Targeting BDNF/TrkB pathways for preventing or suppressing epilepsy

Abstract

Traumatic brain injury (TBI) and status epilepticus (SE) have both been linked to development of human epilepsy. Although distinct etiologies, current research has suggested the convergence of molecular mechanisms underlying epileptogenesis following these insults. One such mechanism involves the neurotrophin brain-derived neurotrophic factor (BDNF) and its high-affinity receptor, tropomyosin related kinase B (TrkB). In this review, we focus on currently available data regarding the pathophysiologic role of BDNF/TrkB signaling in epilepsy development. We specifically examine the axonal injury and SE epilepsy models, two animal models that recapitulate many aspects of TBI- and SE-induced epilepsy in humans respectively. Thereafter, we discuss aspiring strategies for targeting BDNF/TrkB signaling so as to prevent epilepsy following an insult or suppress its expression once developed.

This article is part of the special issue entitled ‘New Epilepsy Therapies for the 21st Century - From Antiseizure Drugs to Prevention, Modification and Cure of Epilepsy’.

1. Introduction

Brain derived neurotrophic factor (BDNF) is the most widely studied neurotrophin in the mammalian nervous system. Using an in vitro bioassay detecting neuroprotective activity, BDNF protein was purified from porcine brain by Yves Barde (Barde et al., 1982). The peptide sequence enabled the molecular cloning of the cDNA for BDNF (Leibrock et al., 1989). This was followed shortly thereafter by the discovery of the receptor mediating its biological effects, tropomyosin related kinase B (TrkB) (Klein et al., 1991). Together these discoveries have provided tools and biological insights that have fueled three decades of extraordinary advances.

The epilepsies are just one of countless diseases in which BDNF/TrkB signaling is thought to play a critical role. The discovery that BDNF expression dramatically and transiently increases following prolonged seizures has fueled early interest in this topic (Isackson et al., 1991). Subsequent recognition that development of epilepsy in the kindling model is impaired in BDNF heterozygous mice ultimately led to the hypothesis that endogenous BDNF/TrkB signaling promotes epileptogenesis (Kokaia et al., 1995). An extensive literature has followed for which a series of thoughtful reviews have appeared (Binder et al., 2001; Hu and Russek, 2008; McNamara and Scharfman, 2012). Here we focus specifically on the contribution of BDNF/TrkB signaling to epileptogenesis, the process by which a normal brain becomes epileptic. We focus particularly on models in which selective perturbation of BDNF/TrkB signaling is introduced following the causal insult and in which such perturbations modify the insult-induced hyperexcitability. We focus particularly on two distinct injuries resulting in epileptogenesis - cortical axonal transection and status-epilepticus (SE) - that model human TBI- and SE-induced epilepsy.

2. Axonal transection models of post-traumatic epilepsy: BDNF/TrkB signaling

Traumatic brain injury (TBI) is a common cause of acquired epilepsy in humans with spontaneous recurring seizures emerging after a latent interval of months to years following the injury. Here we address the following questions: does TBI perturb BDNF/TrkB signaling and, if so, does this contribute to development of post-traumatic epilepsy (PTE)? Currently, there are multiple animal models available to investigate TBI. For the purposes of this review, we focus on experimental models utilizing axonal transection. The reasons are threefold: A) such pathology is seen clinically in TBI; B) animal models of this injury have been developed; and C) investigations of the causal role of BDNF/TrkB signaling have been performed.

The fact that undercutting cerebral cortex can induce hyperexcitability of the overlying cortex was established in studies of cats and monkeys several decades ago (Prince and Futamachi, 1970; Purpura and Housepian, 1961; Sharpless and Halpern, 1962). These models have been extended to guinea pigs, rats, and mice (Hoffman et al., 1994; Prince and Tseng, 1993; Salin et al., 1995). White matter underlying small regions of cerebral cortex is transected to isolate the overlying cortex while leaving vascular supply intact (Hoffman et al., 1994; Prince et al., 2012). Electrophysiological study of slices containing the lesion isolated weeks afterward reveal epileptiform events detected in field potential or whole cell recordings in >80% of animals (Hoffman et al., 1994); the duration of such events is typically a few hundred milliseconds, a pattern consistent with interictal spikes detected on scalp electroencephalogram (EEG) of patients with epilepsy. Longer duration events consistent with electrographic seizures can be induced by incubating slices from undercut cortex in low concentrations of the GABA_A receptor antagonist, bicuculline (Gu et al., 2018). Although video-EEG studies of spontaneous recurrent seizures detected in vivo are not commonly reported, animals undergoing undercut cortex do exhibit enhanced sensitivity to chemoconvulsant induced seizures (Gu et al., 2018).

