Post-Traumatic Epilepsy and Comorbidities: Advanced Models, Molecular Mechanisms, Biomarkers, and Novel Therapeutic Interventions

A. Fluid Percussion Injury Model

Fluid percussion injury (FPI) is perhaps the most extensively used and studied model of PTE, largely due to its ability to easily modify the severity of injury, impact site, and species used. FPI can be applied centrally over the sagittal suture between bregma and lambda (midline FPI) or laterally over the parietal cortex (lateral FPI) and has been adapted for mouse, rat, rabbit, cat, and pig models. FPI produces a fluid pulse injury directly onto the surface of the dura after a craniectomy. The fluid percussion device consists of an adjustable pendulum that strikes the piston end of a fluid-filled cylinder, generating a fluid pulse that is transmitted along the horizontal axis of the cylinder onto the brain.

The model was initially described by McIntosh et al. (1989) in rats with a 4.8 mm craniectomy mid-distance between bregma and lambda and centered between the sagittal suture and lateral ridge. An injury cap is positioned over the craniectomy and secured with dental cement or glue to ensure the fluid bolus remained within the cranial cavity. The injury is produced by a pressure pulse lasting 20 milliseconds, and the severity of injury can be altered depending on the location and magnitude of pressure used (low 1.5 atm to high 3.5 atm in rats). Parasagittal and lateral FPI are common models for studying PTE. After FPI, electrodes can be implanted either immediately or after a short recovery period.

The FPI model can reproduce neuropathology associated with both diffuse or focal injuries as well as other aspects of human TBI, such as acute hypertension, bradycardia, increased plasma glucose levels, hemorrhages, inflammation, and cognitive deficits (Eakin et al., 2015). The damage of FPI is most severe in the ipsilateral cortex, hippocampus, and thalamus, although mild lesions and cell loss has also been detected contralaterally. Additionally, many laboratories have demonstrated the development of epileptogenesis, including a reduction in seizure threshold, presence of epileptiform discharges, and subgroups with SRS (Kharatishvili et al., 2006; Mukherjee et al., 2013; D’Ambrosio et al., 2004).

Seizure incidence after FPI has been inconsistent, with ranges from 0% to a cumulative probability of 100% epilepsy. A major limitation in many of the earlier reports is the lack of continuous long-term EEG recording. Random or intermittent epochs of recording do not allow researchers to determine latency to seizure onset, and in many cases, seizure occurrence will be underestimated or over-reported due to missing data. In one of the most cited reports of FPI-induced seizures, Kharatishvili et al. (2006) found approximately 50% of rats developed PTE over a period of 12 months when monitored with 24/7 video-EEG. Seizures were described as partial or secondarily generalized with a duration between 67–59 seconds; however, seizure frequency was quite low, averaging 0.3 ±0.2 seizures/d/animal. Neuronal loss was observed within the ipsilateral hippocampus, and aberrant mossy fiber sprouting was evident within the dentate gyrus. Similar data have now been collected from several laboratories (Zhang et al., 2008; Andrade et al., 2017; Mukherjee et al., 2013; Shultz et al., 2013; Carver et al., 2021).

B. Controlled Cortical Impact Model

Controlled cortical impact (CCI) has also been widely characterized since its first description by Lighthall in 1988 (Lighthall, 1988). Originating as a model of TBI, CCI has been adapted for studying PTE with the addition of EEG-recording electrodes and measurements of electrophysiological changes. To date, CCI techniques have been replicated in mice, rats, swine, monkeys, and ferrets (Xiong et al., 2013). This method requires general anesthesia of the subject as well as a craniectomy—similar to the FPI model. A computer-controlled impactor is used to deliver a unilateral strike to the intact dura directly at a velocity and depth specified by the researcher. This model can be customized further by choosing the diameter of the impact tip as well as whether the tip is blunt or rounded. Due to the levels of precision involved, CCI can mimic focal injury or widespread degeneration and can remove a potential source of error regarding the position and depth of the impact site. CCI studies have mimicked acute subdural hematoma, axonal injury, cell and tissue loss, blood-brain barrier disruption, and inflammation (Osier and Dixon, 2016). Additionally, many studies have described psychologic, functional, and cognitive changes associated with TBI (Kochanek et al., 2002; Watanabe et a., 2013; Yen et al., 2018).

Over time, long-term changes in the organization of neural circuitry due to trauma can lead to significant cell loss as well as an imbalance of excitatory and inhibitory neurotransmission. Seizures after CCI and FPI have been described in similar terms, both behaviorally and electrographically; however, CCI-induced spontaneous seizures appear to have a swifter onset compared with lateral FPI in rats (Kharatishvili et al., 2006). Previous studies have suggested limbic involvement may not appear for several months after FPI, resulting in a longer latency between initial injury and seizure occurrence (D’Ambrosio et al., 2004; 2005). This shorter, yet clinically relevant timeline, allows for greater flexibility in pharmacological testing.

A full spectrum of interictal activity including early (within the first week postinjury) or late seizures (after first week postinjury), isolated spikes, epileptiform discharges, absencelike spike wave discharges, and high-frequency oscillations (HFOs) have been successfully reproduced after moderate and severe CCI (Hunt et al., 2009; 2010; Statler et al., 2009; Golub and Reddy, 2022; Konduru et al., 2021). Sham mice do not show evidence of seizure activity, but high-frequency oscillations, interictal spike runs or “absencelike” spike-wave discharges have been observed in both injured and sham mice (Konduru et al., 2021). However, this study also pointed out that sham-injured rodents are not always appropriate controls for EEG studies since sham rodents have sometimes demonstrated interictal spiking or discharge activity. Computed tomography (CT) imaging revealed mild lesions compared with noncraniectomy mice, which could have contributed to abnormal activity.

Similar to FPI studies, seizure incidence after CCI is largely dependent on the severity of impact, ranging between 10% and 85% of mice exhibiting increased epileptic discharge spiking and seizures in the weeks after CCI (Hunt et al., 2010; Bolkvadze and Pitkänen, 2012; Golub and Reddy, 2022). Additionally, CCI studies have demonstrated progressive hyperexcitability in neocortical circuits within the first two weeks after TBI (Yang et al., 2010). These changes in electroencephalography are often accompanied with mossy fiber sprouting, hippocampal lesions, and changes in neurogenesis—all of which are defined as hallmarks of temporal lobe epilepsy (Hall et al., 2008; Hånell et al., 2010). Like FPI studies, CCI injuries provide consistency, reproducibility, and an overall accepted construct validity. This model also has the added benefit of easily altering injury severity, thereby changing the course of epileptogenesis and hippocampal pathology.

C. Weight-Drop or Impact-Acceleration Model

The impact-acceleration model, also known as the weight-drop model, simulates a diffuse injury to the brain and was first described by Mararmou and colleagues (Mararmou et al., 1994). After general anesthesia, rats or mice are placed below the weight-drop device on a foam block or platform intended to provide a consistent placement of the animal’s head and body relative to the apparatus. Skull fracture can be prevented by securing a small impact surface onto the location where the force will be applied, using cement or another adhesive. The impact-acceleration model can produce graded injuries, based on the heaviness and height at which the weight is dropped. This model can also be used as a form of repetitive TBI, producing a number of cumulative, but lesser injuries (Bailes et al., 2014).

The clear advantage to this model is the simplicity to induce trauma. The weight-drop apparatus itself can be constructed using inexpensive supplies and does not require a craniotomy within a stereotaxic rig, allowing for low cost and high throughput efficiency. Although this model leads to many known sequelae of human TBI, a critical downside is that weight-drop only produces PTE at extreme intensities, at which most animals do not survive. Furthermore, to identify and analyze abnormalities in EEG activity, such as post-traumatic seizures, typically an electrode is implanted at various focal sites, thereby still requiring access of a stereotaxic apparatus. Although this model is rarely used to identify seizure activity, it has been recently adapted to identify subtle changes in seizure threshold to chemical convulsant compounds, such as pilocarpine, without the need of recording electrode placement (Ben Shimon et al., 2020).

D. Blast Injury Model

Blast-related injuries are a frequent outcome of military exposure to explosive detonation. A leading hypothesis for the mechanism by which an explosive blast causes TBI is through the transmission of shock waves across the target tissue, causing rapid acceleration and deformation of the brain (Magnuson et al., 2012). These oscillating blast waves have sometimes been referred to as the “bobblehead effect” (Rosenfeld et al., 2013). Indirect transmission of kinetic energy from blast shock can travel through vasculature, playing a large role in TBI. Several models of blast TBI have been developed, although open-field blast, blast tubes, and shock tubes are the most frequently used (Kovacs et al., 2014).

Open-field blast occurs when an explosive detonation occurs within an outdoor open area, either by overhead suspension or placement on the ground. This is one of the most accurate representations of the human blast injury condition since subjects are located a standoff distance away from the blast. However, since debris and clouds from the primary explosion can contribute to the injury, blast tubes provide another method of blast injury in which a combined shock wave and blast wind is initiated by an explosion. In this case, animals are placed at the end of a pressurized blast tube, and the head-on explosion occurs at the opposite end of the tube. The torso and abdomen of the mouse are protected from exposure to prevent confounding injuries of the lungs, heart, and gastrointestinal tract. This method allows the rodent to be subjected to a “clean” blast without the reflection of shock fronts from the ground or other surfaces. Another advantage of blast tubes versus open-field blasts is that an equivalent blast intensity can be achieved with a much smaller explosive charge. Lastly, shock tubes use compressed gasses, such as helium, rather than explosives to achieve injury. Shock tubes are generally safer for both the subject and researcher to perform, more cost effective, and can be performed indoors—thereby not affected by weather conditions. However, the physics of shock tubes differs from that of explosive shock waves, which may not be as comparable to the human condition.

Blast injuries re-create several pathophysiological processes that likely play a role in the development of PTE, including intracranial hemorrhages, vasospasm, neuronal damage and degeneration, focal or diffuse axonal injury, and inflammatory reactions (Nakagawa et al., 2011). Although blast models have revealed neurofunctional changes, only one study has reported post-traumatic seizures and reduction in seizure threshold. Bugay et al. (2020) observed 46% incidence of spontaneous seizures in mice within a long-term study after up to three consecutive days of repetitive blast injury. Most seizures were electrographic, with little to no behavioral component observed. They also reported a shortened latency to spiking and hyperpolarization of action potential threshold in patch clamp recording of the hippocampus. These results produced a graded response to the number of blasts each mouse received (i.e., one, two, or three consecutive blasts). Although this is the first study to investigate TBI effects on neuronal excitability, the data clearly demonstrate increased risk of post-traumatic seizures as a measure of severity and repetitive injury.

E. Penetrating Ballistic-Like Brain Injury Model

The penetrating ballistic-like brain injury (PBBI) was designed to model two aspects of high-energy bullet wounds to the head: a large temporary cavity produced from energy dissipation and a permanent injury tract created by the path of the bullet (Williams et al., 2005). The PBBI itself is generated by inserting a custom probe into the brain at the desired location, creating the permanent injury tract, followed by the sudden inflation of an attached balloon to mimic the temporary cavity. The rodent is first placed into a stereotaxic device under anesthesia, scalp incised along the midline, and a small cranial window is created to allow the insertion of the PBBI probe. The probe can be constructed from a thin, 20-gauge stainless steel tube with spatially fixed perforations at one end. The perforations are sealed with airtight elastic tubing, which forms an inflatable balloon when an air pulse is delivered through the steel tube. The probe is then retracted, and the craniectomy is sealed with sterile bone wax or dental cement. Screw or depth electrodes can also be placed either before or after PBBI to obtain EEG recording. Since the relationship between the bullet’s impact velocity and the diameter of the cavity is linear, the parameters of this model can be altered to generate varied injury types or severities. However, this model of TBI mimics a specific and severe form of injury generally seen only in military populations.

One of the most notable consequences of PBBI is intracerebral hemorrhage, which is most common along the route of the probe entry. Additionally, PBBI re-creates pathologies found in the human condition of missile injury, such as extensive zones of radiating neurodegeneration, inflammation, neurologic impairments, edema, and post-traumatic seizures (Wei et al., 2010).

F. Cortical Undercut Model

The cortical undercut model of PTE was developed to mimic penetrating cortical injuries and has been used in both in vivo and ex vivo studies. Cortical undercut has been performed on rats, mice, and cats of varying age (Graber and Prince, 1999; 2006; Chauvette et al., 2016). Typically, rats are anesthetized and placed into a stereotaxic frame before unilateral exposure of the area of interest, typically the sensorimotor cortex. The dura mater and blood supply should be left intact after the craniotomy. A custom-made surgical knife or thin gauge needle, bent at a 90° angle 2mm from the tip is lowered to the white matter below cortical layer VI and rotated 180°, raised, then rotated back to its original position before being withdrawn to transect the underlying white matter. Attention should be paid to avoid any damage to major blood supplies. When a more complete undercut is needed, a second transcortical cut can be made with the needle without rotation (Graber and Prince, 1999; 2006). After a maturation period, animals are sacrificed, and coronal slices are obtained for electrophysiologic recordings and histology.

When preparing for an in vivo EEG study, depth electrodes can be placed in the surrounding or contralateral cortex, hippocampus, or other regions of interest. Array placement of electrodes is common in larger species, such as cats (Nita et al., 2006; Timofeev et al., 2013). Previous experiments demonstrate an immediate reaction to partial cortical deafferentation, resulting in significant reduction of local field potential amplitudes in regions above the transected white matter. Furthermore, 70% of animals experienced a shift in slow oscillatory activity to paroxysmal discharges (Topolnik et al., 2003a; 2003b). Seizure onset in this model evolves from slow oscillation and is characterized by the shortening of silence periods and increased amplitude of depolarization during active periods. Within a month after cortical undercut, electrographic paroxysmal activity spreads to regions distal to the initial transection and can be detected from corresponding electrodes. In mice, generalized seizures can be detected from the leading electrode, located just proximal to the undercut, within hours or days postinjury (Chauvette et al., 2016).

