

79

80 81

82

83

84

85

86

87

88

89

Haegelen C, Perucca P, Chatillon CE, et al. High-frequency oscillations, extent of surgical resection, and surgical outcome in drug-resistant focal epilepsy. Epilepsia 2013; 54: 848–57. Weiss SA, Banks GP, McKhann GM Jr, et al. Ictal high frequency oscillations distinguish two types of seizure territories in humans. Brain 2013; 136: 3796–808. Akiyama T, McCoy B, Go CY, et al. Focal resection of fast ripples on extraoperative intracranial EEG improves seizure outcome in pediatric epilepsy. Epilepsia 2011; 52: 1802–11. Kerber K, Dumpelmann M, Schelter B, et al. Differentiation of specific ripple patterns helps to identify epileptogenic areas for surgical procedures. Clin Neurophysiol 2014; 125: 1339–45. Salami P, Levesque M, Benini R, Behr C, Gotman J, Avoli M. Dynamics of interictal spikes and highfrequency oscillations during epileptogenesis in temporal lobe epilepsy. Neurobiol Dis 2014; 67: 97–106. Levesque M, Salami P, Gotman J, Avoli M. Two seizure-onset types reveal specific patterns of high-frequency oscillations in a model of temporal lobe epilepsy. J Neurosci 2012; 32: 13264–72. Jacobs J, Vogt C, LeVan P, Zelmann R, Gotman J, Kobayashi K. The identification of distinct high-frequency oscillations during spikes delineates the seizure onset zone better than high-frequency spectral power changes. Clin Neurophysiol 2016; 127: 129–42. Xiang J, Tenney JR, Korman AM, et al. Quantification of interictal neuromagnetic activity in absence epilepsy with accumulated source imaging. Brain Topogr 2015; 28: 904–14. Tenney JR, Fujiwara H, Horn PS, et al. Low- and high-frequency oscillations reveal distinct absence seizure networks. Ann Neurol 2014; 76: 558–67. Xiang J, Luo Q, Kotecha R, et al. Accumulated source imaging of brain activity with both low and high-frequency neuromagnetic signals. Front Neuroinform 2014; 8: 57. Amiri M, Lina JM, Pizzo F, Gotman J. High frequency oscillations and spikes: separating real HFOs from false oscillations. Clin Neurophysiol 2016; 127: 187–96. Dümpelmann M, Jacobs J, Schulze-Bonhage A. Temporal and spatial characteristics of high frequency oscillations as a new biomarker in epilepsy. Epilepsia 2015; 56: 197–206. Kharatishvili I, Pitkänen A. Posttraumatic epilepsy. Curr Opin Neurol 2010; 23: 183–88. Rattka M, Brandt C, Bankstahl M, Bröer S, Löscher W. Enhanced susceptibility to the GABA antagonist pentylenetetrazole during the latent period following a pilocarpine-induced status epilepticus in rats. Neuropharmacology 2011; 60: 505–12. Bolkvadze T, Pitkänen A. Development of post-traumatic epilepsy after controlled cortical impact and lateral fluid-percussion-induced brain injury in the mouse. J Neurotrauma 2012; 29: 789–812. Rattka M, Brandt C, Löscher W. Do proconvulsants modify or halt epileptogenesis? Pentylenetetrazole is ineffective in two rat models of temporal lobe epilepsy. Eur J Neurosci 2012; 36: 2505–20. Brima T, Otahal J, Mares P. Increased susceptibility to pentetrazol-induced seizures in developing rats after cortical photothrombotic ischemic stroke at P7. Brain Res 2013; 1507: 146–53. Mishra AM, Bai X, Sanganahalli BG, et al. Decreased resting functional connectivity after traumatic brain injury in the rat. PLoS One 2014; 9: e95280. Brandt C, Tollner K, Klee R, Bröer S, Löscher W. Effective termination of status epilepticus by rational polypharmacy in the lithium-pilocarpine model in rats: window of opportunity to prevent epilepsy and prediction of epilepsy by biomarkers. Neurobiol Dis 2015; 75: 78–90. Bröer S, Löscher W. Novel combinations of phenotypic biomarkers predict development of epilepsy in the lithium-pilocarpine model of temporal lobe epilepsy in rats. Epilepsy Behav 2015; 53: 98–107. Filibian M, Frasca A, Maggioni D, Micotti E, Vezzani A, Ravizza T. In vivo imaging of glia activation using ¹H-magnetic resonance spectroscopy to detect putative biomarkers of tissue epileptogenicity. Epilepsia 2012; 53: 1907–16. Wieser HG, Bancaud J, Talairach J, Bonis A, Szikla G. Comparative value of spontaneous and chemically and electrically induced seizures in establishing the lateralization of temporal lobe seizures. Epilepsia 1979; 20: 47–59.

