An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Noebels JL, Avoli M, Rogawski MA, et al., editors. Jasper's Basic Mechanisms of the Epilepsies. 5th edition. New York: Oxford University Press; 2024. doi: 10.1093/med/9780197549469.003.0039
A biomarker is a measurable characteristic or indicator of normal biologic or pathogenic processes. Biomarkers are needed to (1) identify individuals who have the highest risk of developing epilepsy after an epileptogenic insult (susceptibility/risk biomarkers), (2) improve early diagnosis, possibly even during the presymptomatic latency period following an insult (diagnostic biomarkers), (3) characterize the disease course (prognostic biomarkers), and (4) assess disease severity or response to treatment (predictive biomarkers/surrogate endpoints). Attempts to harmonize the methodologies and procedures of biomarker data collection and analysis will increase the statistical power of preclinical and clinical studies. A strategic roadmap is needed for proposing research priorities in biomarker discovery for epileptogenesis, regulatory issues, and optimization of the use of resources, similar to those devised for cancer and Alzheimer disease research. This chapter will discuss potential genetic and molecular imaging biomarkers, with examples for each, emphasizing susceptibility/risk, diagnosis, prognosis, and predictive markers. In the future, pathophysiological knowledge on epileptogenesis will need to be translated from experimental and animal studies into clinical research to provide more insights into the development and progression of symptomatic epilepsy.
In theory, epilepsies caused by insults such as traumatic brain injury (TBI) or stroke could be prevented, or the disease course modified, once preventive or disease-modifying treatments become available. Only a small subset of people will develop epilepsy after stroke or TBI, complicating treatment development (Galovic et al., 2021). Even after severe penetrating head trauma, it can take 20 years for seizures to appear (Annegers et al., 1998). Clinical trials of antiepileptogenic agents may be impossible, unless individuals at highest risk of developing epilepsy can be identified (Galovic et al., 2018), or the epileptogenic process can be quantified in other ways. Alternative metrics are very important for daily clinical management of people with epilepsy, which rely primarily on subjective seizure self-reporting (Cook et al., 2013). With more personalized treatments on the horizon and ambitions to treat upstream pathophysiological mechanisms, we are lacking biomarkers that can inform prognosis and response to treatment, and so can support claims of disease modification. We urgently need surrogate markers that help us go beyond the mainly convulsive seizure count in patient with established epilepsy.
A biomarker is a measurable characteristic or indicator of normal biologic or pathogenic processes. The FDA-NIH Biomarker Working Group (https://www.ncbi.nlm.nih.gov/books/NBK326791/) defined specific biomarker categories (Table 39–1). The distinctions among biomarker categories help to clarify stages of disease pathogenesis and progression.
Biomarker Definitions.
A susceptibility/risk biomarker would assess the chance of developing epilepsy after a specific event like encephalitis, but a diagnostic biomarker would not, because the individual would still be asymptomatic. Diagnostic biomarkers may determine whether a patient has epilepsy and should be treated or enrolled in a clinical trial studying a particular syndrome.
Overlap occurs. An encephalogram (EEG) may be a susceptibility biomarker after infection, as well as a diagnostic marker in a person with suspected seizures. A few EEG patterns, such as 3 Hz generalized spike-waves, are considered diagnostic of epilepsy even if no clinical events have been observed. Susceptibility biomarkers can help plan preventive strategies and intervention trials after a potentially epileptogenic event.
A prognostic biomarker may help identify how likely it is that epilepsy will progress, remain static, or remit. It might, for example, indicate the chance of further unprovoked seizures after a first provoked seizure, helping to guide clinicians’ monitoring and treatment approaches. They may inform clinical trial eligibility, facilitating subject selection to increase statistical power. Predictive biomarkers identify the chance of response to an intervention. A predictive biomarker can also facilitate trial enrichment by selecting patients most likely to respond. Some markers may serve all three roles: 3 Hz spike-wave discharges suggest untreated seizure recurrence prognosis, predict response to ethosuximide, and measure the drug’s effect (Covanis et al., 1992).
Each biomarker suffers from specific issues based on the data used to validate it. Interpretation of a diagnostic biomarker, for example, could be affected by measurement error and inexact knowledge of normal ranges; no biomarker has perfect clinical sensitivity and specificity. Desirable characteristics of biomarkers include the following ones: objectively measured and evaluated, reproducible, reasonably cheap to obtain, appropriate for children, quantifiable, with high accuracy, sensitivity, and specificity. Optimally, it would reflect the effect of antiepileptogenic intervention. Overreliance on biomarkers in clinical practice may lead to unneeded medical or surgical treatment.
This chapter will discuss potential genetic and imaging biomarkers, with examples for each, emphasizing susceptibility/risk, diagnosis, prognostic, and predictive markers. EEG and blood biomarkers are discussed in the following chapters.
A genetic marker that serves as a biomarker typically presents as a genetic variant like a single-nucleotide polymorphism (SNP) or a copy number variant (CNV). Genetic variation may contribute to the epileptic phenotype and to seizure development after acquired brain injuries. Multiple genes may influence epileptogenesis; their interaction with brain lesions is a major research topic (see Table 39–2).
Summary of Potential Evidence for Genetic Biomarkers Associated with Epilepsy in Humans.
A vast amount of preclinical studies have shown the influence of genetic background on epilepsy outcome (Schauwecker, 2012; Loscher et al., 2017; Bertoglio et al., 2017), but dedicated studies investigating genetic effects on epileptogenesis are sparse. The influence of risk variants on disease development and severity is typically studied in genetically modified mice or animals with different genetic backgrounds. Mutations affecting the brain inhibitory–excitatory balance have been linked with epileptogenesis following TBI.
As an example, mice with a CD1 background have increased risk for epileptogenesis and a greater incidence of epilepsy after controlled cortical impact TBI as compared with C57BL/6J mice (Zhu et al., 2019; Kang et al., 2018). Modifications in individual gene expression on post-TBI seizures and epilepsy showed an effect on acute postimpact status epilepticus, late (N7 d post injury) seizure susceptibility, and epileptogenesis.
