Introduction

Globally, an estimated 2.4 million people are diagnosed with epilepsy each year. Thus, a new person is diagnosed with epilepsy every 13 s [1]. Epileptogenesis refers to the development and extension of tissue capable of generating spontaneous seizures, resulting in (a) the development of an epileptic condition and/or (b) progression of epilepsy after it is established [2]. In 60% of those affected, epileptogenesis is initiated by structural causes such as traumatic brain injury (TBI) [3, 4]. Recent epidemiologic data indicate that approximately 2.5 million people experience TBI annually, both in Europe and the USA. The risk of epileptogenesis increases according to the severity of TBI: about two- to four-fold after mild, eight-fold after moderate, and 16-fold after severe TBI [5, 6]. Up to 53% of patients with penetrating TBI develop epilepsy [7, 8]. Post-traumatic epilepsy (PTE) is estimated to account for approximately 5% of all epilepsies and 20% of structural epilepsies [9]. Mild TBI comprises over 90% of all TBI [10], and thus the total number of patients developing epilepsy after mild TBI can be expected to be greater than that of patients developing epilepsy after severe TBI, which has been the focus of experimental and clinical PTE studies (Fig. 1).

Fig. 1
figure 1

Potential mechanisms and therapeutic strategies to treat post-traumatic epilepsy. A schematic of the healthy, injured, and epileptic brain. Mechanisms that have been implicated in post-traumatic epileptogenesis are indicated with a red question mark. Mechanisms that may be targeted to reduce brain injury are indicated in blue. Mechanisms that may be targeted to reverse or prevent epileptogenesis are indicated in green. We also point out that small molecule and genetic screens may be able to identify novel mechanisms and treatment strategies

Full size image

TBI refers to an alteration in brain function, or other evidence of brain pathology, caused by an external force [11]. Seizures after TBI have been classically categorized into immediate (seizures ≤ 24 h post-TBI), early (≤ 7 days), or late (> 7 days) seizures. A person with PTE suffers repeated unprovoked seizures that result from TBI and occur more than a week after the initial injury [12]. According to the current International league Against Epilepsy definitions, PTE belongs to structural epilepsies, and is diagnosed if the subject experiences one unprovoked seizure > 7 days post-TBI [3, 12, 13]. Approximately 80% of TBI patients who eventually develop epilepsy will receive a PTE diagnosis within 2 years after the TBI [5, 15]. Both clinical and basic science studies have investigated a wide range of biological processes that may be involved in the transition from an injured brain to an epileptic brain. Changes occurring with TBI, including neuroinflammation, neuronal cell death, and changes in synaptic abundance and function, just to name a few, have been studied for their role in PTE [16,17,18]. While there are diverse cellular, molecular, metabolic, and circuit-level changes, we have not been able to identify a causative molecular or cellular event and its temporal relation to the occurrence of PTE. Innovation in both clinical care and basic science suggests that we may be closer than ever to identifying novel therapeutic approaches to prevent PTE. In this review, we examine past failures in preventing PTE, consider diagnostic and genetic information that could guide targeted interventions, and highlight exciting new approaches with may help reduce the prevalence and impact of PTE.

Old and New Strategies for Prevention of Epilepsy After TBI—Still No Treatments in Clinic

In a review of pharmacologic prophylaxis for PTE, Rapport II and Penry [19] cite the first anti-epileptogenesis studies conducted in head-injured patients using diphenylhydantoin [20, 21]. Since then, the concept of using compounds designed to suppress epileptic seizures (anti-seizure drugs) to prevent the complex molecular and cellular processes that drive epileptogenesis [anti-epileptogenic drugs (AEGs)] was expanded to carbamazepine, phenobarbital and valproic acid (for review, see [22]). According to ClinicalTrials.gov, initial studies on PTE are also planned using the third-generation anti-seizure drugs lacosamide (NCT01110187), levetiracetam (NCT01463033, NCT02631759, NCT00566046), and topiramate (NCT00598923). In addition, a study using acetylcholinesterase inhibitor, huperzine A (NCT01676311) was planned but is now discontinued. Biperiden, a cholinergic antagonist acting in the muscarinic receptor, is still under investigation (biperiden (NCT01048138)) [23]. So far, the use of anti-seizure drugs has not resulted in favorable anti-epileptogenic effects, and their use has been recommended only for the first post-injury week to suppress immediate and early seizures (Brain Trauma Foundation Guidelines) [24].