855

Review

90

91

92

93

94

95

96

97

98

99 100 101 102 103

104 105 106

107

108

109

110

111

856

Barba C, Barbati G, Di Giuda D, et al. Diagnostic yield and predictive value of provoked ictal SPECT in drug-resistant epilepsies. J Neurol 2012; 259: 1613–22. Barba C, Di GD, Fuggetta F, Colicchio G. Provoked ictal SPECT in temporal and extratemporal drug-resistant epileptic patients: comparison of statistical parametric mapping and qualitative analysis. Epilepsy Res 2009; 84: 6–14. So N, Gloor P, Quesney LF, Jones-Gotman M, Olivier A, Andermann F. Depth electrode investigations in patients with bitemporal epileptiform abnormalities. Ann Neurol 1989; 25: 423–31. Guo Y, Gao F, Wang S, et al. In vivo mapping of temporospatial changes in glucose utilization in rat brain during epileptogenesis: an ¹⁸F-fluorodeoxyglucose-small animal positron emission tomography study. Neuroscience 2009; 162: 972–79. Shultz SR, Cardamone L, Liu YR, et al. Can structural or functional changes following traumatic brain injury in the rat predict epileptic outcome? Epilepsia 2013; 54: 1240–50. McGinnity CJ, Koepp MJ, Hammers A, et al. NMDA receptor binding in focal epilepsies. J Neurol Neurosurg Psychiatry 2015; 86: 1150–57. Syvanen S, Labots M, Tagawa Y, et al. Altered GABAA receptor density and unaltered blood–brain barrier transport in a kainate model of epilepsy: an in vivo study using ¹¹C-flumazenil and PET. J Nucl Med 2012; 53: 1974–83. Filiou MD, Arefin AS, Moscato P, Graeber MB. ‘Neuroinflammation’ differs categorically from inflammation: transcriptomes of Alzheimer’s disease, Parkinson’s disease, schizophrenia and inflammatory diseases compared. Neurogenetics 2014; 15: 201–12. Vezzani A, Maroso M, Balosso S, Sanchez MA, Bartfai T. IL-1 receptor/Toll-like receptor signaling in infection, inflammation, stress and neurodegeneration couples hyperexcitability and seizures. Brain Behav Immun 2011; 25: 1281–89. Vezzani A, French J, Bartfai T, Baram TZ. The role of inflammation in epilepsy. Nat Rev Neurol 2011; 7: 31–40. Vezzani A, Aronica E, Mazarati A, Pittman Q J. Epilepsy and brain inflammation. Exp Neurol 2013; 244: 11–21. Ravizza T, Balosso S, Vezzani A. Inflammation and prevention of epileptogenesis. Neurosci Lett 2011; 497: 223–30. Vezzani A, Fujinami RS, White HS, et al. Infections, inflammation and epilepsy. Acta Neuropathol 2016; 131: 211–34. Tomkins O, Feintuch A, Benifla M, Cohen A, Friedman A, Shelef I. Blood–brain barrier breakdown following traumatic brain injury: a possible role in posttraumatic epilepsy. Cardiovasc Psychiatry Neurol 2011; 2011: 765923. Pulli B, Chen JW. Imaging neuroinflammation—from bench to bedside. J Clin Cell Immunol 2014; 5: 226. Amhaoul H, Staelens S, Dedeurwaerdere S. Imaging brain inflammation in epilepsy. Neuroscience 2014; 279: 238–52. Barry JM, Choy M, Dube C, et al. T2 relaxation time post febrile status epilepticus predicts cognitive outcome. Exp Neurol 2015; 269: 242–52. Gershen LD, Zanotti-Fregonara P, Dustin IH, et al. Neuroinflammation in temporal lobe epilepsy measured using positron emission tomographic imaging of translocator protein. JAMA Neurol 2015; 72: 882–88. Shinnar S, Hesdorffer DC, Nordli DR Jr, et al. Phenomenology of prolonged febrile seizures: results of the FEBSTAT study. Neurology 2008; 71: 170–76. Cacheaux LP, Ivens S, David Y, et al. Transcriptome profiling reveals TGF-beta signaling involvement in epileptogenesis. J Neurosci 2009; 29: 8927–35. Weissberg I, Wood L, Kamintsky L, et al. Albumin induces excitatory synaptogenesis through astrocytic TGF-beta/ALK5 signaling in a model of acquired epilepsy following blood–brain barrier dysfunction. Neurobiol Dis 2015; 78: 115–25. Shlosberg D, Benifla M, Kaufer D, Friedman A. Blood–brain barrier breakdown as a therapeutic target in traumatic brain injury. Nat Rev Neurol 2010; 6: 393–403.