Cortical undercut is a valuable model for investigating the details of functional and structural alteration of neocortical GABAergic interneurons and pyramidal neurons occurring at the site of a focal injury without the spread of widespread damage and inflammation. However, this model is rarely used for pharmacological research due to its lack of translatability to the human condition. Clean transection of the white matter is rarely seen in human TBI without more extensive injury. Furthermore, there are few behavioral studies with this model, although available data indicate no significant motor deficits or cognitive impairment (Graber and Prince, 1999).

G. Repetitive Traumatic Brain Injury and Concussion Model

The majority of TBIs are mild in severity and are often underreported and, therefore, undertreated. Close-head impact injuries are a common cause of concussion and TBI. The consequences of repetitive traumatic brain injury (rTBI) and concussions have gained increasing attention with emerging reports of altered mood, behavior, and neurologic function. Concussions are extremely common in sports-related injuries, especially in contact-collision sports, such as boxing or American football, putting athletes at a higher risk for neurologic injuries, such as chronic traumatic encephalopathy and PTE (Mez et al., 2017).

The rTBI model mimics cellular and molecular changes induced by diffuse TBI, representing mild, concussive TBIs. Shandra and Robel (2020) recently published a detailed protocol for reproducing rTBI using a modified weight-drop model in mice. Briefly, an anesthetized mouse is placed in an induction chamber with its head positioned under the weight-drop tube on a foam pad. The pin on the weight-drop tube is released so that a 100 g weight is dropped from a height of 50 cm directly onto the scalp. A total of 2–4 weight drops are induced, with a 45-minute recovery period in between. EEG electrode implants can be placed via stereotaxic surgery either on the same day or the next day.

The rTBI model is characterized by a lack of focal lesion to the brain, loss of consciousness, high survivability, and late seizure onset (Shandra and Robel, 2020). Progressive tauopathy has also been observed in both experimental rTBI models and after repeated sports-related injuries (McKee et al., 2009; Tagge et al., 2018). Contrary to FPI or CCI models, the rTBI model does not require a scalp or cranial opening, reducing the risk of increased inflammation or infection.

H. Large Animal Models of Post-Traumatic Epilespy

Large animal species, such as pigs, are used in translational research because of their gyrencephalic neuroanatomy and significant white matter composition. However, one limitation of using these animals is that the laminar structure of the pig hippocampus has not been well characterized compared with that of a rat or mouse. Nevertheless, the Wolf group has described a porcine model of brain injury-related hyperexcitability and PTE and has been working toward elucidating the structure and characterization of the pig hippocampus (Ulyanova et al., 2018).

In the pig model of TBI, the primary neuropathological finding is diffuse axonal injury; however, hippocampal axonal and synaptic dysfunction as well as regional hyperexcitability have been observed, suggesting this model can be adapted for PTE (Meaney et al., 1995; Johnson et al., 2016; Wolf et al., 2017). Closed-head rotation induces a diffuse brain injury using a HYGE pneumatic actuator at controlled rotational acceleration levels to obtain the intended injury severity (Cullen et al., 2016). Briefly, the animal’s head is secured to a custom large-animal stereotaxic rig equipped with a padded snout clamp. The HYGE pneumatic actuator is mounted to the surgical rig using a custom linkage assembly that converts the linear motion to an angular motion. Rapid head rotation is performed within the coronal plane at velocities between 131 and 195 radians/s. Pathologies observed in the pig model of TBI include axonal shearing, tau accumulation, inflammation, and increased network excitability in the hippocampus (Smith et al., 1999; Johnson et al., 2016; Wolf et al., 2017; Grovola et al., 2020). Furthermore, motor and cognitive dysfunction has been documented in pigs after TBI (Friess et al., 2007).

A major downfall of larger animal models is the additional resources and time required to perform the appropriate surgical procedures and care. Therefore, pig studies have been inconsistently used due to their difficulty to implement and increased cost. Furthermore, induction of TBI in cats and dogs have been established but have not been widely used for studying PTE (Morganti-Kossmann et al., 2010). More recently, however, naturally occurring canine epilepsy has been proposed as a translational platform for novel therapeutics for epilepsy disorders (Davis et al., 2011). The prevalence and pathology of naturally occurring canine epilepsy are similar to the human condition (Löscher et al., 1985). In a study evaluating risk of seizures in dogs after head injury, patient records from 1343 diverse breeds were reviewed for previous head injury and recurrent seizures (Steinmetz et al., 2013). Of the 236 dogs with previous head injury, 18.6% exhibited early and/or late post-traumatic seizures. Observed seizure types included convulsive status epilepticus, partial and generalized tonic-clonic seizures, and cluster seizures. Although these data seem promising, the study has a few shortcomings, including difficulty obtaining enough questionnaires or telephone interviews to obtain statistical power. Retrospective studies such as these are not as feasible for the testing of therapeutic interventions.

The broad etiology after TBI presents a challenge for a singular paradigm to re-create all pathologies associated with PTE. The most common causes of human brain injury are car accidents, falls, recreational or sports injuries, and military incidents—all of which present differently within the clinic. Furthermore, acceleration-deceleration injuries differ from blunt force closed-head injuries or penetrating trauma in tearing, scar formation, and contusion (Dixon, 2017). Desirable features in animal models include a high frequency of epilepsy with an absence of extreme seizure clustering, low intersubject variability in seizure presentation, and a rapid and defined evolution of epileptogenesis. Ideally, this model would also be high throughput and low cost. Although the current animal models of PTE have provided much needed insight, no current model can fully recapitulate the full experience of human TBI. Therefore, it is important to understand the strengths and shortcomings of each model to determine which has the optimum conditions to evaluate specific research questions.

I. Translational Relevancy of Animal Models for Post-Traumatic Epilepsy

There has been much debate about the translational relevancy of animal models to the human condition for brain trauma research. As outlined above, numerous animal models have been developed to replicate various aspects of TBI and used for testing potential treatments. Although larger animals are closer in size and physiology to humans, rodents, such as rats and mice, are most commonly used in PTE research due to convenience research operations. The most widely used models include the CCI model, the FPI model, the weight drop-impact model, and the blast injury model. However, these models show intense negative effects, such as skull fracture, intracerebral hemorrhage, axonal injury, neuronal cell, and tissue death. Figure 1 outlines important considerations of translational relevancy between using small rodents to model the human condition. Like poststatus epilepticus models (Reddy and Kuruba, 2013), small animal models are most frequently used in preclinical post-TBI studies, with the aim to improve and develop better understanding of the recovery mechanisms and discover new biomarkers or clinical therapeutics. However, there are differences between small animals and humans as well as limitations to consider. Humans have 23 pairs of chromosomes, whereas rats and mice have 21 and 20, respectively. Although the known human genes associated with disease pathologies have corresponding orthologs in the rat and mice genome, their rates of synonymous substitution are different in the remaining genes. Furthermore, there are many differences in brain anatomy and complexity between small animals and humans. There are also some analogies, i.e., cerebrovascular parameters (Cernak, 2006), but the ratio of white:gray matter differs tremendously, making interpretation of behavioral alterations more challenging (Cordeiro and Horn, 2015). Furthermore, human TBI tends to be much more heterogenous in both injury location and cause (e.g., fall, car accident, sports injury, military events, etc.) than controlled laboratory experiments. However, there are many well-established neurologic and functional tests that can identify sensitive changes in recovery, cognition, or psychiatric function. These functional assessments are critical in comparing injury severity and outcomes not only between cohorts, but also other laboratory groups, similar to how the Glasgow coma scale is the gold standard for identifying injury severity upon clinic arrival. We discuss some of these neurologic and functional assessments in detail in section V of this manuscript, “Comorbidities of PTE.”

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1

Translational relevance of rodent to human TBI. Despite significant investment in advancing technology and basic science to increase knowledge of human TBI pathology, translation from bench-to-bedside into therapeutic advances has been slower than expected. One of the factors limiting the translation of scientific knowledge from preclinical studies into the clinic is the limitation of small rodent in vivo disease models. Although these models have been developed to simulate and mimic the human condition, there are innate differences between rodents and humans, which can limit the impact of these studies. Likewise, there are many important similarities as well as practical hints that can be used to overcome these limitations. This figure discusses important considerations of rodent to human translational relevance.

TBI is also associated with greater risk and frequency of neurodegenerative diseases, such as Alzheimer’s disease, but also enhanced risk of epileptic seizures. Temporally, seizures after brain injury have been categorized into immediate, early, and late seizures. Due to anesthesia protocols and difficulty in recording EEG directly after impact, immediate seizures in rodents are rarely identified; however, studies have reported early and late seizures after FPI or CCI-induced trauma (see sections II and VI). Animal models allow for experimental strategies in determining cellular and molecular interactions within the latency period and epileptic onset, thus provide a reasonable platform to develop new therapeutic interventions for PTE.

As trauma is frequently associated with damage of skin and soft tissues, differences of wound healing between rodents and humans should also be considered. The epidermis and dermis of small animals is thinner than in humans, which creates a challenge for wound suturing. However, wound clips provide a suitable alternative to classic suturing techniques, which is both time and cost effective. Moreover, rats and mice are at a lower risk for infection compared with humans due to faster wound healing processes and their ability to convert L-gluconogammalactone to vitamin C (Abdullahi et al., 2014).

TBI-induced coagulopathy manifests as disseminated intracranial hemorrhage, systemic bleeding, or intracerebral hematoma and is closely correlated to poor clinical outcomes and early onset seizures (Abdelmalik et al., 2016). An animal model using Sprague Dawley rats mimics this specific clinical scenario, and the study completed by Gangloff et al. (2018) suggests great similarity to human acute traumatic coagulopathy in terms of temporality, type of injury, compensatory mechanisms, and impairments in the coagulation systems. However, quantitative results in blood coagulation are not entirely transferable due to species-specific differences in clotting factors in the serum, such as coagulation factors (F)V, FII, FXII, and FXIII, which were all elevated in rats compared with healthy human donors (Karges et al., 1994). Additionally, coagulation factors FVIII, XI, X, and XI were all reduced in rats when compared with pooled human plasma. Other studies have identified decreased platelet count and reduced responsiveness to thrombin in small animals compared with humans (Derian et al., 1995). These factors may affect identification of biomarkers for TBI-induced pathologies.

Another caveat of rodent models is that animals are anesthetized at the time of injury, in comparison with humans, who are typically awake upon injury impact. Ethical reasons prevent a study in which the injury is given without anesthesia, and therefore the impact of anesthesia at time of injury on subsequent neuropathology and behavioral manifestations is unknown. Perhaps future studies can be designed to administer the trauma event in an unanesthetized animal to truly understand the impact of TBI, but this requires working with Animal Care and Use Committees locally or nationally. When planning and performing trauma-related experiments, the “3 R’s” of Reduction, Replacement, and Refinement, published by Russell and Burch in 1959, should always be considered and reflect both the complex pathophysiology and the immunologic alterations induced by the trauma event. Although there appear to be several differences between humans and small animals, there are also many processes in basic trauma and regeneration that have been conserved, making these models suitable for translational research.

A. Neuroinflammation

Local inflammation is intended as a beneficial protective measure after tissue insult; however, aberrant inflammatory responses can alter neuronal function and lead to serious consequences, such as BBB disruption and seizure development (Vezzani et al., 2013). Activated microglia and astrocytes play a large role in inflammation by releasing proinflammatory cytokines into the neuronal environment and promoting scar formation around tissue injury. Cytokine cascades in the brain regulate important pathways, such as neuroendocrine function, synaptic plasticity, metabolism of neurotransmitters, neurogenesis, and the kynurenine pathway (Paudel et al., 2018). These innate processes play significant roles in cell excitability and survival, thereby promoting network hyperexcitability. In particular, the interleukin (IL)-1/toll-like receptor (TLR) signaling pathway is disrupted and the associated receptors IL-1R1, TLR2, TLR3, and TLR4 are rapidly upregulated after both cell injury and seizures (Ravizza and Vezzani, 2006). The excitatory effects of IL-1β have been reported in several brain regions (Vezzani et al., 2011). IL-1β reduces GABA inhibition within the Cornu Ammonis area 3 (CA3) of the hippocampus and increases neuronal excitability in the CA1 by reducing N-methyl D-aspartate (NMDA) and voltage-gated calcium channel efflux (Zhang et al., 2010). Furthermore, lipopolysaccharide-induced inflammation is associated with reduction in seizure threshold in both postnatal and adult rodent models (Galic et al., 2009). This effect on seizure threshold can be reversed by blocking cytokine induction in activated microglia (Galic et al., 2009).

Induction of cyclooxygenase-2 (COX-2) has also been shown to promote epileptogenesis and contribute to neuronal damage in several animal models of epilepsy (Kulkarni and Dhir, 2009). Overexpression of COX-2 intensifies kainic-acid–induced seizures and mortality in mice (Kelley et al., 1999). Wei et al. (2018) confirmed COX-2 mRNA expression was significantly elevated after maximal electroshock. Although the modulation of the COX-2/prostaglandin E2 (PGE2) pathway has been pursued as an alternative therapeutic strategy for controlling seizures, careful study of COX-2 inhibitors could not fully prevent the appearance and development of spontaneous seizures in a rat model of status epilepticus (Holtman et al., 2010). Furthermore, inhibition of COX-2 has been found to either exacerbate or attenuate epilepsy-induced neurodegeneration, depending on the strategies used to interfere with the COX-2 pathway (Baik et al., 1999; Polascheck et al., 2010). These data highlight the ways in which the COX-2 pathway affects epileptogenesis, but mediation of this pathway alone is not sufficient for preventing seizure development.

An important consideration of inflammation in PTE is its contribution to progressive cell loss after injury. Free radicals and proteases accumulate during periods of inflammation, supporting lipid and protein peroxidation, DNA damage, mitochondrial dysfunction, and induction of apoptosis (Vezzani et al., 2013). Tissue damage, stress, and their subsequent cytokine release also adversely interfere with neurogenesis and neuroplasticity through their interactions with brain-derived neurotrophic factor (BDNF) and tropomyosin receptor kinase B (TrkB) receptor signaling (Goshen and Yirmiya, 2007; Ibrahim et al., 2016; Reddy et al., 2020). In a healthy brain, BDNF plays a crucial role in neuron maturation by regulating chloride levels and modifying inhibitory GABAergic signaling from depolarizing to hyperpolarizing (Rivera et al., 2002). However, within the context of injury, the upregulation of BDNF and its receptor TrkB are believed to promote aberrant mossy fiber sprouting (Dinocourt et al., 2006). Furthermore, brain injury causes a selective cluster of differentiation-74 (CD74)-dependent peripheral lymphocyte activation that may exacerbate neurodegeneration (Tobin et al., 2014).