Genetic studies have evaluated the contribution of a risk variant to epileptogenesis and epilepsy severity. Common polygenic variant burden for epilepsy, measured by polygenic risk score (PRS), is distributed differently among patients with epilepsy and controls and estimates the individual’s overall genetic susceptibility to develop epilepsy. In addition, a combination of rare and common variants that may predispose to epilepsy provides a more informative risk prediction (Leu et al., 2019). Common variants are associated with febrile seizures, and specifically with measles, mumps, and rubella (MMR) vaccine-related febrile seizures (Feenstra et al., 2014). Several SNPs were associated with development of temporal lobe epilepsy (TLE) independent of underlying hippocampal sclerosis, but none were replicated in larger studies (Sueri et al., 2018).
Genes influencing inflammation pathways might be involved in epileptogenesis, such as the ones encoding for sestrin 3 (SESN3) (Johnson et al., 2015), interleukin-1 (IL1) (Diamond et al., 2015), kelch-like ECH-associated protein 1 (KEAP1), and nuclear erythroid 2- related factor 2 (NERF2) (Liu et al., 2015).
The incidence of PTE after TBI was not significantly associated with APOE variants; only about half of individuals with APOE E4/E4 genotypes developed posttraumatic seizures (PTSs) after severe TBI, suggesting that the E4/E4 genotype is not a reliable risk biomarker for PTE (Miller et al., 2010). In a military cohort study, methylenetetrahydrofolate reductase (MTHFR) C677T variant was associated with an increased probability of developing epilepsy after TBI (Scher et al., 2011). Variants in glutamic acid decarboxylase 1 (GAD1), an enzyme synthesizing the inhibitory neurotransmitter GABA, may increase PTS risk, in particular the three genotypes rs3828275, rs3791878, and rs769391 (Darrah et al., 2013). Glutamatergic neurotransmission also may affect outcome after TBI. Glutamate transporter gene SLC1A1 and SLC1A3 mutations were reported to increase risk for epilepsy after TBI (Kumar et al., 2019; Ritter et al., 2016). Variants in ADORA1, encoding adenosine receptor A1, which promotes the neuroprotection after brain injury, were linked with increased risk for PTE (Wagner et al., 2010). Specific variants of the ADK gene encoding adenosine kinase, an enzyme that phosphorylates adenosine to monophosphate adenosine, and the NT5E gene, encoding 5′-nucleotidase, were found to influence rapid progression toward epileptogenesis after TBI (Diamond et al., 2015).
A SNP of the TRPM6 gene which encodes for the transient receptor potential cation channel subfamily M member 6 and mainly acts as a magnesium transport regulator, has been associated with susceptibility to develop poststroke epilepsy (Fu et al., 2019).
Overall, there are no currently validated genetic susceptibility/risk biomarkers for epileptogenesis, due to the low level of evidence and lack of replication studies.
Only a few epilepsies have an established genetic etiology, either with Mendelian inheritance, oligogenic or polygenic contribution. De novo pathogenic single gene mutations are identified in 30%–50% of patients with developmental and epileptic encephalopathies, with >100 epilepsy-related genes discovered in recent years associated with severe early-onset epilepsies (Heyne et al., 2019). These are rare genetic variants with large effect, ranging from large deletions with high, on average ~7-fold risk for epilepsy to single, causative de novo variants. These variants can be diagnostically relevant, and their identification is crucial for precision medicine. Genetic diagnostic biomarkers will contribute to development of targeted treatments and development of new drug-screening protocols.
While rare variations of large effect have a clear impact in clinical practice for rare epilepsy syndromes, the value of routine genetic testing is not established for the majority of common epilepsies. Common epilepsies usually have oligogenic or polygenic contribution, implying involvement of multiple genes, for example, in idiopathic generalized epilepsies (IGEs) or self-limited focal epilepsies (Leu et al., 2019).
A whole exome sequencing study of 9,000 individuals with epilepsy highlighted a ubiquitous role for GABAergic inhibition and found an excess of ultra-rare, deleterious variants, with an odds ratio (OR) of 1.4, in patients with common epilepsies, suggesting a convergence of genes implicated in both common epilepsies and epileptic encephalopathies (Feng et al., 2019). Another large genome-wide study found epilepsy-associated copy number variants (CNVs) in 1.5%–3% of patients with common epilepsies, including lesional focal epilepsies (Niestroj et al., 2020).
There is a great need for research on mechanisms of genetic modulators of epileptogenic processes after epileptogenic insults and on their contribution to variable responses to antiepileptogenic therapies. Most genetic factors have been assessed only as prognostic biomarkers for epileptogenesis after TBI, but they could be useful as prognostic biomarkers for epilepsy in patients with any other epileptogenic event.
In some genetic epilepsies, the identification of a genetic etiology can be associated with relevant prognostic information, including seizure and developmental outcome. The impact of genetic mutations must be interpreted in the context of clinical phenotypes. We need better understanding of the impact a given variant on the protein function, and often the genotype-phenotype relation cannot be established. Also the, electrophysiological effects caused by specific variants should be understood (Orhan et al., 2014). For example, benign familial neonatal seizures (BFNSs) can be caused by loss-of-function variants in potassium channel genes KCNQ2 and KCNQ3. This genetic etiology is usually associated with a positive prognosis and seizure freedom in adulthood; early reduction and withdrawal of treatment may be feasible. However, more severe phenotypes, including neonatal epileptic encephalopathies, associated with different variants in the same genes, have emerged. Missense variants in the SCN2A gene usually cause gain-of-function effect and are associated with benign familial neonatal-infantile seizures with good prognosis and early medication withdrawal; nonsense mutations cause far more severe phenotypes such as developmental and epileptic encephalopathies (Reynolds et al., 2020).
Genetic biomarkers could be useful to predict the response to therapy in subpopulations of patients carrying different genetic variants. If there is an underlying specific gene variant affecting brain excitability, a rational treatment strategy might aim to reverse or circumvent the dysfunction. This is the current targeted treatment approach in precision medicine: identification of causative mutations, determination of physiologic effects, and evaluation of an intended treatment’s ability to reverse or inhibit the functional alteration (Balestrini et al., 2018).