Prince and colleagues have conducted extensive studies aimed at elucidating the cellular and circuit mechanisms underlying the epileptiform activity observed in slices isolated weeks following undercutting cortex. Current source density analyses of field potentials revealed the epileptiform activity originated in layer 5 (L5) (Hoffman et al., 1994). Whole cell recordings of L5 principal neurons reveal increased frequency of both spontaneous excitatory post synaptic currents (sEPSCs) and miniature excitatory post synaptic currents (mEPSCs) (Li and Prince, 2002). Enhanced amplitude of AMPA/KA (Kainic acid), but not NMDA, component of the evoked EPSC was evident in L5 cells of undercut cortex (Li et al., 2005; Li and Prince, 2002). Paired pulse depression of the evoked EPSC was reduced, consistent with increased probability of release from glutamatergic terminals (Li et al., 2005). Analyses of biocytin-filled L5 principal neurons revealed no overt change of dendritic morphology yet enhanced length and branching of axons (Salin et al., 1995). Increased immunoreactivity of GAP-43, a marker of axonal sprouting, was observed as well (Prince et al., 2009). Glutamate uncaging-mediated activation of L5 principal neurons provided evidence consistent with enhanced synaptic connectivity among L5 principal neurons, supporting the emergence of enhanced recurrent excitatory synaptic networks arising as a consequence of the undercut cortex (Jin et al., 2006). This provides one attractive mechanism of the increased cortical excitability arising following axonal injury.

In addition to the enhanced excitatory synaptic connectivity, impairments of synaptic inhibition of L5 principal neurons were discovered in slices from undercut cortex. Whole cell recordings of L5 principal cells revealed reduced frequency of spontaneous IPSCs accompanied by reduced frequency but not amplitude of miniature inhibitory post synaptic currents (mIPSCs) (Li and Prince, 2002). Although no reduction in number of parvalbumin (PV) interneurons was found, PV neurons exhibited thinner dendrites and reduced numbers of large boutons (Gu et al., 2017). Additional immunohistochemical studies revealed reduced immunoreactivity of vesicular GABA transporter (VGAT), glutamic acid decarboxylase 65 kD and 67 kD isoforms (GAD65 and GAD67) surrounding soma of L5 principal neurons (Gu et al., 2017).

The structural defects of PV interneurons together with impaired synaptic inhibition of principal cells raised the possibility that undercutting cortex may have somehow reduced trophic support for these neurons. This led Prince and colleagues to examine the contribution of BDNF/TrkB signaling. Under normal circumstances, TrkB is expressed in both principal cells and interneurons widely throughout the adult CNS as evidenced by in situ hybridization detection of mRNA; by contrast, BDNF mRNA is evident in principal cells but at low or undetectable levels in interneurons (Cembrowski et al., 2018). Interestingly, reductions of BDNF mRNA in L5 principal cells were evident 3 weeks after the lesion following undercutting (Gu et al., 2017); if, as is likely, the reduction of mRNA is paralleled by reduction of BDNF itself, this would be expected to reduce activation of TrkB receptors and trophic support of PV interneurons.

Collectively, these findings provided the rationale for testing whether modulation of BDNF/TrkB signaling could impact epilepsy development. One such modulatory method involved a partial agonist of the TrkB receptor, LM22A-4 (Massa et al., 2010). Discovered by an in silico screening strategy using a fragment of BDNF, LM22A-4 exhibits partial agonist properties with TrkB, but not TrkA or TrkC receptors in cellbased assays. Consistent with this, systemic administration in vivo activates TrkB, AKT, and ERK as detected by western blotting of cortical lysates (Massa et al., 2010). Utilizing this partial agonist, Gu et al. (2018) examined the effects of treatment initiated immediately after undercutting and continuing for 2 weeks at which time structural and functional measures of inhibitory neurons were performed. Using whole cell and field potential recordings of slices of undercut cortex, the authors found increased frequency of mIPSCs in L5 pyramidal cells and reduced frequency of interictal bursts following LM22A-4 treatment. LM22A-4 treatment also reduced the frequency of electrographic seizures recorded in slices incubated in the GABA_A receptor antagonist, bicuculline. Notably, no changes of mEPSCs were detected in animals treated with LM22A-4 compared to vehicle. These enhancements of synaptic inhibition were accompanied by enhanced immunoreactivity of VGAT, GAD65 and parvalbumin around soma of L5 principal cells. The effects of LM22A-4 were attenuated by simultaneous treatment with a TrkB kinase inhibitor using a chemical genetic approach described below, thus supporting the interpretation that the beneficial effects of LM22A-4 were mediated by activation of TrkB. Although effects of LM22A-4 on spontaneous recurrent seizures were not reported, treatment with an analog of LM22A-4, did reduce the undercuttinginduced enhanced sensitivity to the chemoconvulsant, pentylenetetrazol.