Prolonged neuroinflammation also greatly affects quality of life and complicates comorbidities, giving cause for identifying therapeutics that explore the mechanistic association between PTE and neurobehavior dysfunction (Paudel et al., 2018). interferon-α can decrease BDNF levels, thereby slowing the rate of cell proliferation in the hippocampus and negatively affecting learning and memory consolidation (Lotrich et al., 2013). Additionally, increased cytokine production causes an imbalance of neurotransmitters, such as serotonin and dopamine, by deregulating the kynurenine pathway and disrupting neurotransmitter transport function (De la Garza and Asnis, 2003). Meta-analyses of existing research have concluded the most reliable biomarkers of inflammation in patients with depression are heightened levels of IL-6, tumor necrosis factor (TNF)-α, IL-1β, and C-reactive protein—all of which are significantly increased with TBI (Miller et al., 2009). Together, these inflammatory processes work in concert to promote depression, anxiety, cognitive impairment, and disturb sleep (Dantzer et al., 2008; Mukherjee et al., 2020).

B. Breakdown of the Blood-Brain Barrier

The blood-brain barrier (BBB) is a particularly important structure for central nervous system (CNS) homeostasis. There is increasing evidence demonstrating the BBB as a multifactorial pathophysiologic process involving faulty angiogenesis, neuroinflammation, altered glial physiology, leukocyte-endothelial interactions, and hemodynamic changes resulting in hyperexcitability (Marchi et al., 2012). Epilepsy disorders and TBI manifest with variable extent of BBB dysfunction; however, the link between BBB permeability and seizures has been posed as “the puzzle of the chicken and the egg” (Friedman, 2011). Acute vascular failure with BBB damage is sufficient to cause seizures in the absence of CNS pathologies or concomitant chemical convulsants (Marchi et al., 2007). Additionally, focal chronic seizures are frequent in patients with vascular malformations, such as cavernous angiomas (Kraemer and Awad, 1994). Magnetic resonance images of cavernous angiomas often present with BBB dysfunction, intracerebral deposits of iron, and albumin accumulation—all three of these factors have been identified as common features of TBI and temporal lobe epilepsy (van Vliet et al., 2007; Raabe et al., 2012).

BBB damage has been demonstrated to both trigger and sustain seizures in animal models and the human experience (Marchi et al., 2007; van Vliet et al., 2007; Raabe et al., 2012). Tomkins et al. (2008) observed greater association of BBB pathology in patients with PTE compared to patients with seizure-free TBI, suggesting a correlation between BBB breakdown and hyperexcitability. Areas of BBB disruption were linked to decreased brain glucose uptake, hypometabolism, and abnormal neuronal activity. After exposure of the cerebral cortex in rats, hypersynchronous epileptiform activity involving glutamatergic and GABAergic neurotransmission as well as significant endothelial tight junction impairment was observed (Seiffert et al., 2004). Accumulated albumin within the parenchyma is associated with downregulation of inward-rectifying potassium channels in astrocytes, affecting buffering capacity and contributing to hyperexcitability (Ivens et al., 2007). Moreover, loss of aquaporins expressed in the end feet of astrocytes affects water flux and potassium regulation, further disrupting the homeostatic environment of the brain (Binder and Steinhauser, 2006). Additionally, BBB damage could allow circulating levels of zinc to gain entry into the brain with devastating consequences, including excessive hyperexcitability and seizures (Carver et al., 2016; Chuang and Reddy, 2019). Zinc is an important neuromodulator, and its ability to persistently block extrasynaptic GABA-A receptors in the brain have dramatic consequences on epileptogenesis.

Neuroinflammation also plays a critical role in BBB permeability. Elevated levels of IL-1β, IL-6, and TNF-α can increase the permeability of the BBB and facilitate the movement of peripherally located cytokines into the CNS. These cytokines bind to receptors in the brain vasculature, producing secondary messengers and toxic by-products that further compromise its integrity (Fabene et al., 2010; Yarlagadda et al., 2009). Furthermore, these factors can trigger the activation of astrocytes and resident microglia, contributing to their dysfunction of neurotransmitter clearance and subsequent secretion of immunoregulatory markers. Systemic injection of lipopolysaccharide has been shown to lower seizure threshold to pentylenetetrazol, suggesting peripheral inflammation leads to a leaky BBB and possible infiltration of peripherally circulating leukocytes (Marchi et al., 2012). BBB dysfunction represents a convergence of pathogenic aspects that often create positive-feedback loops for further exacerbation of inflammation, functional impairment, and BBB permeability. For full review on how the breakdown of BBB affects PTE development, see Dadas and Janigro, 2019.

C. Epigenetic Modifications

Epigenetics refers to the plastic changes in gene expression that occur without alteration of the DNA sequence itself. Under normal conditions, epigenetic modifications are essential for growth, development, learning and memory, and the immune response (Hwang et al., 2017). Epigenetic modifications, such as DNA/Histone methylation, acetylation, and phosphorylation etc., have been implicated in a vast number of diseases, most notably cancer (Weber, 2010). Evidence suggests that epigenetic regulation of gene expression may play a critical role in the physiology of both epilepsy and TBI (Younus and Reddy, 2017; Nagalakshmi et al., 2018). Reddy et al. (2018a) demonstrated the histone deacetylase (HDAC) inhibitor, sodium butyrate, significantly slowed the kindling process in a mouse model of temporal lobe epilepsy when administered prior to electrical stimulation. This study suggests HDAC inhibitor compounds may possess antiseizure effects with an ability to curtail the process of epileptogenesis. Moreover, valproate has been administered as an antiseizure medication for decades, although its inhibitory effect on HDACs was unknown until 2001 (Göttlicher et al., 2001).

Histone modification is perhaps the most widely studied epigenetic modification in both epilepsy and TBI. Reduced H4 acetylation has been observed after pilocarpine administration at GluR/Gria2 promoter loci, a region that encodes for AMPA receptor subunits and limits calcium permeability (Huang et al., 2002). Downregulation of GluR/Gria2 is associated with hyperexcitability and initiating epileptogenesis. The same study also noted H4 acetylation at the BDNF promoter, increasing after seizure activity. H3 phosphorylation is thought to promote acetylation of histone proteins, and multiple studies have found striking evidence of H3 phosphorylation after pilocarpine and kainic-acid induced seizures (Crosio et al., 2003; Sng et al., 2006). Furthermore, hyperactivity of HDAC proteins occurs at early timepoints after lateral FPI (Zhang et al., 2008). This increased H3/H4 acetylation can be found throughout the hippocampus but is particularly visible in the CA3 (Gao et al., 2006). Increased HDAC activity leads to seizure susceptibility and post-traumatic epilepsy in both experimental models and in the clinical setting (Huang et al., 2012; Dash et al., 2009).

Changes in cell-specific global DNA/histone methylation have been shown to persist for up to 8 months post-TBI (Haghighi et al., 2015). Many of the affected genes have been associated with hyperexcitability, disruption of the sleep cycle, and neuropsychiatric disorders, such as Nos1, Il1r1, Homer1, Per3, and the Aanat gene, which encodes the enzyme responsible for catalyzing the serotonin to melatonin conversion (Haghighi et al., 2015). DNA methylation also plays a role in the inflammatory response to injury. Within 24 hours post-TBI, hypomethylation of microglia promotes active gene transcription in areas of widespread necrosis (Zhang et al., 2007). Furthermore, a study of patients with intractable temporal lobe epilepsy found expression of Dnmt1 and Dnmt3a were significantly higher in epileptic versus healthy controls, suggesting aberrant DNA methyltransferases may contribute to the pathogenesis of seizures (Zhu et al., 2012). DNA methyltransferase inhibitors have shown some promise for suppressing neuronal excitability in hippocampal neurons (Nelson et al., 2008; Levenson et al., 2006).

D. Reorganization of Neural Circuitry

The culmination of neuroinflammatory cascades, weakened BBB integrity, and epigenetic modification leads to consequent reorganization of neural circuitry through progressive cell loss, aberrant axonal sprouting, and neurogenesis. Several experimental models have highlighted the loss of inhibitory interneurons coupled with recurrent excitatory circuits as a basis for hypersynchronous epileptiform activity (Dudek and Spitz, 1997; McCormick and Contreras, 2001; Golub and Reddy, 2022). The hippocampus is a model system to study circuitry changes since it is particularly susceptible to injury and undergoes structural reorganization after TBI and in epilepsy disorders (Kharatishvili et al., 2006; Hunt et al., 2009;).

GABA_A receptors are responsible for the majority of inhibitory signaling in the brain. GABAergic interneurons form robust local synaptic connections with excitatory principal cells to control activity in two primary ways: phasic (synaptic) and tonic (extrasynaptic) inhibition (Farrant and Nusser, 2005; Chuang and Reddy, 2018). Phasic inhibition refers to the rapid transmission of information and activation of receptors at the synaptic junction after exposure to high concentrations of GABA released from presynaptic vesicles. Tonic inhibition, on the other hand, is mediated by extrasynaptic GABA_A receptors, persistently activated by low concentrations of ambient GABA. A common histopathologic feature of PTE is the drastic loss of inhibitory interneurons in the dentate gyrus and hilar regions (Lowenstein et al., 1992). Loss of these cells is correlated to an increase in tonic current amplitude in the dentate gyrus contralateral to TBI (Mtchedlishvili et al., 2010). Additional studies have reported changes in the subunit configuration of GABA_A receptors after CCI injury, which may also affect inhibitory control (Gupta et al., 2012; Raible et al., 2012).

Neurodegeneration after TBI affects both principal neurons and interneurons, although it was unclear whether one population is preferentially lost. Carron et al. (2020) examined how TBI affects different populations of interneurons, observing heterogenous changes of calbindin, parvalbumin, calretinin, neuropeptide Y, and somatostatin expressing interneurons in the hippocampus. Their findings suggest a differential vulnerability of interneurons across various brain regions and function after TBI. In a recent study, we performed a time-course of unbiased stereological quantification in the hippocampus contralateral to CCI at days 1, 3, 7, 30, 60, and 120. Populations of principal neurons and inhibitory PV+ GABAergic interneurons had decreased by roughly 30% and 45%, respectively, at 4 months postinjury; however, the degeneration in interneurons was accelerated to that of excitatory cells. Furthermore, the steep decline in interneurons coincided with the onset of spontaneous seizures (Golub and Reddy, 2022). Figure 3 highlights the linear regressions of cell loss during epileptogenesis and their temporal association to seizure onset. Our data agree with a previous study by the Hunt group that found a dramatic shift of interneuron diversity and loss after contusion injury (Frankowski et al., 2019). As previous studies have pointed to, the dentate gyrus and hilar regions showed the greatest loss of inhibitory interneurons (Hunt et al., 2011; Gupta et al., 2012). Loss of cells, whether excitatory or inhibitory, forces reorganization of these neural circuits, contributing the ongoing pathology of epileptogenesis. Restoration of the excitatory and inhibitory balance may be possible with transplantation of neuronal stem cells (Ngwenya et al., 2018). These animals also exhibited improved recovery and novel object recognition compared with nontransplant FPI animals.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3

Relationship of long-term neurodegeneration and spontaneous seizures. TBI induces a state of immediate inflammation and hyperexcitation in the brain, which exacerbate cell loss both ipsilateral and contralateral to the lesion. PTE was induced via a severe 2.0 mm depth CCI model of TBI. After injury, mice were tethered to 24/7-videoEEG for up to 4 months and seizures were identified by a customized MATLAB script and validated by unbiased researchers. Stereological quantification of two cell populations in the contralateral hippocampus was performed at days 1, 3, 7, 30, 60, and 120 post-TBI in subsets of these recorded mice. (A) Linear fit of remaining NeuN+ principal neurons overlayed the linear regression of average seizure output from responding mice to highlight the temporal relationship between cell loss and seizure occurrence. (B) Linear fit of remaining PV+ GABAergic interneurons overlayed the linear regression of average seizure output from responding mice to highlight the temporal relationship between cell loss and seizure occurrence.

Excitatory dentate granule cells are not typically connected to each other. However, several laboratories have demonstrated the reactive plasticity of these circuits in TBI and epilepsy models (Kharatishvili et al., 2006; Hunt et al., 2009; 2010; Bolkvadze and Pitkänen, 2012). Aberrant mossy fiber sprouting refers to the germination of axon collaterals from dentate granule cells into the inner molecular layer, forming functionally recurrent excitatory circuits. These local circuit changes are easy to detect with Timm’s immunohistochemistry and have been consistently reproduced in human and rodent tissues (Sutula et al., 1989; Hunt et al., 2010). Mossy fiber sprouting is generally more robust after severe versus mild TBI, suggesting degree of sprouting is correlated to both severity of injury and seizure risk (Hunt et al., 2012). Sprouting is most often noted in the hippocampus ipsilateral to TBI; however, damage is not constrained to the injured hemisphere and may influence circuitry reorganization with increased mossy fiber density (Pischiutta et al., 2018). Aberrant sprouting may provide a means for regional network synchronization, which is particularly vulnerable if basal inhibition is lost or impaired by cell loss or dysfunction.