The recent rapid progress in modeling PTE provides an opportunity to vigorously assess preclinical candidate treatments in different clinically relevant epileptogenic injuries, mimicking the heterogeneity of TBI in humans with PTE (Table 1). Long-term video-electroencephalogram (EEG) monitoring studies have shown that epileptogenesis can be triggered in different strains of rats and mice by various injury types, including focal (e.g., controlled cortical impact-induced TBI, CCI), diffuse (repetitive weight-drop), mixed type injury (e.g., lateral-fluid percussion TBI), and blast injury. Although gaps in animal models remain, including epileptogenesis in females and younger animals, currently available animal models have identified candidate molecular, cellular, and network epileptogenic mechanisms that will pave the way for the discovery of novel treatments for PTE.

Table 1 Summary of in vivo recorded changes in excitability in different models of traumatic brain injury. Only the data that was collected at least 1 week post-injury is included
Full size table

Table 2 summarizes the current in vivo treatment studies on epileptogenesis in relevant animal models. Most of them have administered small molecules with a variety of mechanisms of action as a monotherapy. Commonly targeted mechanisms include oxidative stress, neuroinflammation, neuroprotection, and restoration of inhibitory GABAergic function. Unlike in status epilepticus-induced epileptogenesis models, gene therapy, administration of monoclonal antibodies, or treatments targeting DNA or RNA have not yet been tested in PTE models. Some laboratories have applied ketogenic diet, hypothermia, focal passive cooling, treadmill exercise, or transplantation of GABAergic progenitors (Table 2). These pharmacological treatments have typically been initiated within hours after the injury. Studies delivering anti-epileptogenic interventions at later time points would also greatly inform future clinical studies and would complement the development of biomarkers to identify high-risk patients at later timepoints. Ex vivo analysis of tissue excitability or in vivo analysis of seizure susceptibility, incidence of epilepsy, or characteristics of epilepsy phenotype (seizure frequency, duration or behavioral severity of seizures) have been used as outcome measures. Assessment of a 50% responder rate has been challenging as it would require large animal numbers as typically 25–50% of animals with TBI develop epilepsy within the 4–6 months follow-up (Table 2). So far, none of the treatments except the transplantation of GABAergic progenitors has been able to prevent the development of epilepsy.

Table 2 Preclinical proof-of-concept studies in animal models of traumatic brain injury (TBI). Only the studies with treatment initiated after the TBI have been included. Studies are in chronological order
Full size table

Advancing Biomarkers from Basic and Clinical Studies

One of the biggest challenges in preventing post-traumatic epilepsy is identifying patients who are at the greatest risk following injury. Only a fraction of those who suffer a traumatic brain injury will go on to develop epilepsy. Clinicians are often hesitant to use aggressive or experimental interventions when there is little certainty that a patient is at significant risk of developing PTE. Therefore, developing biomarkers that are predictive of PTE would be extremely useful in stratifying patients into those that have a low probability of developing epilepsy from those that are at significantly higher risk. For patients at high risk, targeted interventions may slow or prevent the development of PTE. With this goal in mind, much work has been devoted, both in clinical and basic epilepsy research, to identifying useful biomarkers predictive of the later development of PTE (Table 3).

Table 3 Prognostic and diagnostic biomarkers for epileptogenesis, post-traumatic epilepsy, and tissue epileptogenicity after TBI
Full size table

A biomarker is a characteristic that can be objectively measured as an indicator of normal biologic processes, pathogenic processes, or responses to an exposure or intervention, including therapeutic interventions. Biomarker modalities include molecular, histologic, radiographic, or physiologic characteristics. To improve the understanding and use of biomarker terminology in biomedical research, clinical practice, and medical product development, the FDA-NIH Joint Leadership Council developed the BEST Resource (Biomarkers, EndpointS, and other Tools). The seven BEST biomarker categories include (a) susceptibility/risk biomarkers, (b) diagnostic biomarkers, (c) monitoring biomarkers, (d) prognostic biomarkers, (e) predictive biomarkers, (f) pharmacodynamic/response biomarkers, and (g) safety biomarkers.

The current status of biomarker development in PTE field is summarized in Table 3. Many types of biomarkers have been examined to predict the development of epilepsy after different epileptogenic etiologies (for recent review see [88]), including molecular biomarkers (serum proteins, non-coding RNAs, cerebrospinal fluid (CSF)), imaging biomarkers (MRI, PET), and EEG based biomarkers. Unfortunately, no useful clinical biomarkers have been rigorously validated using sufficient patient populations. That said, there are a few exciting studies that suggest clinical biomarkers of PTE may be not that far down the road. A study of 256 Caucasian adults showed that high levels of the inflammatory cytokine IL-1β in the CSF, relative to levels in the serum, were associated with an increased risk of developing PTE over time [73]. While quite promising, this study had a small sample size and did not include statistical analysis of the predictive value of IL-1β CSF/serum ratios in PTE. None the less, this is an exciting line of research worthy of further development. In two other studies, EEG-based biomarkers suggest that epileptiform activity (epileptiform discharges, lateralized periodic discharges, generalized periodic discharges, or lateralized rhythmic delta activity) [79] and early seizures [80] are predictive of later development of PTE. These studies underscore the importance of robust, quantitative analysis in defining biomarkers of PTE and in applying statistical modeling to pinpoint the most predictive parameters in the data matrix as biomarkers. Moreover, analyses of the sensitivity, specificity, quantitative cut-off values for PTE prediction are needed to rigorously apply a biomarker in an individual subject [e.g., receiver operating characteristic analysis (ROC)]. While increased abnormal EEG and early post-injury seizures seem like straightforward potential biomarkers, the heterogeneity in the human population, in the injury itself, and in the type of clinical data collected has made it extremely challenging to pinpoint a specific metric useful in predicting PTE. More work is needed to develop these important clinical tools.