Four genes and seven RNA probes were identified in cortical tissue from 24 people with epilepsy, predicting seizure-free outcome following anterior temporal lobectomy with amygdalo-hippocampectomy (Gallek et al., 2016).
At present there are no genetic markers that can be used in clinical practice to monitor treatment response. However, there is some evidence that antiseizure medications (ASMs), like sodium valproate, can exert their regulatory effects on gene expression as epigenetic modifiers, that is, determining a heritable modulation in gene expression in the short term (Tremolizzo et al., 2012). Epigenetic mechanisms include DNA methylation, histone modification, and untranslated RNA regulation.
Metabolism of ASMs is mostly mediated by the cytochrome P450 (CYP) family. Some CYPs have genetic (allelic) variants, encoding isoforms of differing activity, which may affect serum ASM concentrations, with subsequent potential risk of drug toxicity (Balestrini et al., 2018). For example, single-nucleotide polymorphisms (SNPs) in CYP2C9 and CYP2C19 genes can lead to significant differences in ASM serum levels (López-García et al., 2017). CYP2C9 accounts for about 90% of the metabolism of phenytoin. Individuals carrying CYP2C9 alleles encoding variant enzymes, for example, CYP2C9*2 (rs1799853) and CYP2C9*3 (rs1057910(C)), with reduced activity, metabolize phenytoin at a considerably slower rate compared with individuals homozygous for the wild-type (CYP2C9*1; rs1057910(A)) allele and have a greater risk of developing concentration-dependent neurotoxicity. Despite extensive evidence, pretreatment pharmacogenetic testing for CYP2C9 variants is not recommended in routine practice, with monitoring for clinical signs of toxicity and serum drug level being the standard approach.
In terms of side effects, there is an association between certain human leukocyte antigen (HLA) alleles and increased risk of idiosyncratic adverse drug reactions. In patients from Han Chinese and other South Asian ethnic groups, there is a high prevalence of the HLA-B*15:02 allele, which is associated with a risk of Stevens-Johnson syndrome induced by carbamazepine (Chung et al., 2004) and other aromatic ASMs (Cheung et al., 2013). Genotyping for HLA-B*15:02 before commencing treatment with carbamazepine, and possibly also with phenytoin and lamotrigine, is recommended in patients from these ethnic groups to avoid giving carbamazepine to carriers of this allele, if possible (Amstutz et al., 2014). The presence of the HLA-A*3101 allele is associated with carbamazepine-induced hypersensitivity reactions among subjects of Japanese and Northern European ancestry (McCormack et al., 2011). Genetic variation in the complement factor H-related 4 (CFHR4) gene and complement factor H (CFH) genes has been associated with phenytoin- induced maculopapular exanthema in European descent patients, suggesting a new link between regulation of the complement system alternative pathway and phenytoin-induced hypersensitivity (McCormack et al., 2011).
Nearly all book chapters or brain imaging reviews start by reminding us how imaging revolutionized the way we understand and investigate epilepsy. Magnetic resonance imaging (MRI), positron emission tomography (PET), and single-photon emission computerized tomography (SPECT) are routinely used in the clinic to guide epilepsy surgery in patients with focal drug-resistant epilepsy. However, advanced neuroimaging techniques provide quantitative surrogate markers for targets at molecular, cellular, and neuronal system levels, with an untapped potential to monitor a myriad of biologic processes participating in epileptogenesis or disease progression. MRI is noninvasive, safe, and easily translatable to clinical epileptogenesis studies, emphasizing its value as a platform for biomarker analysis. However, while neuroimaging has high sensitivity to detect an abnormality, its specificity for epileptogenesis is uncertain. Most human data have been obtained in patients with established, often drug-resistant epilepsy. So far, only a few experimental and clinical studies of posttraumatic epilepsy have applied imaging to biomarker discovery.
Preclinical studies might identify true susceptibility or risk biomarkers indicating potential for developing epilepsy in a healthy individual. Pitkänen and Immonen (2014) investigated whether quantitative T2, T1ρ, and diffusion assessed with MRI at 9 d, 23 d, or 2 months after TBI in perilesional cortex, thalamus, and hippocampus predict seizure susceptibility in the pentylenetetrazol (PTZ) seizure-susceptibility test 12 months after TBI. The highest predictive value for development of seizure susceptibility was achieved by coassessment of diffusion in perilesional cortex and thalamus at 2 months after TBI. Assessing individual MRI parameters in perilesional cortex or thalamus at 9 d after TBI also provided high sensitivity and specificity for predicting increased seizure susceptibility at 12 months after TBI, for example, T1ρ, AUC 0.929, with 90% sensitivity and 80% specificity (see Table 39–3; Higashida et al., 2011). Interestingly, the hippocampal parameters did not increase sensitivity. Differences between animals with and without epilepsy, however, were not evaluated, and the study relied on the assumption that animals with increased seizure susceptibility are undergoing epileptogenesis.
Summary of Potential Evidence for Imaging Biomarkers Associated with Epilepsy.
Measurements of myo-inositol levels, a marker of astrocyte activation, were measured in a model of epilepsy induced by pilocarpine-evoked status epilepticus (SE) (Pascente et al., 2016). Myo-inositol levels increased transiently in the hippocampus of SE rats not developing epilepsy while this increase persisted until spontaneous seizures onset in epilepsy-prone rats, being associated with a local increase in S100β-positive astrocytes. These findings suggest that myo-inositol levels have potential clinical relevance for identifying individuals at risk for developing epilepsy following exposure to epileptogenic insults and, consequently, for designing adequately powered antiepileptogenesis trials.
More recently, Huttunen and coworkers (2018) hypothesized that functional MRI (fMRI) can be used to identify brain areas with hyperexcitability during postinjury epileptogenesis. Functional MRI was used to detect the onset and spread of blood oxygenation level-dependent (BOLD) activation after PTZ-induced seizures at 2 months following lateral fluid-percussion injury (FPI) in rats. In 56% of rats with TBI, the PTZ-induced BOLD response started in the ipsilateral perilesional region and was followed by activation of contralateral cortex. In the remaining 44% of TBI rats, as well as sham-operated controls, BOLD response onset was bilateral. These findings suggest that perilesional cortex is sensitive to PTZ-induced hyperexcitability after TBI.