Neural precursors proliferate in areas both proximal and distal to TBI impact. Much of this proliferation makes up the astrogliotic scar that forms around the injury site (Kernie et al., 2001). Changes in the rate of neurogenesis have also been found after TBI, and ectopic migration of these newborn cells may affect the excitability of the neural circuitry. TBI-induced newborn cells have increased dendritic branching proximal to the soma and wider dendritic reach that persists through cell maturity (Villasana et al., 2015). Neurogenesis has been a point of controversy in epileptogenesis, with some reports suggesting increased cell proliferation after TBI (Dash et al., 2001; Gao et al., 2009), whereas others observe reduced neurogenesis (Rola et al., 2006). However, differences in the rate of neurogenesis may be in part due to proximity to injury, timepoint of tissue sampling after TBI, or even the possibility of selective death of vulnerable newborn cells (Gao et al., 2008). Yu et al. (2008) observed both upregulation of type-1 quiescent progenitor cell activation in the injured hippocampus as well as progressive elimination of type-2 doublecortin-expressing progenitors. At the same time, the contralateral hippocampus also saw upregulation of type-1 progenitors, suggesting TBI may differentially impact damage versus compensatory signaling. Furthermore, severity of injury affects neurogenesis at different stages (Wang et al., 2016a). Regardless of location, severity, or timing after injury, fluctuation of cells born into the hilar and molecular layers of the dentate gyrus have been suggested to play a role in epileptogenesis (Danzer, 2019). Villasana et al. (2015) found TBI-induced newborn granule cells receive a normal balance of excitatory and inhibitory inputs and are involved in information processing, but suggested TBI-induced anatomic changes and dendritic projection patterns may be the root cause of maladaptive neurogenesis network properties. Insulinlike growth factor-1 overexpression has been found to increase the survival of newly born granule cells while inhibiting ectopic migration, the main implication of neurogenesis-associated circuitry changes (Carlson et al., 2014; Littlejohn et al., 2020). Therefore, conditional expression of astrocytic IGF-1 may be beneficial in reducing reactivity of astrocytes and preserving cognition after TBI.

E. Mammalian Target of Rapamycin Pathway Hyperactivity

The mammalian target of rapamycin (mTOR) pathway regulates several physiologic functions, and, in the brain, it is involved in cell proliferation and survival, neuronal morphology, and protein synthesis (Bockaert and Marin, 2015). Dysregulation of this pathway has been implicated in several brain disorders, including tuberous sclerosis complex, ganglioglioma, and focal cortical dysplasia—all of which may potentially or certainly lead to epilepsy (Liu et al., 2014a). Moreover, a role of mTOR signaling has been identified in brain trauma, although it is shrouded in controversy (Chen et al., 2007a). Some studies suggest the inhibition of mTOR, via administration of the mTOR inhibitor rapamycin, prevents neuronal injury and cell death after TBI (Erlich et al., 2007; Nikolaeva et al., 2016), whereas others suggest increasing mTOR signaling promotes greater recovery of function and regeneration, and this transient increase in mTOR signaling after TBI may be critical in stimulating neural stem cell proliferation (Wang et al., 2016b).

Hyperactivation of mTOR seems to play a critical role in the pathogenesis of acquired epilepsy, such as PTE, and rapamycin administration has prevented epileptogenic mechanisms and reduced seizure burden in certain models. In a rat hippocampal organotypic slice culture model of PTE, inhibition of Akt, PI3K, or mTOR reduced both ictal activity and cell death (Berdichevsky et al., 2013). In the clinical setting, rapamycin and its derivatives have been tested mainly on severe, refractory epilepsy disorders, such as tuberous sclerosis complex. Rapamycin and everolimus treatments improved seizure control in phase I/II studies (Krueger et al., 2013), and, in some cases, patients experienced complete cessation from previously intractable seizures (Perek-Polnik et al., 2012). The mechanisms by which mTOR inhibition reduces seizure activity in experimental models is still largely unclear but seems to point toward neurocircuitry reorganization. Inactivation of phosphatase and tensin homolog induces aberrant mossy fiber sprouting in animal and human hippocampal granule cells. Sutula and Dudek demonstrated that phosphatase and tensin homolog deletion was sufficient to trigger spontaneous seizures and that mTOR hyperactivation played a central role in this process (Sutula and Dudek, 2007). Furthermore, rapamycin inhibition of mTOR signaling reduces abnormal axonal sprouting and other pathologies associated with epileptogenesis, including neuronal excitability (Zeng et al., 2009). Although these data are positive, other studies have found rapamycin treatment reduces mossy fiber sprouting but has little effect on decreasing seizure frequency or duration (Buckmaster et al., 2009). Inhibitors of mTOR possess low efficacy in halting seizures within the preclinical models since multiple days of treatment are needed to achieve an antiseizure impact and beneficial effects typically cease after drug discontinuation.

When considering the mechanisms of epileptogenesis for prophylactic approaches, it is important to consider that both adaptive and maladaptive processes are activated by brain injury. The former can contribute to functional and beneficial recovery, whereas the latter may contribute to epileptogenesis. Questions emerge to determine whether these processes are separate or concurrent, if they are distinctly different forms of plasticity involved in either functional recovery or epileptogenesis, and if these processes are not distinct, do they differ quantitatively in their timing or intensity? Preventative or curative treatments may interfere with both epileptogenesis and development of comorbidities. It is likely that multiple agents will be needed to provide full spectrum symptomatic relief to patients. The answers to these questions are critical to pharmacological progress in epilepsy and head trauma.

A. Imaging Biomarkers

Imaging within the context of TBI has thus far focused largely on primary lesion formation and the evolution of gliosis; however, imaging techniques have the potential to capture progression of pathologies of PTE. Neuroimaging biomarkers are incredibly appealing not only due to their noninvasive procedures but also because they are routinely performed as part of a patient’s routine medical care. Furthermore, imaging can detect patterns across specific structures or within the whole brain (Reddy et al., 2019).

CT scans have been used for decades to assess global structural damage after TBI (D’Alessandro et al., 1988). Characterization of early CT scans have identified an increased risk of PTE development in patients with depressed skull fracture, dural penetration, and intraparenchymal, subdural, or epidural hemorrhage. Additionally, patients with cortical/subcortical contusions or large lesions in the temporal lobe have shown higher rates of PTE incidence, regardless of injury severity (Englander et al., 2003; Tubi et al., 2019).

Positron emission tomography (PET) allows researchers to visualize inflammatory responses and the metabolic impact of neuronal injury after TBI. 18F-Fluorodeoxyglucose (FDG) is a radioactive tracer of cerebral metabolism often used in PET experiments. Hypometabolism has been observed within the first 24 hours after kainic acid-induced status epilepticus, suggesting abnormal metabolism may play a role in epileptogenesis (Jupp et al., 2012). Conversely, FDG-PET scans of patients with severe TBI demonstrate hyperglycolysis occurs up to 2 weeks after initial damage (Bergsneider et al., 1997). This increase in glucose consumption may be due to the rise in inflammatory cell populations surrounding the impact site.

Gliosis appears bright on T2-weighted MRI and fluid attenuated inversion recovery techniques. Inflammation and glial scarring are extremely common around the impact site after TBI in both rodent models and in the clinical setting (Dixon, 2017). Activation of astrocytes and microglia through maladaptive inflammatory cascades have been correlated to excitotoxicity, mitochondria dysfunction, and cell loss—all of which have also been indicated as epileptogenic factors (Alyu and Dikmen, 2017; Shandra et al., 2019). Magnetic resonance spectroscopy is also a useful tool for providing quantitative data on altered metabolite profiles related to inflammation and hyperexcitability. Reductions in glucose and GABA neurotransmitter, as well as increased glutamate have been observed after TBI using magnetic resonance spectroscopy (Friedman et al., 1999). To further assess metabolite profiles as a function of epileptogenesis, Filibian et al. (2012) investigated glutathione levels, an antioxidant produced by activated astrocytes, in rats with pilocarpine-induced status epilepticus. Glutathione was negatively correlated with seizure frequency, providing strong evidence of astrocytic involvement in seizure generation.

MRI has excellent resolution and tissue classification, using strong magnetic fields to form images of the anatomic and physiologic processes of given organs (Reddy et al., 2019). Structural MRI studies have revealed minor, but significant, changes in the ipsilateral hippocampus relative to baseline in rats experiencing post-traumatic seizures versus nonepileptic rats one week after lateral FPI (Shultz et al., 2014). The same study boasted high predictability of PTE incidence using a multivariate logistic regression model of serial FDG-PET parameters among all injured and noninjured cohorts. Similarly, hyperexcitability in the hippocampus has been correlated to early 3-hour postinjury decreases and later 1- to 12-month postinjury increases in hippocampus diffusion (Kharatishvili et al., 2007). Furthermore, mossy fiber sprouting scores, which have been long considered a hallmark of epilepsy, were correlated with diffusion values after lateral FPI, providing greater evidence that quantitative diffusion MRI is a potential tool for facilitating the prediction of increased seizure risk after TBI. In a follow-up study, the same group accurately predicted seizure susceptibility to pentylenetetrazol (PTZ)-administration using similar MRI diffusion patterns, with the greatest accuracy to be found with a combinational biomarker calculated from diffusion in the ipsilateral somatosensory cortex and thalamic regions at 2 months post-TBI (Immonen et al., 2013).

Functional MRI can be used to analyze connectivity, plasticity, and remodeling within the brain after injury. There is extensive reporting on functional changes after TBI and clinical outcome, but studies focusing on epileptogenesis and PTE are scarce and inconsistent. Evidence of both hypoconnectivity and hyperconnectivity have been observed in several networks after TBI, and these abnormalities have been linked to changes in behavior, cognitive impairment, and motor control (Hillary et al., 2011; Tang et al., 2011; Stevens et al., 2012; Palacios et al., 2013). For example, abnormal frontoparietal network connectivity after mild TBI affected performance in working memory tasks, suggesting that interhemispheric connectivity between left and right inferior frontal gyri may contribute to learning and memory impairments seen in the clinical setting (Kasahara et al., 2010). A single experimental study has investigated functional MRI biomarkers in relation to epileptogenesis using lateral FPI and PTZ susceptibility as a model. Mishra et al. (2014) observed decreased connectivity between the ipsilateral and contralateral parietal cortex and between the hippocampus and parietal cortex in the injured hemisphere compared with sham-operated rats. However, no significant relationship was found between functional connectivity and seizure susceptibility during the PTZ test.

Blood-brain barrier (BBB) dysfunction has long been suggested to play a key role in seizure susceptibility after TBI and in epilepsy disorders (Cornford and Oldendorf, 1986). Recent studies demonstrate that abnormalities in the BBB can be visualized using dynamic contrast-enhanced MRI and fluid attenuated inversion recovery techniques. Disrupted BBB function has been found with increased frequency and to a larger extent in patients with post-traumatic seizures versus nonepileptic patients with TBI (Tomkins et al., 2011). Furthermore, reduced BBB integrity within the cortical regions surrounding the impact site or within the piriform network have been suggested as sensitive predictors of epilepsy (Bar-Klein et al., 2017). There are several hypotheses that associate cerebrovascular permeability with epileptogenesis, including imbalance of ion and molecule distribution and disturbance of neuronal homeostasis (Dadas and Janigro, 2019).

Two of the biggest challenges regarding imaging biomarkers is the heterogeneity of PTE-related injuries and the rarity of longitudinal studies. Thus, a consistent and validated imaging biomarker for PTE has yet to be discovered. For a full review of imaging biomarkers in PTE, see Garner et al. (2019) and Immonen et al. (2019).

B. Electrographic Biomarkers

Electrographic biomarkers may predict the onset of seizures and epileptogenesis, allowing for development of targeted preventative therapies. Currently, there are no validated electrophysiological biomarkers for PTE; however, experimental EEG studies using lateral FPI and CCI in rodents have identified potential candidates, including pathologic high-frequency oscillations (HFOs), reduction in sleep spindle duration, changes in theta oscillations, dominant frequency at the stage III to rapid eye movement sleep, and epileptiform spiking/discharges preceding seizure onset.

HFOs are commonly classified into ripples (80–250 Hz) and fast ripples (250–500 Hz) and are believed to be a naturally occurring phenomena involved in both physiologic and pathologic processes (Zijlmans et al., 2012). Although physiologic and pathologic HFOs cannot be differentiated on spectral frequency alone, increased rhythmic patterns and power of HFOs have been correlated to epileptic foci (Jirsch et al., 2006; Staba, 2012). Early investigators of HFOs questioned whether HFOs could be evaluated by standard invasive macroelectrodes, but recent reports have shown that HFOs can be detected for long periods of time by a wide variety of methods, including the standardized scalp EEG (Andrade-Valenca et al., 2011).

The premise that HFOs may play a role in epileptogenesis was first hypothesized by Bragin et al. (2000) who observed fast ripple occurrence preceded spontaneous seizures by weeks in a kainic acid model of status epilepticus. A follow-up study found that 19 out of 26 rats developed pathologic HFOs in the first month postkainic acid injection and that all 19 rats later developed epilepsy, whereas the remaining seven rats did not exhibit pathologic HFOs or seizures (Bragin et al., 2004). In a lateral FPI study, repetitive HFOs occurred at a significantly higher rate after moderate or severe injury, compared with mild or sham-injured animals, providing additional evidence that injury severity correlates to epilepsy (Reid et al., 2016). Among these FPI rats, HFOs tended to occur more frequently in the early stages of the study before seizure onset. These studies suggest a pattern of intense and increasing abnormal epileptic activity during the epileptogenic period, starting with increased HFOs and ending the latency period with spontaneous seizures (Fig. 4). Nevertheless, alterations in characteristics of HFOs in people at risk for epilepsy or epileptogenesis may serve as valuable noninvasive biomarker of epilepsy or risk prediction. For full review of HFOs as biomarkers for epilepsy, see Perucca et al. (2019) and Jacobs and Zijlmans (2020).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4

Evolution of TBI-induced hyperexcitability and seizure activity. Electrographic biomarkers may predict onset of seizures, as hyperactivity in the brain progresses over time to the culmination of spontaneous recurrent seizures. TBI induces a state of heavy inflammation, disrupting both neurotransmitter and metabolic homeostasis. The emergence of these abnormal electrographic activities may reflect different stages of the epileptogenic process postinjury. Pathologic HFOs often precede seizures by weeks, followed by reduced frequency and duration of sleep spindles during the transition between stage 3 and REM sleep. Disruption of normal sleep spindles contributes to several sleep-wake disorders reported by patients with TBI. EEG spiking and discharges represent an advanced hyperactive disturbance that has been described as epileptiform abnormalities in animal brain slices and in vivo at various time-points postinjury. The final stage is the end of latency indicated by the occurrence of spontaneous seizures; however, epileptogenesis can continue to progress even after the first seizure.