Many promising predictive biomarkers of PTE have also been identified in basic science studies and have promise as potential clinical biomarkers. For example, in the fluid-percussion model of PTE, MRI and FDG-PET imaging at 1 week and 3 months post-TBI time points were able to predict which animals would go onto develop PTE [43]. Multiple EEG biomarkers, which may eventually be useful in treating human PTE, have also been identified as predictive in animal models. These include shortening of sleep spindles [81] and the presence of pathological high frequency oscillations (HFOs) [82]. In these preclinical studies, the predictive power of these metrics was confirmed using robust statistical approaches. In the near future, we hope that these metrics can be investigated in relevant clinical populations to determine if their utility applies to human TBI/PTE.

Identifying and validating biomarkers of PTE is of critical importance for three main reasons. First, aggressive interventions for those at the highest risk of developing PTE could prevent the development of epilepsy. If those interventions have adverse effects, however, clinicians may hesitate to treat unless the risk of developing PTE is known to be high. Validated biomarkers can give clinical confidence, on a case-by-case basis, that using interventions are worthwhile based on patient risk. Second, it is extremely challenging and costly to carry out clinical trials for PTE prevention. Because only a subset of patients with TBI develops PTE, a significant number of patients must be used to conclusively ascertain if a treatment is effective. If a validated biomarker could show which patients were at high-risk, clinical studies could be enriched for patients with the greatest likelihood of developing PTE. This would dramatically reduce the cost and number of patients needed to test new PTE treatments. This is particularly important for mild TBI, where patients often do not seek immediate clinical care after injury, but are still at risk of developing PTE. Perhaps most importantly, biomarkers enable patients and care givers to understand their risk and plan accordingly. Clearly, the development and validation of a clinical biomarker for PTE would greatly impact the field and patients who suffer from TBI.

Genetic Modifiers of PTE

It is not well understood why some people develop PTE after TBI, while others do not. Many factors are likely at play including injury severity, inflammatory response, age at time of injury, time after injury, secondary “hits,” including other TBIs, stress, disruption of sleep, and many others. Of course, one must also consider that underlying genetic variations may contribute to the progression of secondary injury and manifestation of PTE. The effect may not be linear but rather depend, for example, on injury severity. A number of human and animal studies suggest that there is reason to believe that genetic variability that is not pathological under normal circumstances predisposes individuals to developing PTE.

Human Studies

While only preliminary analysis of genetic modifiers of PTE have been identified, a number of candidate gene variants have been suggested [89]. Methylenetetrahydrofolate reductase (MTHFR) is an enzyme involved in amino acid metabolism. The C677T MTHFR variant has been examined as genetic risk factor for epilepsy, may be over-represented in epilepsy patients, and is suggested to be linked to migraine and alcohol withdrawal seizures. In a study of 800 epileptic patients and 800 controls, the C677T variant was enriched in patients who had documented PTE [90]. While exciting, larger studies need to be done to determine if C677T MTHFR is strongly linked to PTE. Glutamic acid decarboxylase (GAD) is an enzyme critical to generating the inhibitory neurotransmitter GABA. Single-nucleotide polymorphisms (SNPs) in GAD1, one of the two GAD isoforms, were shown to be linked to post-traumatic seizures occurring shortly after TBI (< 1 week post-TBI) and PTE (1 week–6 months post-TBI) [91]. Again, these preliminary studies are intriguing but require further investigation to confirm their functional significance and strength of linkage to PTE. Changes in GABAergic inhibition are likely associated with PTE as GABAergic interneurons are lost after TBI [16, 92,93,94] and restoring GABAergic inhibition after TBI can prevent PTE [72, 95] in animal models. The adenosine A1 receptor (A1R) is a G-protein coupled receptor that is powerfully anti-convulsant and neuroprotective due to its ability to activate G-protein coupled inwardly rectifying K (GIRK) channels and inhibit presynaptic Ca2+ channels. In a study of over 200 patients with a severe TBI, SNPs in the A1R gene were linked to post-traumatic seizures occurring within one week following injury [96]. Again, A1Rs are closely linked to epilepsy as adenosine acting at A1Rs is thought to play a significant role in terminating seizures [97, 98] and genetic deletion of the A1R gene results in increased mortality in a rodent model of TBI [99]. Perhaps most interestingly, a SNP in the interleukin-1beta (IL-1b) gene, an inflammatory cytokine, has been shown to be linked to PTE risk over time [73]. The functional effects of this IL-1b SNP are unknown, but as mentioned above, the CSF/serum IL-1b ratio may also serve as a biomarker of PTE, strongly tying IL-1b to TBI/PTE. Finally, while not linked to PTE, there are several genetic polymorphisms, including APOE4, that are linked to poor outcomes after TBI [100, 101]. By combining genetic modifiers that affect TBI outcomes and early seizures, with those that affect PTE, we can build a more comprehensive understanding of how underlying genetic variation contributes to epileptogenesis. Properly powered clinical studies are critical to this goal.