Another multimodel imaging study using the controlled cortical impact (CCI) model of TBI in rats, investigated diffusion tensor imagery with MRI and translocator protein (TSPO) as a neuroinflammation PET marker that binds to the 18 kDa translocator protein, showed that these measures on their own or in combination were able to predict variability in different long-term sequalae post TBI such as behavioral deficits and PTZ sensitivity (Missault et al., 2019).
The difficulty with these studies is that they rely on surrogate markers of excitability such as PTZ threshold, as often few animals develop spontaneous recurrent seizures or very long lead times are necessary.
Certain features of TBI or stroke are highly prognostic for development of epilepsy. Imaging can identify risk factors, but not valid biomarkers for acquired epilepsies. Following ischemic stroke, there is an increased risk of poststroke epilepsy depending on stroke severity and middle cerebral artery involvement, but these factors are only relevant when assessed within the context of other clinical features, such as large-artery atherosclerotic etiology, early seizures, and cortical involvement (Galovic et al., 2018).
Structural etiologies for epilepsy have predictive value for ASM or surgical treatment response. MRI evidence of hippocampal sclerosis is not necessarily related to seizure severity and may occur in individuals who never had seizures, as in asymptomatic relatives of patients with familial mesial TLE, suggesting that hippocampal sclerosis in this condition is not a consequence of recurrent seizures and is determined by a strong genetic predisposition (Kobayashi et al., 2002). An abnormality detected in an otherwise healthy individual is of little relevance for the development of epilepsy, unless it is a known epileptogenic abnormality, like a long-term epilepsy associated tumor (LEAT) (Thom et al., 2012). One could argue though that a LEAT would not be an incidental finding in an otherwise healthy individual.
Blood–brain barrier (BBB) damage is a risk factor for epileptogenesis following either TBI or stroke. BBB dysfunction that results in brain entry of blood components to the extracellular space can lead to epilepsy or aggravate epilepsy. In experimental epilepsy models, BBB leakage is commonly detected by tracers, and brain extravasation can be detected and quantified using in vivo imaging studies (Chassidim et al., 2015). Precise understanding of BBB dynamics during epileptogenesis may be of importance for assessment of future therapies, including those targeting BBB leakage. In patients, gadolinium (Gd)-based MR contrast agents are routinely used to detect BBB opening. When BBB is disrupted, intravenously administered MR contrast agent leaks out of blood vessels and accumulates in brain parenchyma (Rüber et al., 2018). MR signal changes can be used to localize BBB disruption and quantify the relative degree of BBB permeability. However, quantitative BBB permeability imaging is not used in routine clinical settings due to complicated and demanding dynamic scanning protocols, including tracer injection during scanning and long scanning sessions. As understanding of the critical role of the BBB in brain pathologies continues to advance, the need for establishment of practical BBB imaging methods in clinical settings is becoming urgent. In the post-pre approach the differences between T1-weighted MRI scans before and after tracer infusion are calculated, whereas the dynamic approach with fast T1 mapping detects dynamic changes during tracer infusion (Rüber et al., 2018).
Patients with mild to moderate TBI who developed PTE had a higher likelihood of abnormal BBB permeability with more affected cortex on dynamic contrast enhanced MRI traumatic patients without epilepsy (Tomkins et al., 2011). Receiver operating characteristic curve (ROC) analysis revealed BBB dysfunction in the injury area may be a risk biomarker for epileptogenesis after TBI with an AUC of 0.85 (Tomkins et al., 2008). These results further support a role for microvascular pathology, specifically BBB dysfunction, in PTE pathogenesis. Even though BBB opening per se may lead to increased seizure occurrence, BBB permeability may serve as a candidate biomarker for epileptogenesis (Vezzani et al., 2011).
Some studies suggest that human herpes virus 6 (HHV-6) is associated with a history of childhood complex or prolonged febrile seizures or febrile status epilepticus (FSE) and may lead to persistent brain infection and later MTLE (Lewis et al., 2014; Akinsoji et al., 2020). Febrile seizures have been associated with early epilepsy onset and hippocampal volume loss in patients with MTLE. The FEBSTAT study suggested that increased T2 signal associated with febrile status epilepticus (FSE) is most prominent in the hippocampal region CA1, with HHV-6 as a risk factor (Lewis 2014).
The diagnosis of epilepsy and its associated syndromes is based on clinical symptoms and scalp EEG. These data increasingly are supplemented by noninvasive neuroimaging modalities. The range of clinically relevant imaging modalities is rapidly expanding and includes MRI, PET, SPECT, computed tomography (CT), and ultrasound for neonatal and pediatric seizures. MRI is the imaging modality of choice to evaluate patients for structural abnormalities, particularly for assessment as surgical intervention candidates.
Structural imaging is important for identifying individuals with a particular subtype of epilepsy, either by detecting a relevant pathognomonic abnormality, for example, hippocampal atrophy for mesial TLE, or by exclusion as for IGE. Particular advances have been made in identifying subtle structural and functional imaging abnormalities in the latter group of patients, but these are almost exclusively group findings, and not of benefit yet for the individual patient (see Ratcliffe et al., 2020, for review on the interplay between imaging abnormalities and cognitive function in IGEs). More advanced imaging analysis allows to further refine diagnosis of structural epilepsies like focal cortical dysplasia (FCD) into subtypes (Hong et al., 2017).
The presence or absence of a structural or functional imaging abnormality has primarily predictive or prognostic value, and for this reason these imaging technologies are discussed below.
Understanding the consequences of newly discovered single gene mutations causing human epilepsy has the potential to yield new insights into the underlying mechanisms of this disorder. A mutation of the gamma2 subunit of the GABAA receptor, which substitutes glutamine for arginine at position 43 (R43Q) has been found in a familial generalized epilepsy. Individuals affected by the GABRG2(R43Q) mutation had reduced [11C]-FMZ binding likely linking this gene defect with the epilepsy phenotype (Fedi et al., 2006). Succinic semialdehyde dehydrogenase (SSADH) deficiency is an autosomal recessive disorder of GABA metabolism characterized by elevated levels of GABA and gammahydroxybutyric acid (Pearl et al., 2009). SSADH-deficient patients showed widespread downregulation of GABAA binding site availability, in particular affecting the amygdala, hippocampus, cerebellar vermis, frontal, parietal, and occipital cortex, which suggests a potential mechanism for neurologic dysfunction in a serious neurodevelopmental disorder with high endogenous brain GABA levels.