Sleep spindles occur at a frequency between 10–20 Hz (typically 12–14 Hz) with a duration of 500–2000 milliseconds. Spindle generation involves interaction between inhibitory neurons in the thalamic regions to function as pacemakers. An increase in spindle frequency is known to occur immediately preceding REM sleep in both rodents and humans (Vyazovskiy et al., 2004; Purcell et al., 2017). Sleep disturbances are well documented after TBI, and disruption of normal spindle activity may contribute to epileptogenesis (Duclos et al., 2014). A study by Andrade and colleagues discovered 92% of detected seizures occurred during the transition period between stage III and REM sleep after lateral FPI in rats (Andrade et al., 2017). Sleep spindles in these epileptic rats were significantly shorter and slower during the transition from slow-wave N3 to REM sleep, compared with nonepileptic rats. These changes were identified at 9 weeks postinjury but were not observed within the first 2–3 weeks (Bragin et al., 2016), providing greater evidence of the progression of abnormal EEG over time. However, these results somewhat contradict those of a more recent study in which Konduru et al. (2021) confirm shorter spindle duration and lower band frequency in injured versus noncraniectomy mice but did not find significant differences in these features between TBI mice with and without a seizure response. Furthermore, sham-injured animals have been known to exhibit abnormal EEG activity compared with noncraniectomy groups, suggesting sham cohorts may not always be the best control for all parameters.

EEG spikes, sometimes called interictal spiking, have been observed in both rats and mice after brain injury (Statler et al., 2009; Bolkvadze and Pitkänen, 2012). These spikes can take the form of isolated spikes, runs of spike-wave discharges, or absence-like spike-wave discharges and represent abnormal fluctuations in brain waves but have yet to be correlated with the onset of SRS. Furthermore, stimulation-evoked hyperexcitability in the hippocampus has been demonstrated after weight drop and in neocortical brain slices after CCI (Golarai et al., 2001; Yang et al., 2010). It is unclear whether these electrophysiological disturbances are associated directly with the injury state or with epileptogenesis. It should be noted that not all studies show a correlation between animals with spiking activity and documented seizures, although recordings in this case were short 1-week recordings (Konduru et al., 2021). Recently, Golub and Reddy (2022) observed short bursts of high-energy activity lasting between 3 and 9 seconds after CCI injury in mice. Although this definition of epileptiform discharges is intermediary to interictal spikes and seizures, it does provide further evidence of the evolution of abnormal EEG activity to seizure development.

Although there are currently no validated electrophysiological biomarkers for PTE, experimental studies continue to identify potential candidates that may be instrumental in predicting and, therefore, preventing epilepsy with targeted therapeutic approach. These EEG abnormalities include pathologic or repetitive HFOs, sleep spindle disturbances, and abnormal interictal spiking. Further experimental studies using a controlled means of EEG analytics are needed to determine which features may be reliable biomarkers of PTE in the clinical setting. Machine learning, which has been recently instituted for seizure detection (Abbasi and Goldenholz, 2019), may also help to classify these abnormal EEG patterns in more meaningful ways. Furthermore, standardization of detection protocols, analysis algorithms, and sampling of EEG recordings will aid the progress of neurophysiologic biomarker discovery.

C. Molecular Biomarkers

Biofluid markers are useful for determining TBI severity and play a critical role in monitoring disease progression and clinical prognosis (Sharma and Laskowitz, 2012). Molecular biomarkers refer to nonimaging factors that have biophysical properties allowing for measurement in biologic samples, such as blood, plasma, cerebrospinal fluid (CSF), saliva, or biopsy. Circulating biofluid markers, such as microRNAs (miRNAs), proteins, extracellular vesicles, and cytokines have been widely studied in both TBI and epilepsy conditions (see reviews Engel et al., 2013; Agoston et al., 2017), but few studies have combined their efforts to identify biomarkers for epileptogenesis and PTE. Furthermore, there is little evidence to suggest biomarkers for TBI overlap with those for epileptogenesis; therefore, this subsection will briefly cover molecular biomarkers of TBI that have been indicated as potential risk factors for seizures.

Severe and penetrating TBI represent the highest risk of PTE development due to the extent of tissue damage, bleeding, inflammation, and bone fracture. These processes produce a graded increase in circulating levels of inflammatory chemo- and cytokines, regulators, and bone morphogenic proteins, providing a molecular basis for classifying injury severity (Heggeness et al., 2017). Deposition of bone particles and other foreign bodies in the brain parenchyma are among the most important risk factors for developing seizures after TBI (Salazar and Grafman, 2015). Likewise, increased serum and/or CSF levels of claudin-5, VEGF, occludin, aquaporin-4, and von Willebrand factor may indicate a breakdown of the BBB and/or vascular injury (Croll et al., 2004; De Oliveira et al., 2007; Jiao et al., 2011; Thrane et al., 2011; Ahmed et al., 2015).

Iron levels, due to excessive bleeding, increase risk of PTE (Ding et al., 2016). Accumulation of iron in the blood can be cytotoxic, resulting in mitochondrial dysfunction and oxidative stress through the generation of free radical particles. In the clinical setting, patients with TBI and low ceruloplasmin, an important protein involved in iron metabolism and injury repair, develop increased intracranial pressure that can lead to post-traumatic seizures (Dash et al., 2010a). Neuroinflammation around the impact site also contributes to the rise in intracranial pressure. Increased cytokines and inflammatory proteins, such as IL-1, IL-6, TNFα, CD53, fibrinogen, and MIP1α, have been implemented in prolonged inflammation in human and animal studies (McManus et al., 1998; Katayama et al., 2009; Woodcock and Morganti-Kossmann, 2013). A genetic biomarker study demonstrated higher CSF:serum IL-1β ratios were associated with increased PTE incidence (Diamond et al., 2015a). Furthermore, due to the astrocytic roles in both inflammation and glucose metabolism, CSF and/or serum levels of glial fibrillary acidic protein (GFAP) can provide insight into more than one pathology associated with PTE. It is likely that a combination of disrupted neural connectivity due to cell loss, metabolic dysregulation, and inflammation are critical components of immediate and early seizure onset after TBI (Tubi et al., 2019).

MiRNAs are noncoding RNA molecules that have emerged as potential molecular biomarkers for a number of neurologic disorders. miRNAs are considered good biomarker candidates since they are more stable than mRNA and proteins, are present in biofluids, such as blood and CSF, and are inexpensive to assay. The human genome contains approximately 2300 miRNAs, and about 500–700 of these are also present in laboratory rodents, such as mice and rats (Alles et al., 2019). In plasma, miRNA can be either bound to argonaute2 protein or held within extracellular vesicles. Extracellular vesicles are involved in intercellular communication and can carry biomolecules such as DNA, mRNA, miRNA, proteins, and lipids. Raoof et al. (2017) conducted a study including patients with temporal lobe epilepsy and status epilepticus that identified miR-19-3p was largely argonaute2-bound in both epileptic conditions, and miR-21-5p was mostly carried within extracellular vesicles in status. A follow-up study discovered the proportion of argoaute2-bound miR-328-3p increases after a spontaneous seizure in temporal lobe epilepsy (Raoof et al., 2018). Lastly, patient levels of transfer RNA fragments 5′GluCTC, 5′AlaTGC, and 5′GlyGCC have been found to be upregulated in preseizure versus postseizure samples (Hogg et al., 2019). Since extracellular vesicles can be extracted from all biologic fluids, they have exciting potential for identifying biomarkers of post-traumatic epileptogenesis. However, most clinical research in this area have so far investigated miRNA levels in patients with pre-existing epilepsy, rather than conditions with an epileptogenic trigger.

Dysregulation of miRNAs has been observed in both patients with epilepsy and TBI; however, no study has been done to identify biofluid miRNA biomarkers of epileptogenesis in humans. There is little overlap between TBI and epilepsy in potential miRNA candidates, and previous studies have presented a high frequency of contradictory data regarding serum and plasma miRNA dysregulation and their usability in disease trajectory for PTE. Table 3 shows a list of potential miRNA biomarker candidates and whether they have been shown to be upregulated or downregulated in epilepsy, TBI, and/or PTE. Data from this table are mixed between preclinical and clinical studies, as there is limited knowledge in this field at present (Simonato et al., 2021). miRNAs, such as miR-21-5p, miR-27b-3p, miR-93, miR-135a, miR-146a, miR-155, miR-203, and miR-451, have all been shown to be upregulated in various epilepsy conditions as well as after brain injury, whereas miR-27a-3p, miR-128, and miR-221-3p are downregulated (Redell et al., 2010; Gorter et al., 2014; Liu et al., 2014b; Roncon et a., 2015; Atif and Hicks, 2019; Brennan et al., 2020). These data hold promise for PTE biomarker discovery in that there seems to be some similarity between these two neurologic conditions. However, there are many other candidates with opposing dysregulation, such as miR-23a, miR-30a-5p, miR-153, miR-182, miR-219, miR-300, miR-328-3p, and miR-574. We are currently at the beginning of the process to adequately validate the first biomarkers for epileptogenesis, and more time and effort will need to be dedicated toward discovery before their subsequent clinical use.

The first step in establishing plasma miRNA or protein biomarker discovery is to harmonize protocols and procedures used to collect data and perform experiments, providing an objective framework for quality control. Recently, there has been a great effort across three international sites (University of Eastern Finland, University of Melbourne/Monash University, and the University of California, Los Angeles) to standardize the protocols for miRNA biomarker validation and analysis for PTE (van Vliet et al., 2017; Kamnaksh et al., 2019). These studies highlight the need for rigorous quality assessment, as hemolysis and anesthesia presented as confounding factors. Moreover, the study suggests improved training is needed for technicians to obtain a more precise venipuncture, faster blood draws, and less coagulation in the catheter lines.

D. Genetic and Genomic Influences

A genetic biomarker is a sequence of DNA that causes or is associated with susceptibility of a disease. Genetic biomarkers typically present as a genetic variant, such as a copy number variant or single-nucleotide polymorphism in the clinical setting. Likewise, genomic biomarkers reflect the expression, function, and regulation of a gene and its interrelationships to identify a combined influence within a biologic state or within the context of disease progression (WHO, 2002). Genomic biomarkers include both DNA and RNA characteristics. DNA characteristics include single nucleotide polymorphisms, DNA modifications (e.g., methylation or acetylation), variability within short sequence repeats, insertions, deletions, copy number variations, or cytogenetic rearrangements (e.g., duplications, deletions, inversions, translocations). Characteristics of potential RNA biomarkers include changes in RNA expression levels and sequence, alteration of RNA processes, such as splicing or editing, and variation of microRNA levels. These two terms “genetic” and “genomic” biomarkers are similar in that they represent a change at the level of DNA or RNA but should not be used interchangeably. The main difference between genetic and genomic biomarkers is their respective focus on a specific gene and heredity versus an organism’s entire genetic make-up including coding and noncoding DNA and their interactions. For example, Huntington’s disease is caused by a genetic mutation in the HTT gene, representing a genetic biomarker; whereas, significantly altered mRNAs found in the peripheral blood of patients with Huntington’s serve as a genomic biomarker (Borovecki et al., 2005). Genomic biomarkers could be a useful prognostic marker for PTE by identifying individuals with a higher risk of epileptogenesis and enriching the population of antiepileptogenesis trials.

Preclinical rodent trials have included genetically modified mice and/or animals with varying genetic backgrounds to help identify genomic influence on epileptogenesis and TBI. Adenosine is known to exhibit some anticonvulsant effects within epilepsy (Knutsen and Murray, 1997; Avsar and Empson, 2004). Knockout of the Adora1 gene, affecting the adenosine receptor A1, was found to increase the incidence of acute postimpact status epilepticus (Kochanek et al., 2006). Adenosine receptor deficiency also exacerbated the microglial and neuronal damage response after TBI (Haselkorn et al., 2010). In the clinical setting, variants of adenosine kinase influenced the rate of epileptogenesis after TBI (Diamond et al., 2015b). This same study also identified the genotype rs1143634, a variation in the IL-1 proinflammatory cytokine profile, increased risk for developing seizures.

Several studies have focused on genes related to plasticity of the extracellular matrix. Pijet et al. (2018) used matrix metallopeptidase 9 (MMP-9) knockout and overexpression mice to determine the role of extracellular matrix restructuring in post-traumatic epileptogenesis. Two peaks of MMP-9 expression were found at 30 minutes and 6 hours postinjury. Their results demonstrate overexpression of MMP-9 resulted in greater seizure frequency and lowered PTZ seizure threshold after TBI. Prevalence of post-traumatic seizures was also correlated to increased lesion volume in these mice. APP/PS1 mice, which are predisposed to plaque deposition and gliosis, were found to have more pronounced epileptogenesis and robust comorbidities, such as cognitive impairment, after TBI (Miszczuk et al., 2016). A series of studies investigated the role of the extracellular matrix proteinase urokinase-type plasminogen activator and its receptor on PTE. Neither mutation of the Plua or Pluar gene, resulting in the deficiency of urokinase or its receptor, affected the progression of PTE after TBI. The authors concluded that epileptogenesis and seizure susceptibility was not worsened with urokinase-type plasminogen activator or its receptor deficiency, although comorbidities, such as cognitive impairment and motor function, were exacerbated (Bolkvadze et al., 2016). However, a recent follow-up study using a double knockout of both Plau and Plaur genes found significantly increased susceptibility to PTZ without brain injury (Kyyriäinen et al., 2019).

In the clinical setting, mutations affecting the balance of inhibitory-excitatory circuitry have been linked with an increased risk of PTE. Variations in the GAD1 gene, responsible for producing the enzyme that catalyzes the production of inhibitory GABA from glutamate, increases a patient’s risk for exhibiting post-traumatic seizures. Three high-risk genotypes have been identified for higher risk thus far: rs769391, rs3791878, and rs3828275 (Darrah et al., 2013). Genetic variation in the SLC1A1 gene responsible for neuronal glutamate transporters has also been associated with an increased risk for post-traumatic seizures and excitotoxicity (Ritter et al., 2016). Moreover, a major neuroprotective and inhibitory molecule, adenosine, has become increasingly important in identifying risk of PTE. Multivariate analysis of rs3766553 revealed a strong link between variability in the adenosine A1 receptor and increased risk of late seizures after injury (Wagner et al., 2010). Furthermore, a military-cohort study found a common variant in the methylenetetrahydrofolate enzyme C677T may predispose an individual to PTE and other epilepsy disorders (Scher et al., 2011). Lastly, the role of APOPE gene has also been studied within the context of clinical PTE. Although no statistically significant associations were found, Miller et al. (2010) found half of the individuals with the E4/E4 genotype of APOPE had exhibited chronic post-traumatic seizures, suggesting this variant may be at greater risk for delayed PTE. The lack of studies reporting the effect of genetic and genomic factors on epileptogenesis after TBI limits the progress to be made in determining targeted measures for therapeutics and clinical trials (Kumar et al., 2019). There is a great need to continue research on these mechanisms to contribute to the progress in preventative therapies for PTE.