In a distinct axonal transection model, implemented by Scott Thompson and colleagues (McKinney et al., 1997), knife cuts were placed in rodent hippocampus to transect the Schaffer collateral axons of CA3 pyramidal cells that form synapses with CA1 pyramidal cells. Such lesions in cultured hippocampal slices induced sprouting of CA3 pyramidal cell axons, marked increases of a marker of synaptic vesicles, and increased GAP-43 immunoreactivity within axons (a marker of axon growth) (McKinney et al., 1997). Whole cell recordings of CA3 pyramidal cells revealed enhanced numbers of sEPSPs and sIPSPs in the lesioned hippocampi which correlated with levels of GAP-43 immunoreactivity. Dual whole cell recordings revealed increased numbers of synaptically coupled pairs of CA3 pyramidal cells in the lesioned compared to control hippocampus. Incubation with low concentrations of the GABA_A antagonist, bicuculline caused epileptiform bursting in slices from lesioned but not sham controls (McKinney et al., 1997). The presence of large numbers of spontaneous and evoked IPSPs following the lesions together with similar levels of GABA immunoreactivity in lesion and control slices supported the integrity of GABA mediated synaptic inhibition. Collectively, these data support the authors’ conclusion that transecting the axons of the CA3 pyramidal cells induced axonal sprouting and enhanced excitatory synaptic connectivity among these cells ultimately leading to hyperexcitability.

In search of a molecular mechanism, the previously demonstrated contribution of neurotrophins to axonal outgrowth (Huang and Reichardt, 2001) led Thompson and colleagues to examine the role of BDNF/TrkB signaling. Knife cuts induced a striking increase of BDNF content as detected by ELISA measurements, peaking 24–48 h after the lesion followed by a return to normal by 72 h. This was associated with increased content of TrkB (Dinocourt et al., 2006). Further, the knife cut-induced increase of GAP 43 immunoreactivity was virtually eliminated in slices cultured from transgenic mice exhibiting low levels of TrkB (Dinocourt et al., 2006), thus supporting a causal role of BDNF/TrkB signaling in the lesion-induced axonal sprouting.

Findings with this in vitro model provided the rationale for investigating the contribution of BDNF/TrkB signaling to axonal sprouting and hyperexcitability induced by a knife cut in vivo.

Similar to findings from the cultured hippocampal slices, epileptiform bursting of CA3 pyramidal cells was detected in slices following knife cuts but not in sham controls. Abnormal bursting was detected 7–21 days after lesioning but not earlier, evidence of a latent period between lesion and emergence of hyperexcitability (Aungst et al., 2013). Western blot analyses revealed biochemical evidence of enhanced activation of TrkB in slices isolated at 24 h following the lesion but not at later time points. GAP-43 immunoreactivity was increased at 7 days following the lesion but not at days 1, 14, or 21 after the lesion (Aungst et al., 2013).

To circumvent potential developmental confounds from the TrkB mutant mice examined in the in vitro studies described above, Aungst et al. (2013) deployed a powerful chemogenetic method that provided the temporal control afforded by pharmacology with molecular specificity afforded by genetic perturbations. This method utilized TrkB^F616A mice in which an alanine was substituted for phenylalanine in the kinase domain of TrkB; this rendered the mutant TrkB kinase inhibitable by a small molecule, 1NMPP1. Because 1NMPP1 does not inhibit the kinase in the WT mice, effects of 1NMPP1 in mutant but not wild type mice can be ascribed specifically to inhibition of TrkB kinase. Importantly, in the absence of 1NMPP1, the wild type and mutant TrkB kinases function equally well.

Aungst et al. (2013) found that knife cuts of hippocampus produced hyperexcitability and axonal sprouting to a similar extent in both wild type and TrkB^F616A mice when treated with vehicle. However, treatment of the mutant mice with 1NMPP1 prevented both the increase of GAP-43 immunoreactivity and development of the hyperexcitability. Treatment of wild type mice with 1NMPP1 did not inhibit either the hyperexcitability or the axonal sprouting measures. Separately, Gill et al. (2013) demonstrated that treatment with the BDNF scavenger, TrkB-Fc, attenuated the lesion-induced increase of action potential firing, increased evoked fEPSPs along with axonal sprouting post lesion in cultured hippocampal slices. These data implicate a causal role of BDNF in development of hyperexcitability in this model. In sum, these data clearly demonstrate that inhibition of BDNF and TrkB kinase prevents the epileptiform consequences of the knife cut that transect the axons of CA3 pyramidal cells as well as expression of the axonal sprouting marker, GAP43.