Hippocampal sclerosis is generally viewed as a diagnostic marker for TLE, but it is also a prognostic marker for drug resistance and for material-specific cognitive impairment related to laterality. The typical signs of hippocampal sclerosis, that is, volume reduction, volume asymmetry, increased T2-weighted signal, and loss of internal architecture, can be distinguished reliably on MRI. The specific pattern of hippocampal atrophy relates to the severity and type of memory deficits in people with epilepsy (Postma et al., 2020). Hippocampal abnormalities can be detected in unaffected siblings of individuals with TLE and may represent a disease endophenotype (Long et al., 2020). On the other hand, cortical thinning in entorhinal and parahippocampal cortex, and outside the temporal lobes, is usually not found in unaffected relatives of people with epilepsy and may, thus, be an acquired phenomenon mainly driven by disease-related factors.
Frequently affected subcortical structures include thalamus and basal ganglia (Wheelan et al., 2018). Involvement of these areas may increase propensity of seizures to generalize (He et al., 2020) and odds of poor postsurgical outcome or sudden unexpected death in epilepsy (Wandschneider et al., 2015).
Emerging evidence from large-scale cross-sectional and longitudinal MRI studies suggests epilepsy is a progressive rather than static condition. This is reinforced by the notion that epilepsy-related cortical changes are mainly observed in adults, whereas pediatric populations have no or little cortical thinning.
A large number of cross-sectional studies linked the severity and extent of brain atrophy with longer epilepsy duration (Caciagli et al., 2017). These studies have suggested that damage to cortex and hippocampus accumulates during the course of epilepsy. However, a cross-sectional design does not allow definitive conclusions because disease duration is strongly intercorrelated with participant age, that is, older individuals having a longer duration of epilepsy.
Longitudinal imaging is a powerful method for assessing ongoing neurodegeneration in epilepsy and may be helpful for development of disease-modifying treatments. It is generally accepted that current ASMs do not have a disease-modifying effect on epilepsy. In contrast, successful epilepsy surgery may lead to “cure” if a sufficient proportion of the epileptic network is removed. Recent evidence suggests that epilepsy surgery may be neuroprotective, because it prevents further cerebral damage (Galovic et al., 2020). Around three-quarters of individuals with epilepsy under follow-up in a tertiary epilepsy center show progressive cortical thinning (Galovic et al., 2019). In a population-based cohort, progressive changes can be detected in around half of those with chronic epilepsy and a third of those with new-onset seizures (Liu et al., 2003). Due to these observations it seems unlikely that suppressing seizures with ASMs or surgery may interrupt the ongoing neurodegeneration in epilepsy.
Functional MRI (fMRI) is used to probe the functional cognitive anatomy in vivo and to identify disturbances associated with neurological dysfunction. Paradigms of particular relevance to epilepsy are language and memory, which are commonly impaired, particularly in those with a long history of epilepsy, a high burden of seizures, and taking medication that affects cognition.
Auditory and visual naming fMRI activate posterior temporal lobes; activation strength is correlated with clinical naming performance. In TLE, a later age of epilepsy onset was associated with stronger activation in the temporal lobe, and an earlier age of onset was associated with less deactivation extra-temporally, in parts forming the default mode network (DMF), that need to be suppressed for effective cognition (Trimmel et al., 2018).
Further, in those with left TLE, shorter epilepsy duration and lower seizure frequency were related to stronger temporal lobe activations, and longer epilepsy duration and higher seizure frequency were related to reduced deactivation. A high seizure burden, long epilepsy duration, and early onset age are associated with greater temporal lobe language network dysfunction. Longitudinal studies will help to unravel causative factors, assess potential reversibility, and assess usefulness as a biomarker to evaluate therapeutic strategies.
Working memory is fundamental to many cognitive tasks. fMRI studies of working memory have shown functional connectivity of networks integral to working memory is abnormal in TLE and associated with deranged structural connectivity (Stretton et al., 2013). Memory encoding fMRI for verbal and visual material activates hippocampus with material specificity. Lower seizure frequency and a lesser duration of epilepsy were associated with less disruption of the memory encoding network; conversely, higher seizure frequency and longer duration were associated with more inefficient, extra-temporal reorganization of memory encoding networks (Sidhu et al., 2015).
Current hypotheses for pharmacoresistance in epilepsy include, among others, the transporter and the target hypothesis, but none is able to explain convincingly how drug resistance arises (Loescher et al., 2013). Molecular imaging techniques have helped to investigate these proposed mechanisms of pharmacoresistance in epilepsy.
Multidrug efflux transporters such as P-glycoprotein contribute to drug resistance by pumping ASM from the cell at the BBB. To study the contribution of P-glycoprotein to pharmacoresistance in vivo, an experimental PET protocol involving a radiolabeled P-glycoprotein substrate, [11C]verapamil, and partial, half-maximum P-glycoprotein blockage with tariquidar has been proposed (Bankstahl et al., 2008). The brain uptake of [11C]verapamil would correlate with the magnitude of P-glycoprotein action at the BBB.
PET imaging showed increased Pgp transporter activity in pharmacoresistant epilepsy comparing patients with refractory TLE and seizure-free TLE patients (Feldman et al., 2013). Pharmacoresistant patients had a lower baseline influx rate constant (K1) corresponding to a higher P-glycoprotein activity in the hippocampus temporal neocortex with seizure frequency correlating positively with P-glycoprotein activity in the hippocampus and on a whole brain level. The findings of [11C]verapamil PET were validated in ex vivo specimens of five patients who underwent epilepsy surgery. Increased P-glycoprotein function before surgery and a decrease postoperatively were associated with optimal surgical outcome, indicating that P-glycoprotein expression responded dynamically to therapeutic procedures or changes in seizure frequency (Bauer et al., 2014).