Discovery for PTE biomarkers is at an early stage. Each potential biomarker discussed has both advantages and disadvantages; therefore, it is unreasonable to expect a single biomarker to measure the progression of a heterogenous disease, such as PTE. It is more likely that a combination of multimodal biomarkers will be needed to accurately identify epileptogenesis after brain injury. Relatively noninvasive biomarkers, such as imagining, surface EEG, and blood/plasma samples, are promising for patients with TBI in which the likelihood for seizures is uncertain. However, there is little evidence to suggest that biomarkers for epileptogenesis after brain injury overlap with those of TBI in general. These differences may be exaggerated by variance in models, detecting methods, analysis platforms used, and stage of epileptogenesis. Furthermore, translating preclinical studies into clinical biomarkers is challenging. The timeline of molecular and cellular changes that occur in the development of epilepsy is much shorter in rodents than in humans, explaining why dozens of successful experimental pharmacotherapies have failed in clinical trials (Agoston et al., 2019). Longitudinal data in both rodent and clinical studies are lacking; therefore, no standard temporal window comparing imaging pathology or sampling has been followed. A concerted effort to standardize biomarker efforts for TBI, PTE, and other epilepsies may be an optimal strategy for discovery of novel biomarkers that are translatable throughout these related disorders.

A. Sensorimotor Abnormalities

Risk of complications increases with the severity of trauma, although mild TBI can also result in disabilities that interfere with daily life (van der Naalt, 2001). Patients may experience sensory problems, especially complications with vision (Ripley and Politzer, 2010). One of the brain’s primary functions is to integrate information from the outside environment, process it, and determine an appropriate reaction or response. Disturbances in vision, either in recognition or registration of objects, can lead to clumsiness and poor hand-eye coordination. Double vision also affects depth perception and ambulatory balance. Other sensory deficits, such as those affecting hearing, taste, smell, or touch, are less common but not unlikely. Damage to regions of the brain that process taste or smell may cause the perception of bitter and/or noxious smell. Likewise, injury to sensory pathways can trigger neuropathic itch, skin tingling, or pain (Oaklander, 2011).

The motor and somatosensory cortices are among the most vulnerable brain regions affected by TBI and diffuse axonal injury due to their superficial position. Damage to these regions is associated with impaired motor control and function. Patients with TBI often report difficulty with balance, changes in ambulatory stride, and loss of fine motor control. Recently, a first-of-its-kind study characterized the presence of a newly defined neurologic disorder called vestibular agnosia in patients with TBI (Calzolari et al., 2021). This condition results in the loss of vertigo perception and imbalance. The Seemungal group found that patients with TBI who exhibit vestibular agnosia have worse balance problems and are unlikely to experience dizziness. Therefore, these patients are at higher risk of subsequent falls or TBI.

In preclinical studies, researchers have focused on changes in motor functionality as a parameter of recovery outcomes. These changes in motor function have been measured with numerous behavioral tests, including comprehensive neurologic scoring, beam walk, and rotarod testing. Furthermore, tasks, such as open field and water maze, can also provide insight on functional recovery by evaluating walking and swim speed between injured and noninjured cohorts. Rodents with TBI and/or PTE demonstrate significantly worsened sensorimotor complications compared with sham-injured controls (Gold et al., 2013; Nissinen et al., 2017; Golub and Reddy, 2022). Insulin-like growth factor-1 overexpression attenuates post-traumatic motor dysfunction, suggesting sensorimotor recovery may be influenced by overactive inflammatory signals or reactive astrocytes (Madathil et al., 2013). Although locomotive ability progressively recovers over time, sensory and cognitive deficits often persist for months or years after injury. Therefore, it is critical to assess sensorimotor outcomes as an indicator of rehabilitation in both preclinical and clinical trials.

B. Memory and Cognitive Dysfunction

There are many forms of cognitive dysfunction, whether it be difficulty retaining or recalling information, disrupted focus, or higher order impairments, such as inability to plan, lack of motivation, or psychosocial disability. Psychometric tests found patients with PTE exhibited a significantly reduced ability to plan, showed a lack of initiative, and had a higher incidence of disinhibited behaviors compared with patients with TBI and no seizures (Mazzini et al., 2003). Furthermore, clinical studies have found evidence for impairments in attention (Fenwick and Anderson, 1999), problem solving (Cazalis et al., 2006), short-term and working memory (Vallat-Azouvi et al., 2007), and mental fatigue (Ziino and Ponsford, 2005).

Preclinical research has hypothesized altered synaptic plasticity and neurodegeneration induced by TBI may play a critical role in cognitive impairment. Mild to moderate TBI is associated with altered hippocampal bursts, with longer duration and lower interburst spike frequency (Munyon et al., 2014). Shorter interval bursts in CA1 hippocampal neurons provoke long-term potentiation and plays a role in synaptic plasticity, thus affecting information coding after TBI and resulting in hippocampal-associated cognitive impairments (Thomas et al., 1998; Ouyang et al., 2017).

TBI is known to affect both short- and long-term memory (Enomoto et al., 2005; Carron et al., 2020). Preclinical in vivo studies often use behavioral tests, such as the Morris water maze or Barnes maze, to evaluate spatial learning and memory (Barnes, 1979; Morris, 1984; Reddy, 1988). Both tests involve training rodents to use visual spatial cues to escape the arena. Over a series of trials, rodents remember the location of the hidden platform or escape box and complete the task progressively faster. Another common task is the Novel Object Recognition Test, which was initially described in 1988 (Ennaceur and Delacour, 1988). The Novel Object Recognition Test examines recognition memory by exposing rodents to two identical objects during a familiarization phase, followed by the replacement of one of those objects with a novel object. Healthy rodents recognize the new object, spending more time investigating it, using hippocampal-dependent recognition memory (Bevins and Besheer, 2006). Several research groups have evaluated short- and long-term cognitive deficits within the context of PTE, finding PTE cohorts perform poorly at these tasks compared with uninjured controls (Scheff et al., 1997; Lu et al., 2015; Nissinen et al., 2017; Golub and Reddy, 2022). These studies reflect similar cognitive deficits found after TBI (Paterno et al., 2017). Due to the extent of cognitive impairment often seen after TBI and how this relates to poorer quality of life in the human condition, it is our opinion that the success of future clinical trials of PTE depends on preclinical models that incorporate both the measurement of cellular and molecular pathologies associated with hyperexcitability as well as memory and behavior tasks.

C. Depression, Mood Disorders, and Anxiety

Cognitive deficits resulting from brain injury can also overlap with development of affective disorders. For example, emotion-recognition difficulties, such as facial affect recognition disorder, contribute to a suite of social functional impairments in patients with TBI (Babbage et al., 2011). Social dysfunction in TBI is well-documented and can negatively affect a person’s ability to form relationships, impairs empathy, and results in low social participation and stress (Hammond et al., 2004).

Among the general population, anxiety and depression have the highest prevalence of any other group of psychiatric disorders, with a lifetime occurrence reported at approximately 30% (Barlow, 2004). A growing body of research indicates that mood and anxiety disorders are even more prevalent in patients with TBI. Approximately 70% of patients with TBI experience a psychiatric illness within the first year after initial injury (Bombardier et al., 2010; Ponsford, 2017). The most common manifestations of psychiatric disorders among adolescents and adults with TBI are generalized anxiety, depression, phobias, post-traumatic stress disorder, and obsessive-compulsive disorder. A long-term neuropsychiatric study found that patients with PTE showed higher incidence of irritability and agitated behaviors, aggression, and personality disorders compared with nonepileptic patients with TBI (Mazzini et al., 2003). Aggressive behaviors may limit access to rehabilitation treatment, participation in employment and social activities, as well as contribute to drifting friendships and romantic relationships.

Neuropathological changes associated with TBI can lead to dysfunction of the lateral and dorsal prefrontal cortices and increased activation of the limbic and paralimbic structures, including the amygdala. Altered amygdala connectivity has been identified as a possible biomarker of both comorbid depression and anxiety in patients with TBI. Relative increases in amygdala connectivity were found in the left dorsomedial and right dorsolateral prefrontal cortices and thalamus as well as the brainstem with spatially dissociable patterns of correlation between this increased connectivity and symptom severity (Han et al., 2015). Another study investigating psychiatric symptoms associated with TBI found that patients with a history of major depression often exhibit comorbid anxiety (77%) and aggressive behavior (57%) (Jorge et al., 2004). These patients had significantly worse executive functionality compared with their nondepressed counterparts. Furthermore, neurotransmitters, such as norepinephrine, serotonin, dopamine, and GABA, are mediators of anxiety and depression symptoms. Disruption of these systems, by either direct impact of TBI or indirect pathologies, such as chronic stress or inflammation, can negatively influence comorbid development of psychiatric disorders.

D. Attention Deficit Hyperactivity Disorder

A meta-analysis documented a significant association between mild TBI and attention deficit hyperactivity disorder (ADHD) (Adeyemo et al., 2014). ADHD is a neurodevelopmental, childhood onset, and persistent disorder characterized by increased impulsivity and risk-taking behavior in individuals. Not only has ADHD been shown to heighten risk of experiencing a TBI, but pre-existing diagnosis of ADHD may also result in worsened recovery outcomes after injury (Bonfield et al., 2013). Compared with patients with TBI but not ADHD, individuals with both diagnoses were significantly more impaired on individual scores of working memory, planning/organization, metacognition indices, and behavioral regulation (Ponsford and Kinsella, 1992; Biederman et al., 2015).

ADHD that develops as a comorbidity of TBI is referred to as secondary ADHD and has been shown to occur in 10%–20% of patients post-TBI (Gerring et al., 1998). Many studies support the hypothesis that dysfunction in the prefrontal cortex, basal ganglia, and their related neurotransmitter systems underlie deficits in inhibitory regulation found in patients with ADHD (Dickstein et al., 2006). Therefore, damage to these areas have the potential to manifest as secondary ADHD or other psychiatric disorders in children and adults. Secondary ADHD may be more prevalent in children and adolescents due to the underdevelopment of brain regions linked to inattention and hyperactivity. Anatomic studies of youth with developmental ADHD demonstrate a loss of volume of frontal and striatal structures without lesions. One case study describes secondary ADHD development in a 10-year-old boy who suffered head trauma and early post-traumatic seizures (Ceylan and Akca, 2013). Cranial CT scans demonstrated decreased density of the right basal ganglia and loss of corticomedullar differentiation in the right frontal area. Six months after the accident, reports from the patient’s parents and teachers describe a change in the child’s behavior of overactivity, boredom at class and home, difficulty maintaining concentration, and depressive-like symptoms. Social disinhibition, hyperactivity, poor impulse control, forgetfulness, and lack of judgement/foresight are among the chronic sequelae of closed head injuries (Kaitaro et al., 1995).

E. Sleep Disorders

Sleep-wake dysfunction after brain injury is common, affecting up to 70% of patients (Viola-Saltzman and Watson, 2012). Common diagnoses after TBI include insomnia, parasomnia, idiopathic hypersomnia, narcolepsy, sleep apnea, and circadian rhythm disorder (Morse and Garner, 2018). This sleep-wake cycle is regulated by a concerted effort between circadian rhythms, sleep-wake homeostasis, metabolism, and external environmental factors, such as diet, stress, medication, and surroundings. Pre-existing sleep conditions can increase the likelihood of post-TBI sleep-wake disturbances. Additionally, comorbid sleep disorders contribute to psychologic instability, resulting in behavioral problems, mood or emotional lability, and worsened daily functioning (Shay et al., 2014; Reddy et al., 2018c).

Alertness and cortical arousal are mediated by several pathways that project from the brainstem near the junction of the midbrain and pons to innervate the thalamus, posterior hypothalamus, and forebrain (Fuller et al., 2006). These ascending pathways are populated by key cell types, including cholinergic, histaminergic, noradrenergic, dopaminergic, and serotoninergic neurons, that fire in a distinct pattern to promote wakefulness. Orexin peptides produced in the lateral hypothalamus reinforce the arousal state. However, this system is inhibited roughly every 24 hours by sleep-active GABAergic and galaninergic neurons of the ventrolateral preoptic nucleus (VLPO) to promote sleep onset (Schwartz and Roth, 2008). The VLPO is a known sleep center influenced by the suprachiasmatic nucleus and activated by endogenous sleep-promoting substances, such as adenosine and prostaglandin D2. Melatonin production from the pineal gland plays a role in REM sleep and aids in synchronizing circadian rhythms.

The pathophysiology of post-traumatic sleep disorders has not been fully explored; however, decreased levels of wake-promoting neurotransmitters, such as hypocretin (orexin) and histamine, are believed to play a contributing role. Post-mortem assessments of brains from patients with TBI demonstrate significant reductions in hypocretin neurons (Shekleton et al., 2010). Even after 6 months of recovery, many patients with TBI exhibit persistently low hypocretin-1 levels, suggesting sleep disorders as a chronic consequence of TBI (Baumann et al., 2007). Disturbed orexin signaling contributes to sleep-wake disorders by causing excessive daytime sleepiness and circadian rhythm disruption (Nishino et al., 2000; Baumann et al., 2005). Moreover, decreased CSF levels of orexin have been associated with poorer clinical outcomes and higher risk of depression and sleep disorders (Brundin et al., 2007). Low CSF histamine levels have also been reported in patients with narcolepsy and after TBI (Kanbayashi et al., 2009). Injury may cause lifestyle changes that can affect sleep and arousal, including changes to diet or medication, exercise routines, or environmental factors. Furthermore, insomnia may develop because of psychologic trauma resulting from personal assaults or accidents that resulted in TBI. Lastly, direct injury to sleep-wake brain centers, such as the VLPO or suprachiasmatic nucleus, may also explain development of circadian rhythm disorders.