Animal Studies

Only five studies have used genetically modified mice in PTE studies (Table 2). APP/PS1 mice showed increased prevalence of epilepsy [102]. Pijet et al. [103] showed increased prevalence of epilepsy, increased seizure frequency and susceptibility to PTZ-induced seizures in Mmp-9 over-expressing mice. Interestingly, deficiency in another extracellular matrix system (urokinase-type plasminogen activator) had minor if any effect on epileptogenesis [59, 60]. Adenosine A1R knockout mice show increased susceptibility to acute post-TBI seizures, but no longer-term risk of epileptogenesis was reported [99].

Overall, understanding the contribution of genetics on the evolution of epileptogenesis is at its infancy. Perhaps the most convincing evidence is the observation of higher incidence of epileptogenesis in CD1 mouse strain as compared to B6 strain [14] (Table 1). Even less is known about the contribution of genetics to therapy response. Further studies to replicate the current observations in larger study populations are warranted. It remains to be studies whether genetic markers can be used in stratification of patient populations in epileptogenesis clinical studies.

In Vitro Approaches to Understand Epileptogenesis and Develop Anti-epileptogenic Treatments

The development of PTE is likely driven by multiple factors which over time transform healthy brain networks in vivo into networks that generate seizures. Because of the in vivo nature of PTE, relatively few purely in vitro approaches exist to model epileptogenesis. One exception to this rule is the organotypic slice culture model of epilepsy developed by the Staley Lab. In this approach, hippocampal brain slices are prepared from neonatal rodents and cultured in vitro for weeks to months. Interestingly, over this time window, organotypic slices undergo dynamic changes in neuronal activities that eventually evolve into in vitro seizure-like events. In the first week of culture, only spike-like activity occurs. By 3 weeks in culture, > 50% of slices display prolonged, ictal-like activity [104, 105]. This evolution of ictal-like activity has been harnessed to investigate candidate mechanisms of epileptogenesis, like PI3K-Akt signaling [106, 107] as well as to perform drug screening aimed at identifying novel anti-convulsant and anti-epileptogenic therapies [108]. When combined with in vivo validation, this approach may hold promise as an alternative to purely in vivo–based drug discovery for anti-convulsant and anti-epileptogenic therapies.

As novel approaches are developed to grow cerebral organoids in vitro from induced pluripotent stem cells (iPSCs), new ways to model TBI and PTE will emerge. Intriguing recent work examines how contusional injury, similar to the controlled cortical impact model, affects neuronal viability, neurotransmission, and cell signaling in cerebral organoids [109]. This study shows that neurons die, that glutamate is released, and that pAKT and GSK3b signaling are reduced after physical injury in cerebral organoids. Whether cerebral organoids go onto develop something like PTE remains to be seen, but this is an exciting step towards developing additional in vitro assays for TBI and later PTE.

In addition to this purely in vitro approach, by coupling in vivo injury with in vitro assays, many studies have identified potential mechanisms of PTE and assayed the effects of various drugs. These include neuronal cell death, excitatory neuron sprouting, neuroinflammation, changes in metabolic activity, glial cell dysfunction, and more.

Conclusions

Recent successes in modeling various epileptogenic human brain injuries in rodents and larger animals [110], in identifying subjects at risk for PTE (high CSF/serum Il-1β, N3-REM spindles), and in pinpointing epileptogenic regions in the injured brain (HFOs) provide promise that novel approaches to identify and treat PTE are closer than ever before. Recent emphasis on clinically relevant study designs and outcome measures, combined with statistically powered preclinical multicenter studies can be expected to advance the field remarkably over the next years [111]. There is hope on the horizon.