PET studies in focal epilepsies of temporal and extra-temporal lobe origin demonstrate widespread abnormalities of [11C]-flumazenil (FMZ), which binds to the benzodiazepine binding site of the γ-aminobutyric acid (GABAA) receptor complex containing α1-, α2-, α3-, and α5 subunits. In mesial TLE patients, decreases in benzodiazepine receptor (BZR) density were usually limited to the anterior mesial temporal lobe (Koepp et al., 1997) with atrophy not accounting for the decreased hippocampal BZR density (Koepp et al., 1998). Focal BZR decreases in mesial TLE show a high rate of concordance with the ictal onset zone determined by intracranial ictal EEG monitoring (Ryvlin et al., 1998), but they can be altered by recent or increased seizures shortly before PET data acquisition (Bouvard et al., 2005). In a mixed group of nonlesional focal epilepsy patients, increased seizure frequency was inversely correlated with uptake in the frontal piriform cortex independent of the site of seizure onset (Laufs et al., 2011). Similarly, a crucial epileptogenic area has been described in the prepiriform cortex of rats and monkey and termed “area tempestas” (Piredda & Gale, 1985).
In patients with TLE and normal MRI, [11C]Ro15-4513, which binds predominantly to α5- and α1-containing GABAA receptors, showed an increase of the α5-selective component of the signal in the hippocampus, temporal lobe, anterior cingulate gyri, and the piriform cortex (area tempestas) (McGinnity et al., 2021). While [11C]FMZ is mainly indicative of the distribution of α1-subunit, [11C]Ro15-4513 is a partial inverse benzodiazepine agonist for α5 subunit-containing GABAA receptors. Consistent with previously observed loss in [11C]FMZ binding close to the seizure focus, α1-containing receptors were reduced compared to those in healthy controls, suggesting a possible loss in subunits α1, α2, or α3 in the seizure focus, whereas receptors containing α5 are up-regulated.
The N-methyl-d-asparate (NMDA) glutamate receptor has been implicated in excitotoxic neuronal damage and epileptogenesis. Multiple attempts of developing suitable PET ligands for imaging TLE failed usually showing reduced tracer uptake, possibly reflecting reduced NMDA receptor density, reduced perfusion, or focal atrophy. [11C]ketamine, which binds to the open NMDA receptor ion channel, showed a modest average ipsilateral binding reduction of 14% in mTLE (Kumlien et al., 1999).
Similarly, the [18F]GE-179 PET tracer also binds to the phencyclidine site within the NMDA ion channel pore, and so indicating the activated state of the receptor (McGinnity et al., 2014). The utility of [18F]GE-179 PET as a use-dependent marker of NMDA receptors was validated by showing focal [18F]GE-179 uptake during electrical deep brain stimulation (DBS) of pig hippocampus and the demonstration of specificity for the phencyclidine site by blockade with S-ketamine (Vibholm et al., 2020). Patients with epilepsy had a significantly increased global radioligand binding compared to controls, suggesting increased NMDA receptor activation in a mixed group of refractory epilepsy patients that could point to ongoing epileptogenesis (McGinnity et al., 2015). The most surprising finding was a markedly decreased global binding in three epilepsy patients taking antidepressants, which fell below that of healthy controls. The potential influence of antidepressants and, possibly, depression on activation of NMDA receptors warrants further research.
Aberrations in the number and function of glutamate AMPA receptors are thought to be involved in epileptogenesis. In an exploratory clinical study including patients with epilepsy, increased [11C]K-2 uptake was detected in the epileptogenic focus of patients with mesial TLE, which was closely correlated with the local AMPA receptor protein distribution in surgical specimens from the same individuals (Miyazaki et al., 2020).
Other advances include the development of new tracers for subtype 5 of the metabotropic glutamate receptor (mGluR5) (Ametamey et al., 2007). Whereas upregulation of mGluR5 in surgical hippocampal specimen of epilepsy patients has been reported (Das et al., 2012), a more consistent finding is the decrease of mGluR5-mediated long-term depression in a rat model of spontaneous seizures (Kirschstein et al., 2007). Using [11C]ABP688 PET in a similar epilepsy rat model, Choi et al. (2014) showed reduced global binding in the acute phase after status epilepticus. Using the PET radioligand [11C]ABP688, mGluR5 availability was reduced in patients with focal epilepsy and FCD (DuBois et al., 2021). In addition, there was evidence that ongoing epileptic activity may alter chemoarchitectural brain organization in distant regions that are essential for network integration. Focally reduced mGluR5 availability in the epileptogenic zone might reflect receptor internalization or conformational changes in response to excessive extracellular glutamate, supporting a potential role for mGluR5 as a therapeutic target in human MTLE.
The interest in serotonin has led to the development of several suitable PET tracers that study three different aspects of cerebral serotonin function. First, [11C]α-methyl-l-tryptophan (AMT) PET has been used to address serotonin synthesis (Natsume et al., 2003). The downside of this approach is that serotonin metabolism might be disturbed in disease, and the findings are difficult to interpret in a pathophysiological context. Second, the density of 5-HT1A receptors has been measured with several PET tracers that differ in their pharmacological properties. 2′-methoxyphenyl-[(N-2′-pirydynyl)-p-[18F]fluorobenza-midoethylpiperazine ([18F]MPPF) is a selective antagonist of 5-HT1A with an affinity close to that of serotonin and is, hence, sensitive to endogenous serotonin variations (Merlet et al., 2004). Conversely, [18F]trans-4-fluoro-N-2-[4-(2-methoxyphenyl)piperazin-1-yl]ethyl]-N-(2-pyridyl)cyclohexanecarboxamide ([18F]FCWAY) and [O-methyl-11C]-N- (2- (4- (2-methoxyphenyl)-1-piperazinyl) ethyl)- N- (2-pyridinyl) cyclohexanecarboxamide trihydrochloride ([11C]WAY100635) are high-affinity agonists of 5-HT1A, and they do not interact with endogenous serotonin (Toczek et al., 2003). Finally, [11C]DASB can measure the serotonin transporter 5-HTT, the main terminator of synaptic serotonin effect (Martinez et al., 2013).