Comorbid disturbances in the sleep-wake cycle have also been observed after severe TBI in mice. One critical study tracked nonrapid eye movement sleep for up to 4 months postinjury and found changes in delta power contributed to predictive seizure modeling (Konduru et al., 2021). The authors state that delta power increased in injured mice versus no craniectomy control mice; however, responding injured mice displayed lower delta power at a chronic time point compared with injured mice without seizure activity. Moreover, sleep spindle duration and dominant band frequency is lower in PTE rats compared with sham controls during the transition from stage III to REM sleep (Andrade et al., 2017). These studies suggest sleep-wake disturbances may be a potential biomarker for post-traumatic seizures.

Once the prevalence of PTE-associated comorbidities is established, the search for biomarkers for PTE and recovery may benefit from the neurocognitive network model of affective and/or sleep disorders currently applied to primary epilepsy. Since sensorimotor, cognitive, psychiatric, and sleep-wake comorbidities are so common after TBI, it is important for future work to incorporate quality of life measurements as indicators of recovery outcomes when assessing new therapeutic targets. Additionally, the potential of drug-drug interactions between antiepileptogenesis or antiseizure medications and those that reduce comorbid symptoms should also be evaluated. By examining cellular and synaptic changes in the affected neural circuitry, we can better understand how learning and memory deficits occur. Using newer technologies, such as chemogenetics or optogenetics, could substantiate the relationship between behavior and physiology in a meaningful way.

A. Antiseizure Medications

Antiepileptic drugs, recently renamed as antiseizure medications, are the mainstay for the control of early and late seizures, irrespective of the source of such seizures. More than three dozen ASMs, including first, second, and third, or recent generation, have been extensively tested against post-traumatic seizures (Reddy, 2020). Perhaps the most accessible experiments for PTE are those that evaluate the effect of clinically available ASMs for beneficial effects on disease modification or recovery outcomes. Early studies with positive but uncontrolled results led to widespread use of phenobarbital and/or phenytoin for prophylaxis against immediate and long-term post-traumatic seizures. One survey, conducted in 1973, reported over 60% of clinicians supported immediate pharmacological prophylaxis for head injuries, with 40.3% of respondents prescribing treatment of 1 or more years after injury (Rapport II and Penry, 1973). Current evidence suggests treatment of early seizures does not influence the incidence of PTE (Agrawal et al., 2006). Within the clinical setting, carbamazepine, phenytoin, phenobarbital, and valproate all failed to prevent long-term PTE after head injury (Temkin, 2009). In a recent clinical trial, Hazama et al. (2018) evaluated the benefit of levetiracetam after head trauma. The study consisted of 403 patients, 227 of whom were treated with levetiracetam for early post-traumatic seizure prophylaxis. Although patients receiving levetiracetam treatment trended with lower seizure incidence, this trend never reached statistical significance. These ASMs have shown successful seizure control in other epilepsy disorders but have little effect on preventing or modifying epileptogenesis.

Preclinical research has demonstrated a similar lack of efficacy in reducing PTE incidence with ASMs. Eastman et al. (2010) used the FPI model to systematically detect potential antiepileptic effects of carbamazepine and valproate 1-month postinjury. Carbamazepine (up to 12mg/kg/d for 4.5 days) treatment did not reduce seizure frequency or duration, nor did it reduce incidence of PTE. However, valproate (480 mg/kg/d for 7 days) reduced seizure frequency and cumulative seizure duration. Comorbidities were not evaluated as part of these investigations. In a follow-up study, carisbamate administration was also found to be ineffective at reducing seizure burden or preventing PTE (Eastman et al., 2011) Recently, retigabine, an ASM that was removed from the commercial market in 2017, has shown some promise in reducing seizure burden after CCI in mice. Not only did retigabine administration reduce inflammation, BBB breakdown, and neurodegeneration at the impact site, but also significantly reduced frequency of spontaneous seizures and enhanced susceptibility to chemoconvulsants (Vigil et al., 2020). Gabapentin administration inhibited synapse formation and decreased excitatory synaptic activity after cortical injury (Li et al., 2012). Gabapentin treatment also reduced the expression of astrocytic GFAP expression and thrombospondin-1 protein, as well as the number of fluoro-jade B+ stained cells. These results suggest gabapentin may modify pathways associated with plastic changes in brain excitability, but a controlled in vivo study with continuous EEG recording has yet to confirm an antiepileptogenic effect.

Lastly, Mountney et al. (2013) demonstrated administration of ethosuximide, or phenytoin dose-dependently attenuates in the incidence, frequency, and duration of nonconvulsive seizures after a penetrating blast-like brain injury. In this study, a loading dose of either ethosuximide or phenytoin was given 30-minute post-TBI, followed by a maintenance dose 8 hours later. Four doses of each drug were used to provide a full dose-response curve, with the two highest doses of each drug showing significance in reducing seizure burden. Although these data are promising, this study only follows the rats for up to 72 hours postinjury and identifies immediate and early nonconvulsive seizures, which are considered to be caused by direct impact rather than epileptogenesis. Furthermore, there is little evidence that suggests early seizure prophylaxis to reduce incidence of long-term spontaneous seizures.

B. Targeting Neuronal Excitability Network Pathways

The balance of inhibition and excitation in the brain is very delicate. TBI triggers changes that ultimately result in the disruption of this harmony, leading to increased risk of seizures and epileptogenesis. One approach aimed at minimizing excitotoxicity and preserving cortical inhibition after TBI is the administration of ceftriaxone, a common β-lactam antibiotic with BBB permeability. Ceftriaxone is also a known potent stimulator of glutamate transporter-1 (GLT-1) expression, a critical protein responsible for 95% of total glutamate clearance in the rat brain (Lehre and Danbolt, 1998). Ceftriaxone (200 mg/kg) rescues TBI-induced downregulation of GLT-1 expression within the first week after injury (Goodrich et al., 2013). Authors also report a reduction in regional GFAP expression relative to untreated rats. A follow-up study reported ceftriaxone significantly suppressed both the frequency and duration of post-traumatic seizures weeks after injury (Hameed et al., 2019).

Enhancing GABAergic inhibitory tone is a well-established mechanism for seizure prophylaxis, as many ASMs are approved in this class. GABA_A receptors are also targets of general anesthetics, such as halothane, that can completely suspend seizure activity after FPI, although it leaves the subjects immobile (Eastman et al., 2010). Neurosteroids that modulate tonic inhibition can suppress seizures (Carver et al., 2016; Chuang and Reddy, 2020). Recent attention has been directed toward selective, rather than nonselective (e.g., diazepam), positive allosteric modulators of GABA due to a lower risk for somnolence, motor impairment, and drug abuse (Witkin et al., 2018). KRM-II-81 is an orally bioavailable compound selective for α2/3-containing GABA_A receptors. Witkin et al. (2020) used several models of pharmaco-resistant epilepsy, including kainate-induced mesial temporal lobe seizures, CCI-induced focal injury, and corneal-kindling, to investigate the antiepileptogenic properties of KRM-II-81. They report complete suppression of corneal-kindling seizures in treated mice as well as reduced paroxysmal discharges after kainate injection. Furthermore, repeated injection of KRM-II-81 administration immediately reduced neural hyperactivity for weeks after CCI, suggesting drug resistance or tolerance was negligible. Many ASMs have little therapeutic impact after TBI for controlling post-traumatic seizures; therefore, the cortical excitatory reductions by KRM-II-81 encourage future preclinical and clinical studies.

Lastly, 2-deoxyglucose (2-DG), a competitive inhibitor of hexokinase, the rate-limiting enzyme in glycolysis, has also shown anticonvulsive properties both in vivo and in vitro. Recently, Koenig et al. (2019) explored the compound as a disease modifying agent to prevent epileptogenesis after TBI. Using the CCI model of brain contusion, they report acute 2-DG treatment attenuated hyperexcitability in the brain and prevented development of epileptiform activity in slices taken 3–5 weeks after injury. Additionally, 2-DG treatment reduced the loss of parvalbumin-expressing interneurons, thereby showing neuroprotection against TBI-induced cell loss.

C. Neuroinflammatory Modulation

Targeting neuroinflammation has been a therapeutic strategy in epilepsy disorders for decades, for good reason. Neuroinflammation involves both resident microglia and astrocytes, as well as peripheral immune signaling when the BBB integrity becomes compromised (Reddy et al., 2016b). Furthermore, reactive glial cells can be the drivers of abnormal neuronal activity by impairing the inhibitory action of GABA receptors, reduced neurotransmitter clearance, and disrupted homeostasis (Robel and Sontheimer, 2016). To this end, a handful of compounds have been tested as potential antiepileptogenics for PTE. Although glycyrrhizin, a HMGBI1 inhibitor, was recently found to be ineffective at preventing seizures or reducing susceptibility after TBI, its administration did reduce edema and microglial activation after CCI in pediatric mice (Webster et al., 2019). Toll-like receptor agonists, such as monophosphoryl lipid A and Pam3Cys, significantly slowed amygdala kindling after TBI, demonstrating a reduction in seizure susceptibility (Hesam et al., 2018). Treated rats also exhibited a lessened behavioral response to kindled seizures similar to nontraumatic rats. Given that monophosphoryl lipid A and Pam3Cys are safe and have clinical use as components in vaccines, they have the potential to be used in combination with other agents as a therapeutic strategy for PTE.

Minozac, a suppressor of proinflammatory cytokine upregulation, has been shown to significantly reduce electroconvulsive shock seizures after TBI in mice (Chrzaszcz et al., 2010). These data were coupled with a favorable reduction in TBI-induced metallothionein expression in the CA1, suggesting a reduction in oxidative stress with treatment. Similarly, kineret, an IL-1 receptor antagonist, reduces seizure susceptibility and improved neuropathology associated with epileptogenesis 2 weeks after CCI (Semple et al., 2017). Moreover, mice treated with kineret showed significantly lower seizure frequency compared with vehicle-treated controls at 5 months postinjury. Combined with improved neurobehavioral function, these data provide evidence of IL-1 signaling as a mediator of injury-associated epileptogenesis.

D. Disrupting the Mammalian Target of Rapamycin Pathway

As discussed above, the mTOR signaling pathway is implicated in the regulation of multiple cellular functions that contribute to epileptogenesis. The current hypothesis is such that mTOR signaling is hyperactivated after TBI and triggers multiple downstream mechanisms of PTE. Several studies of mTOR inhibition have demonstrated a beneficial effect of rapamycin, an mTOR inhibitor, in the treatment of epilepsy disorders and improving recovery outcomes after TBI.

Within PTE models, three studies have shown reduced seizure incidence, frequency, or duration after CCI in mice (Guo et al., 2013; Butler et al., 2015; Butler et al., 2016). Guo et al. (2013) observed a significant reduction in seizure incidence, dropping from 50% to 13% in treated versus untreated mice. It was also noted that when rapamycin-treated mice did exhibit seizures, the behavioral component was lessened, suggesting seizure intensity was also reduced by mTOR inhibition. Butler et al. (2015) reported a trend of reduced seizure frequency and incidence with continuous rapamycin treatment but indicated mTOR inhibition was not sufficient to prevent epileptogenesis after CCI. Furthermore, a follow-up study confirmed rapamycin treatment can modify synaptic and tonic GABA_A receptor-mediated currents, hinting toward another mechanism of antiepileptogenic properties (Butler et al., 2016).

Similar to reports on status epilepticus, rapamycin administration resulted in reduced neurodegeneration and attenuated mossy fiber sprouting. Unfortunately, the beneficial effects of rapamycin appear to cease upon discontinuation of the drug in both animal and clinical trials (Buckmaster et al., 2009). Lifelong treatment with mTOR inhibitors is questionable given the known adverse effects, such as immunosuppression, risk of cognitive impairment, and negative influences on normal growth and development in children or adolescents (Bissler et al., 2008).

E. Targeting of Neuronal Plasticity

Neurotrophic factors like BDNF, nerve growth factor, and fibroblast growth factor promote cell survival, growth, and differentiation through activation of signaling pathways after TBI. Abnormalities of interneurons and cell loss play critical roles in epileptogenesis after TBI; therefore, enhancing the function of remaining parvalbumin+ interneurons may be a novel therapeutic approach to preventing PTE. BDNF is known to have positive influences on interneuron growth and function through activation of its receptor, TrkB (Marty et al., 1997). A recent study tested the hypothesis that supporting TrkB function may decrease epileptogenesis after cortical undercut (Gu et al., 2018). LM22A-4 or PTXBD4-3, both partial agonists of the TrkB receptor, were administered up to 2 weeks after injury. PTXBD4-3 reduced seizure susceptibility to PTZ but did not affect seizure duration or latency. LM22A-4, however, decreased the incidence of epileptiform discharges compared with untreated controls. Activation of the TrkB receptor also resulted in increased parvalbumin immunoreactivity, suggesting that partial activation of TrkB may be neuroprotective to interneurons, thereby preserving inhibitory circuity and reducing seizure susceptibility. A different approach explored the use of BDNF blockers in ex vivo organotypic hippocampal slice cultures with collateral lesions and observed reduced hyperexcitability of CA3 pyramidal neurons as well as a reduction in aberrant mossy fiber sprouting (Gill et al., 2013). These differing approaches provide supporting evidence for TrkB signaling modification, but further investigations are needed to confirm antiepileptogenic effects.

F. Hypothermia

Focal cooling can be broadly neuroprotective and has suppressed seizures in animal models, providing evidence that therapeutic hypothermia should be investigated within the context of PTE. Blind, randomized studies of FPI found that a graded cooling up to 2°C significantly reduced seizure frequency and duration in rats. Cessation of ictal activity lasted up to 10 weeks after hypothermia treatment ended (D’Ambrosio et al., 2013). Another study reported reduction of seizure threshold to chemical convulsants and reduced mossy fiber sprouting with hypothermia treatment but did not exhibit neuroprotective effects on cell loss (Atkins et al., 2010). Hypothermia appears to be safe with few negative consequences; however, more research is needed to optimize treatments and define its clinical value.