112 Pitkänen A, Roivainen R, Lukasiuk K. Development of epilepsy after ischaemic stroke. Lancet Neurol 2016; 15: 185–97. 113 Rodriguez-Baeza A, Reina-de la TF, Poca A, Marti M, Garnacho A. Morphological features in human cortical brain microvessels after head injury: a three-dimensional and immunocytochemical study. Anat Rec A Discov Mol Cell Evol Biol 2003; 273: 583–93. 114 Rigau V, Morin M, Rousset MC, et al. Angiogenesis is associated with blood–brain barrier permeability in temporal lobe epilepsy. Brain 2007; 130: 1942–56. 115 van Vliet EA, da Costa Araújo S, Redeker S, van Schaik R, Aronica E, Gorter JA. Blood–brain barrier leakage may lead to progression of temporal lobe epilepsy. Brain 2007; 130: 521–34. 116 Raabe A, Schmitz AK, Pernhorst K, et al. Cliniconeuropathologic correlations show astroglial albumin storage as a common factor in epileptogenic vascular lesions. Epilepsia 2012; 53: 539–48. 117 Wei EP, Hamm RJ, Baranova AI, Povlishock JT. The long-term microvascular and behavioural consequences of experimental traumatic brain injury after hypothermic intervention. J Neurotrauma 2009; 26: 527–37. 118 Ding G, Jiang Q, Li L, et al. Detection of BBB disruption and hemorrhage by Gd-DTPA enhanced MRI after embolic stroke in rat. Brain Res 2006; 1114: 195–203. 119 Veksler R, Shelef I, Friedman A. Blood–brain barrier imaging in human neuropathologies. Arch Med Res 2014; 45: 646–52. 120 Tomkins O, Shelef I, Kaizerman I, et al. Blood–brain barrier disruption in post-traumatic epilepsy. J Neurol Neurosurg Psychiatry 2008; 79: 774–77. 121 Pavlovsky L, Seiffert E, Heinemann U, Korn A, Golan H, Friedman A. Persistent BBB disruption may underlie alpha interferon-induced seizures. J Neurol 2005; 252: 42–46. 122 Pitkänen A, Kharatishvili I, Narkilahti S, Lukasiuk K, Nissinen J. Administration of diazepam during status epilepticus reduces development and severity of epilepsy in rat. Epilepsy Res 2005; 63: 27–42. 123 Shrot S, Ramaty E, Biala Y, et al. Prevention of organophosphateinduced chronic epilepsy by early benzodiazepine treatment. Toxicology 2014; 323: 19–25. 124 Ravizza T, Gagliardi B, Noe F, Boer K, Aronica E, Vezzani A. Innate and adaptive immunity during epileptogenesis and spontaneous seizures: evidence from experimental models and human temporal lobe epilepsy. Neurobiol Dis 2008; 29: 142–60. 125 Dyhrfjeld-Johnsen J, Berdichevsky Y, Swiercz W, Sabolek H, Staley KJ. Interictal spikes precede ictal discharges in an organotypic hippocampal slice culture model of epileptogenesis. J Clin Neurophysiol 2010; 27: 418–24. 126 Erdmann J. All aboard: will molecular tumor boards help cancer patients? Nat Med 2015; 21: 655–56. 127 McDermott JE, Wang J, Mitchell H, et al. Challenges in biomarker discovery: combining expert insights with statistical analysis of complex omics data. Expert Opin Med Diagn 2013; 7: 37–51. 128 Atluri G, Padmanabhan K, Fang G, et al. Complex biomarker discovery in neuroimaging data: finding a needle in a haystack. Neuroimage Clin 2013; 3: 123–31. 129 Humpel C. Identifying and validating biomarkers for Alzheimer’s disease. Trends Biotechnol 2011; 29: 26–32. 130 Rosen C, Hansson O, Blennow K, Zetterberg H. Fluid biomarkers in Alzheimer’s disease—current concepts. Mol Neurodegener 2013; 8: 20. 131 Di Battista AP, Buonora JE, Rhind SG, et al. Blood biomarkers in moderate-to-severe traumatic brain injury: potential utility of a multi-marker approach in characterizing outcome. Front Neurol 2015; 6: 110. 132 Milward EA, Moscato P, Riveros C, Johnstone DM. Beyond statistics: a new combinatorial approach to identifying biomarker panels for the early detection and diagnosis of Alzheimer’s disease. J Alzheimers Dis 2014; 39: 211–17. 133 Gao Z, Roy A, Tan M. Multistage adaptive biomarker-directed targeted design for randomized clinical trials. Contemp Clin Trials 2015; 42: 119–31.

www.thelancet.com/neurology Vol 15 July 2016