Originally, increased tracer uptake was thought to reflect increased serotonin synthesis, though recent evidence in epileptogenic tubers of patients with tuberous sclerosis complex (TSC) point to increased tryptophan metabolism in the presence of neuroinflammation via the kynurenine pathway leading to the production of proconvulsants. The differentiation of epileptogenic and nonepileptogenic tubers was extensively examined with [11C]AMT-PET. and increased tracer binding was consistently demonstrated in tubers that co-localized with ictal EEG findings (Kagawa et al., 2005).
Decrease of 5-HT1A receptor density ipsilateral to seizure focus is a consistent finding in TLE using different PET tracers, such as [18F]MPPF, [11C]WAY-100635, and [18F]FCWAY. The decreases were more pronounced in the hippocampus and in areas involved in seizure generation consistent with the concept of a proconvulsive effect of serotonin depletion. The use of 5-HT1A receptor PET in presurgical epilepsy evaluation has only been tested in small patient samples. 5-HT1A receptor PET showed decreased temporal binding in more than 80% of these cases, and all pathological decreases were congruent with the lateralization of the ictal onset on EEG (Didelot et al., 2008). All patients with lateralizing [18F]MPPF-PET became seizure-free after surgery. These results suggest a higher specificity than FDG-PET; however, they remain to be reproduced in larger patient samples. The sensitivity of 5-HT1A receptor PET can be further improved by comparing tracer binding between both cerebral hemispheres using asymmetry indices. This might increase the sensitivity to above 90% with a specificity of 88%.
Some studies reported abnormalities beyond the temporal lobe, describing decreased [18F]FCWAY binding in the insular cortex and anterior cingulate (Giovacchini et al., 2005). Such changes in the limbic areas were significantly more common in epilepsy patients with concomitant depression compared to those without mood disturbances, and the magnitude of hippocampal binding inversely correlated with depressive symptoms (Theodore et al., 2007). These results suggest a common pathomechanism of epilepsy and comorbid depression due to a decrease of serotonin receptors that extends beyond the temporal lobe and affects limbic structures.
The serotonin transporter 5-HTT was studied using the tracer [11C]DASB in 13 TLE patients and 16 controls (Martinez et al., 2013). There were no regional differences in 5-HTT function between patients and controls. However, epilepsy patients with history of depression had a relatively reduced 5-HTT activity in the ipsilateral insula compared to those without depression. A reduced transporter activity would decrease serotonin reuptake and might represent a compensation mechanism to increase extracellular serotonin concentration. Insular 5-HTT activity correlated with 5-HT1A receptor density measured with [18F]FCWAY-PET, indicating that greater loss of 5-HT1A receptors may lead to decreased reuptake of serotonin.
Several PET tracers have been developed to target neuroinflammation, and its use for epilepsy was recently reviewed (Bouilleret and Dedeurwaerdere, 2021). The most commonly used radioligands are [11C]PK11195 and [11C]PBR28; they bind to TSPO, a marker of activated microglia (Amhaoul et al., 2015). The concentration of TSPO is very low in healthy brain tissue; however, a marked increase can be observed in experimental models of neuroinflammation and reliably measured with TSPO-PET tracers. In rat models of epileptogenesis, increased TSPO expression pointing to neuroinflammation can be demonstrated with PET (Dedeurwaerdere et al., 2012). The inflammatory process peaks 2 weeks after initial SE, but limbic activation of microglia persists into the chronic phase. TSPO PET has shown discriminative power between phenobarbital-responding versus resistant rats with spontaneous recurrent seizures (Bogdanovic et al., 2014). The resistant rats also had higher baseline levels of seizures; therefore, it cannot be completely dissected out whether the TSPO PET signal is reflecting pharmacoresistance or rather intrinsic disease severity. Additional studies in rats have also concluded that modeling of TSPO signal in the brain using PET correlates with seizure burden during life reflecting intrinsic disease severity (Bertoglio et al., 2017). It is tempting to speculate that it could also provide a predictive biomarker for response to anti-inflammatory treatments, but this has not been demonstrated thus far.
Two well-powered studies in TLE showed increased [11C]PBR28 uptake, pointing to activation of microglia, in temporal regions ipsilateral to the epileptic focus and, to a lesser extent, in the ipsilateral thalamus and contralateral temporal lobe (Hirvonen et al., 2012; Gershen et al., 2015). In a follow-up study in patients with neocortical seizure onset, another TSPO-PET tracer, [11C]-DPA713, was used to investigate neuroinflammation in the epileptogenic zone together with [11C]PBR28 (Dickstein et al., 2019). These in vivo findings give support to a local neuroinflammation in human epileptic cortex. Although the causal role of brain inflammation in generating seizures still needs to be confirmed, these observations might provide a rationale for anti-inflammatory treatment in some epilepsy patients.
Imaging biomarkers can assist in determining who might benefit or derive harm from a particular treatment or therapeutic interventions. The vast majority of imaging studies has focused on surgical treatments, both identifying those people who most likely benefit from certain surgical interventions mainly based on structural imaging, whereas functional imaging is primarily used to identify those people whose eloquent cognitive or motor functions will be harmed by the planned surgical resection (Szaflarski et al., 2017).
The best predictor for poor response to ASM is the presence of a brain lesion (Semah et al., 1998). Similarly, a large number of studies have shown the best predictor for a good surgical outcome is the presence of an epileptogenic lesion, which co-localizes with ictal EEG recordings (DeTisi et al., 2011). Focal or regional functional deficits on PET cerebral glucose metabolism or SPECT blood flow imaging also predict good surgical outcome when in agreement with ictal EEG, even if no structural lesion is present (Theodore 2017).
Routine MRI sequences read by expert radiologists failed to detect 57% of epileptogenic lesions (VonOertzen et al., 2002). Dedicated MR protocols can significantly improve detection of abnormalities on structural imaging (Bernasconi et al., 2019). Algorithms can reliably predict seizure relapse (Lamberink et al., 2017), but there are no specific imaging findings to indicate whether someone is at particular risk for medically or surgically refractory epilepsy.