G. Electrical Stimulation

Deep brain stimulation (DBS) has shown to be remarkably effective, safe, and practical for the treatment of movement disorders, such as tremor, dystonia, and Parkinson’s disease (Deuschl et al., 2006). These successes inspired the application of DBS for other neurologic disorders, including epilepsy. The action of DBS is multifaceted and complex, with high-frequency stimulation (>50 Hz) mimicking the effects of ablative procedures by inducing a reversible functional lesion (Benabid et al., 2002). Lower-frequency stimulation has been associated with anticonvulsant effects, changes in adenosine receptor expression, and altered levels of neurotransmitters and hormones in cerebrospinal fluid. The basis of DBS is to improve abnormal synchrony between different brain regions. One case study was found using DBS as a potential therapeutic intervention for PTE. The patient reported a post-traumatic episode during childhood that resulted in subsequent seizures throughout adulthood. ASM treatments, such as carbamazepine, phenobarbital, and clonazepam, had all been unsuccessful in providing adequate symptomatic control of seizures. Bilateral DBS hippocampal stimulation resulted in a progressive decrease in seizure frequency over the 8 years of follow-up, although some clinicians postulate whether the surgical placement of the DBS system may have contributed to the disruption of the epileptic foci in this case (Piacentino et al., 2018). Other studies have found behavioral improvements and functionality using DBS after TBI, although seizure outcomes were not applicable in these cases (Lee et al., 2013; Shin et al., 2014; Rezai et al., 2016).

H. Tau Hyperphosphorylation

Tau is a microtubule-associated protein that has roles in maintaining neuronal health, axonal transport, and microtubule stabilization. Tau phosphorylation is a normal metabolic process critical to tau’s ability to bind to microtubules. Hyperphosphorylation of tau can cause aggregation and form insoluble fibrillar deposits in tissues. Hyperphosphorylated tau is a long-known hallmark of neurodegenerative diseases, such as Alzheimer’s disease and dementia; however, the causal role of TBI-induced tauopathy has been debated for decades (Castellani and Perry, 2019). Recent studies have confirmed moderate and severe brain injury can trigger the formation of pathologic tau aggregates, linking TBI to increased risk of Alzheimer’s disease (Edwards et al., 2020; Wu et al., 2020).

Sodium selenate mitigates hyperphosphorylated tau by antagonizing PP2A activity and can improve TBI outcomes. Liu et al. (2016) administered sodium selenate (1mg/kg/d), a less toxic form of selenium, via osmotic pumps for 12 weeks after FPI. Treatment with sodium selenate ameliorated the enlargement of ventricles and hippocampal atrophy, which often accompanies brain injury. Additionally, the latency period to spontaneous epileptiform activity, such as seizures or discharges, was extended compared with vehicle-treated rats, suggesting possible interruption of the epileptogenic process. Although this is the only study to specifically investigate the effects of sodium selenate on PTE, sodium selenate has been demonstrated to suppress seizures, improve comorbidities, and reduce seizure susceptibility in experimental models of Lafora disease and temporal lobe epilepsy (Jones et al., 2012; Sanchez-Elexpuru et al., 2017).

I. Cell Transplant Therapies

Cell transplantation, including genetically modified cell types, have been tested as a recovery-enhancing treatment after TBI. Outcome measures include the effects of treatment on lesion volume, severity of neurodegeneration, axonal injury, edema, motor ability, and cognitive function (Jackson et al., 2017). Although positive data have emerged from these studies, most reports have not measured development of spontaneous seizures as an outcome. One recent study, however, found the transplantation of GABAergic progenitors derived from the embryonic medial ganglionic eminence migrated, integrated, and restored post-traumatic decreases in synaptic inhibition (Zhu et al., 2019). Using a CCI model of PTE, mice were recorded using continuous 24/7 video-EEG for between 7 and 20 days at 4 months postinjury. Since most models of PTE find the onset of seizures to be between 20 and 90 days, this timeline was assumed to be after onset of spontaneous epileptiform activity. Incredibly, no electrographic seizures were observed in TBI mice that were implanted with medial ganglionic eminence cells, whereas five of eight untreated mice exhibited repeated ictal events. Cell transplantation therapy also resulted in improved memory precision in transplant mice using the novel object location test. Together, these results provide powerful evidence of antiepileptogenesis using cell-based therapies by restoring long-term deficits in both synaptic inhibition and neurobehavioral impairments. However, this study was limited by the relatively small cohort sizes and shortened length of EEG recording. Follow-up studies could easily address these limitations. This work establishes a promising framework for future studies to evaluate other populations of neurons for cell transplantation therapies for PTE.

J. Dietary and Phytochemical Therapeutic Strategies

Studies dating back to the 1920s have shown that diet and exercise can improve seizure control for patients with epilepsy, with a special emphasis on the ketogenic diet (Wilder, 1921). The ketogenic diet is a high-fat, adequate-protein, and low-carbohydrate diet that has been shown to reduce seizure frequency by over 50% in children and adolescents but has not demonstrated evidence of disrupting the epileptogenic process (Neal et al., 2008). Similar to clinical indications, the ketogenic diet was found to not be effective in preventing PTE after brain injury (Schwartzkroin et al., 2010). However, regular treadmill exercise not only lengthened the latency to seizure onset, but also reduced ictal duration and susceptibility to PTZ-induced seizures in rats (Silva et al., 2013). Creatine supplementation to the diet, on the other hand, has produced contradicting results. Gerbatin et al. (2019) report decreased duration of tonic-clonic seizures and reduced PTZ-induced epileptiform discharges. Creatine supplementation was also found to reduce neurodegeneration of parvalbumin+ interneurons in the hippocampus at 1-month postinjury. These results required long-term administration of daily creatine. Saraiva et al., however, found no reduction on convulsive parameters induced by PTZ in the first week after TBI, but did reduce oxidative damage at the impact site (2012). Together, these studies suggest creatine supplementation may not be a sufficient prophylactic for early post-traumatic seizures but may possess long-term downstream effects which stabilize epileptogenesis.

Cannabis and related compounds have recently broke foreground in epilepsy disorders with the USA FDA-approval of Epidiolex for Lennox-Gastaut and Dravet syndromes in 2018 (Golub and Reddy, 2020a). Cannabidiol has yet to be evaluated for post-traumatic epilepsy, but Rimonabant (SR141716A), a CB1 receptor antagonist may modify disease progression. CB1 receptors are present on both excitatory and inhibitory nerve terminals, where they inhibit glutamate and GABA release, respectively. Therefore, agents which act at this receptor may have multiple effects on neurotransmission. Interestingly, rimonabant is a proconvulsant that has been demonstrated to lower threshold to kainate-induced seizures. Blocking CB1 receptors prevented increased seizure susceptibility in an experimental model of febrile seizures (Chen et al., 2007b). Similar results were found in a rat model of PTE in which a single injection of rimonabant reduced long-term hyperexcitability and susceptibility to kainic acid (Echegoyen et al., 2009). Rimonabant administration after LFPI reversed the overexpression of mGluR5 in late-stage brain injury, thereby lengthening latency to PTZ-induced seizures (Wang et al., 2016c; 2016d). However, these studies are limited by their use of secondary convulsants, such as PTZ and kainic acid, and did not use continuous EEG recording. Moreover, a recent report found investigated the therapeutic effects of Δ9-tetrahydrocannabinol on repetitive mild traumatic brain injury and found postinjury administration reduced anxiety, depressive-like behaviors, and mitigated injury-induced deficits in short-term working memory (Bhatt et al., 2020). More research is needed in this field to fully understand how these compounds may positively or negatively affect recovery outcomes.

Atipamezole, a synthetic α2 adrenergic receptor antagonist, is used mainly in veterinary medicine as it is indicated for the reversal of sedative and analgesic effects of dexmedetomidine and medetomidine in dogs. Treatment with atipamezole reduced PTZ seizure susceptibility at 14 weeks after TBI and improved motor performance but did not prevent the development in spontaneous seizures (Nissinen et al., 2017).

Dextromethorphan, typically used as an over-the-counter cough suppressant, has multiple mechanisms of action including acting as a nonselective serotonin reuptake inhibitor, sigma-1 receptor agonist, and blocks NMDA glutamate receptors at high doses. The dextromethorphan derivative, C-10068, was found to reduce non-convulsive seizure frequency and cumulative seizure duration in a penetrating ballistic-like brain injury model of PTE (Lu et al., 2015). The most efficacious dose of C-10068 also reduced inflammation and reactive microglia accumulation around the lesion site. However, this study reported little improvement in neurobehavioral function and continuous EEG recording was only performed for up to 72 hour postinjury.

Lastly, steroid hormones have been proven to be neuroactive and protective in a variety of CNS disorders (for review see Reddy and Estes, 2016a). Neurosteroids regulate the plasticity of synaptic and extrasynaptic GABA_A receptors involved in the pathophysiology of epilepsy. Progesterone is the precursor to allopregnanolone, which acts as a positive allosteric modulator and direct activator of GABA_A receptors to enhance inhibition in the brain. A recent study utilizing the weight-drop model of PTE reported a reduced duration of PTZ-induced seizures after progesterone treatment (Ghadiri et al., 2019). The behavioral component of seizures was also reduced, suggesting decreased seizure intensity and neuroprotection.

K. Novel Epigenetic Interventions

Epigenetic interventions, such as histone modifiers, represent a novel therapeutic pathway that remains to be fully explored. Global and regional changes in gene expression due to epigenetic modification have been observed after TBI and in epilepsy disorders (Golub and Reddy, 2020b). TBI is known to increase HDAC activity in the brain, resulting in reduced H4 acetylation and increased seizure susceptibility (Dash et al., 2010b; Huang et al., 2012; Reddy et al., 2018b). Reversing hyperacetylation improves motor function and reduces the inflammatory response (Zhang et al., 2008). Furthermore, post-translational histone modifications, such as histone methylation and acetylation, have also strongly implicated in TBI-induced neuropsychiatric disorders (Sagarkar et al., 2017).

Epigenetic HDAC inhibitor treatments, such as valproic acid, sodium butyrate, and suberoylanilide hydroxamic acid (SAHA), have been recently identified as potential disease modifying agents. Valproate has been administered as an anticonvulsant drug for decades, yet its inhibitory effect on HDAC activity was not discovered until 2001 (Gottlicher et al., 2001). Valproate administration reduced neuronal damage, improved cognitive outcomes, and decreased BBB permeability in a CCI model of TBI (Dash et al., 2010b). Although experimental and clinical trials using valproate for PTE have not been effective, some of the observed neuroprotective properties may be linked to this mode of action (Temkin, 2009; Eastman et al., 2010).

Recently, Reddy and team discovered inhibiting HDAC hyperactivity via sodium butyrate treatment retarded the rate of hippocampal kindling in a model of temporal lobe epilepsy (Reddy et al., 2018a). Sodium butyrate treatment also reduced aberrant mossy fiber sprouting in the hilar region of the hippocampus. In addition to blocking a broad spectrum of HDAC enzymes, sodium butyrate is also a known anti-inflammatory agent and shows neuroprotection after stroke (Reddy et al., 2017; Park and Sohrabji, 2016). Stroke is a common cause of epileptogenesis in animal models and humans (Reddy et al., 2017). These data point toward the potential of sodium butyrate as a multifunctional approach to reducing several pathologies associated with epileptogenesis.

In a drug screening study with 870 unique compounds, SAHA was identified as a potential antiseizure drug with selective inhibition of HDAC1 and HDAC3 proteins (Ibhazehiebo et al., 2018). Downstream activity of SAHA results in increased transcription factors crucial to expression of genes needed to induce cell differentiation. Additionally, recent study concluded that SAHA administration after TBI protected against neuronal injury by reducing oxidative stress and inflammation by inducing the inducible nitric oxide synthase/nuclear factor-erythroid factor-2related factor/antioxidant response element (iNOS/Nrf2/ARE) pathway (Xu et al., 2018). Mice treated with SAHA exhibited significantly improved grip test scores and reduced water content in the brain compared with untreated mice. SAHA (Vorinostat) is FDA-approved in patients with cutaneous T cell lymphoma; therefore, validating additional indications in the clinical setting will be a much faster and cost-effective process. Currently, an ongoing phase 2 clinical trial is evaluating the safety and efficacy of SAHA in pediatric patients with drug resistant epilepsy (NCT03894826). Results from this study are forthcoming.

As new biomarkers for PTE are identified, future therapeutic strategies could also include miRNA mimics or antisense oligonucleotides. For example, recent preclinical studies of epilepsy have attempted to control epileptogenesis by regulating expression of miR-146a and found that intranasal delivery of miRNA-146a mimics improved seizure onset and reduced hippocampal damage after pilocarpine administration (Tao et al., 2017). An additional study found similar results in immature rats after an intracerebroventricular injection of miR-146a (Wang et al., 2018). Furthermore, miRNA-146a mimics have been shown to ameliorate injury cascades after TBI (Zhang et al., 2020). Although this method has yet to be studied in PTE directly, studies involving epilepsy or TBI as a singular pathology model have been promising.

There are many studies that highlight disease-modifying effects of test treatments on TBI and comorbidities (Yun Ng and Lee, 2019), but evidence of an antiepileptogenic effect is rare. Furthermore, many studies investigating PTE have done so by measuring seizure threshold to chemical convulsant administration after injury. Although this information is valuable, it does not speak toward the agent’s effect on development of spontaneous seizures and prevalence of long-term epilepsy.

Traditional management of epilepsies has involved the evaluation of the electroclinical phenotype; however, seizures are a symptom of many different causes. Interventions for PTE require the disruption of underlying maladaptive network processes as well as protection against functional impairment. Antiseizure medications are often used for prophylaxis against early TBI-induced seizures but are ineffective at preventing long-term PTE (Wat et al., 2019). Although progress in the field continues to be made, the issue remains as to whether more effective agents will be discovered once the distinct causal processes of PTE are determined and whether these more precise therapeutics will have an improved efficacy and beneficial impact when used earlier in the epileptogenic process.