Several PET investigations examined effects of vagus nerve stimulation (VNS) on cerebrospinal fluid (CBF) or glucose metabolism (Dedeurwaerdere et al., 2005). Acute VNS-activation PET studies, performed within the first 24 hours after VNS therapy began, found synaptic activations in the dorsal medullary complex of the vagus, the central pons and midbrain, inferior cerebellum, hypothalami, and thalami, and also showed a combination of activations and deactivations in amygdalae, hippocampi, insulae, and other neocortical sites bilaterally (Henry et al., 1998). Chronic VNS-activation PET studies, performed after 3 months of ongoing VNS therapy, showed synaptic activations in the same brainstem, cerebellar, and diencephalic sites as were observed acutely, but cortical effects were markedly reduced on chronic as compared with acute CBF imaging (Henry et al., 2004). Patients who had greater bilateral thalamic activation went on to experience significantly greater seizure reduction during VNS than those who had little or no thalamic activation. A significant difference in metabolic connectivity evaluated by preoperative FDG-PET was noted between VNS-effective and VNS-ineffective groups (Yu et al., 2018). Relative changes in glucose metabolism were strongly connected among the areas of brainstem, cingulate gyrus, cerebellum, bilateral insula, and putamen in patients with <50% seizure control after VNS.
Chronic treatment with vigabatrin, which increases synaptic GABA availability by inhibiting enzymatic degradation, was associated with receptor downregulation of GABA receptors in FMZ PET studies in childhood epilepsy, possibly due to increased GABA levels potentially exacerbated by elevations of other metabolites (Juhasz et al., 2001).
Research into the effects of ASM using PET techniques has mainly focused on mechanisms of ASM toxicity and to a lesser extent on ASM kinetics and ASM effects on specific neurochemical systems.
Cerebral kinetics of phenytoin were mapped using [11C]phenytoin, which rapidly entered cerebral gray matter and equilibrated within 20 minutes on dynamic PET imaging (Baron et al., 1983). The early kinetics of [11C]phenytoin did not differ between the epileptogenic region and its contralateral homolog. The ASMs levetiracetam and brivaracetam were labeled with the view to image the synaptic vesicle glycoprotein 2A (Nabulsi et al., 2016). Using the SV2A PET tracer [11C]-UCB-J, reductions in tracer binding were observed in the sclerotic hippocampus of patients with unilateral MTLE (Finnema et al., 2020).
The effects of chronic valproate therapy on GABAergic systems was studied in patients with generalized epilepsy showing generalized cerebral decreases in [11C]FMZ binding (Koepp et al., 1997b), but interictal [11C]FMZ activity was normal in patients with absence seizures who were receiving ASMs other than valproate (Prevett et al., 1995). This observation suggests that valproate may act to increase cerebral endozepine concentration, which would reduce the availability of central benzodiazepine receptors to bind FMZ at the GABAA–chloride ionophore complex; reduced affinity of the receptor for FMZ or reduced neuronal expression of the receptor (or of the entire GABAA complex) is less likely to explain the association of valproate use and [11C]FMZ activity reduction.
The effect of ASM on cerebral metabolic rate of glucose has been studied with FDG PET. Overall, absolute values were found to be depressed by drugs acting through different mechanisms. In order of decreasing the effect size, drugs studied included barbiturates (~27%; Theodore et al., 1986b), valproate (~22%; Leiderman et al., 1991), carbamazepine (~16%; Theodore et al., 1989), and phenytoin (~12%; Theodore et al., 1986a), with evidence of moderate regional variability. Chronic use of these drugs did not appear to be associated with altered serotonin-1A-receptor affinity for [18F]FCWAY or with the cerebral density of serotonin-1A receptors (Theodore et al., 2006).
Cognitive fMRI can probe functional anatomy and determine clinical factors that affect eloquent functions acting as a biomarker for dysfunction and response to therapies. Dysfunctional networks appear to extend far beyond the presumed main disease focus, including major brain networks relevant to cognitive function, such as working memory and DMN. The DMN is a key network for both cognitive dysfunction and seizure manifestations. Critical disease hubs co-localize with major DMN subsystems, that is, the medial temporal lobe. Function within these networks is directly modulated by disease factors, such as seizure frequency and disease duration (Danielson et al., 2011).
Language fMRI studies examining ASM effects on cognitive activation and deactivation patterns showed decreased activation in task-positive regions, that is, dominant inferior and middle frontal gyri (IFG and MFG), and failure to deactivate task- negative regions, including DMN (Yasuda et al., 2013; Wandschneider et al., 2014, 2017; Xiao et al., 2018). ph-MRI evaluates networks underlying drug behavioral effects, independent of its mechanism of action. ASMs often target several receptor subtypes with varying regional distribution and target engagement; fMRI can monitor combined effect of these interactions across multiple brain regions at a network level and remotely from regions of highest target receptor densities, whereas PET and molecular studies can define target receptor occupancy and affinity without necessarily translating effects to large-scale networks. A further advantage is that fMRI does not use ionizing radiation and has no known biological side effects.
We do not yet have a useful biomarker for epileptogenesis, which would have to meet stringent criteria. It is rather unlikely that a single readout, like a protein measured in blood or a single imaging or neurophysiology biomarker, will be sufficiently reliable as a prognostic or predictive biomarker. It is more likely we will need to use combinations of different biomarkers in a multidisciplinary approach, complementing, for instance, fluid biomarkers with genetics, imaging, and EEG. Validation of single biomarkers and combinations of biomarkers will be difficult since only a minority of individuals experiencing a potential epileptogenic event will develop seizures, and the process may take many years, hampered as well by small-scale, uncoordinated studies, suffering from inherent design issues. A “strategic roadmap” for research priorities in epileptogenesis biomarker discovery, from preclinical studies through regulatory issues, optimizing use of resources, similar to efforts in cancer and Alzheimer disease research, has been proposed (Simonato et al 2021). Reaching these goals will depend on extensive international collaboration and commitments by funding agencies to long-term studies. Before human studies are started, crucial choices on model systems and trial designs must be made.
This is an open access publication, available online and distributed under the terms of a Creative Commons Attribution-Non Commercial-No Derivatives 4.0 International licence (CC BY-NC-ND 4.0), a copy of which is available at https://creativecommons.org/licenses/by-nc-nd/4.0/. Subject to this license, all rights are reserved.
Your browsing activity is empty.
Activity recording is turned off.
See more...