2.1. Perspective

What accounts for the opposite effects of TrkB signaling on epileptogenesis in these two models of axonal transection induced hyperexcitability? That is, Gu et al. (2018) report that a novel compound with TrkB partial agonist properties prevents development of hyperexcitability in the undercut cortex model; the authors conclude that the beneficial effects are mediated by activation of TrkB. By contrast, Aungst et al. (2013) report that chemical genetic inhibition of TrkB kinase prevents development of hyperexcitability induced by lesions that transect axons of CA3 pyramidal cells.

One issue warranting careful consideration is the possibility that the effects of LM22A-4 are mediated by a target distinct from TrkB. Indeed, Boltaev et al. (2017) failed to detect activation of TrkB by LM22A-4 in a series of quantitative assays measuring receptor activation, downstream signaling, and gene expression, supporting this concern. That said, Gu et al. (2018) demonstrated that chemical genetic inhibition of TrkB kinase prevented the beneficial effects of LM22A-4, clearly implicating a TrkB mechanism of action of LM22A-4.

An alternative possibility, which we favor, is that BDNF/TrkB signaling can have opposite effects on axonal transection induced hyperexcitability in these two different models. The authors provide evidence that the mechanism underlying the hyperexcitability in the undercut neocortex model is impaired synaptic inhibition, likely mediated at least in part by the PV subset of interneurons. By contrast, the mechanism underlying the hyperexcitability caused by hippocampal knife cuts appears due at least in part to formation of enhanced excitatory synaptic connectivity mediated by axonal sprouting. If correct, the net effect of enhancing BDNF/TrkB signaling on excitability of a local circuit in these two locales would be opposite: reduced excitability in neocortex and enhanced excitability in hippocampus.

3. Status epilepticus models: BDNF/TrkB signaling

Clinical and preclinical observations implicate an episode of prolonged seizures, “status epilepticus” (SE), to be one cause of temporal lobe epilepsy. Clinical observations and longitudinal studies (French et al., 1993; Lewis et al., 2014; Shinnar et al., 2008; VanLandingham et al., 1998) support the idea that SE can result in temporal lobe epilepsy (TLE) after a latent period of months to years. Between 13–82% of children who experience afebrile convulsive SE eventually develop epilepsy with a long latent period of up to 10 years (Raspall-Chaure et al., 2006), whereas available data indicate that about 30% of teens or adults who experience SE develop epilepsy within two years (Hesdorffer et al., 1998). The fact that experimental induction of SE by electrical stimulation or diverse chemical methods (Loscher, 2002) is sufficient to transform a normal brain to an epileptic brain supports the idea that SE is one cause of TLE in humans.

Several factors provided the rationale for testing the hypothesis that SE-induced BDNF/TrkB signaling promoted development of TLE. First, abundant biochemical, genetic, and pharmacological evidence implicates enhanced BDNF/TrkB signaling in promoting epileptogenesis using kindling models (reviewed by McNamara and Scharfman, 2012). Notably, Chris Gall and colleagues demonstrated that SE induces dramatic increases in expression of BDNF that persists for several days (Isackson et al., 1991), an increase that correlates with the time course of SE-induced activation of TrkB signaling (He et al., 2010). Optimally, testing the hypothesized causal role of SE-induced BDNF/TrkB signaling requires initiating a perturbation after the SE because intervening prior to or during SE could attenuate the intensity of the insult and thereby confound interpretation of prevention of TLE. Further, the duration of the intervention needs to be limited in time; assessing occurrence of seizures in the continued presence of a drug cannot distinguish preventive from simply anti-seizure effects. For these reasons, the experiments considered here will be restricted to designs in which perturbations were introduced after the SE and continued briefly thereafter.

The lack of highly selective and structurally distinct small molecule inhibitors of TrkB led Liu et al. (2013) to implement a chemical-genetic strategy utilizing TrkB^F616A mice as described above (Liu et al., 2013). One week following stereotaxic implantation of a guide cannula overlying the right amygdala and an EEG recording electrode in the left hippocampus, SE was induced by infusion of kainic acid (KA) into the basolateral nucleus of the amygdala. SE was terminated by systemic administration of diazepam followed 1 h later by lorazepam. Like other models of SE (He et al., 2010), KA-SE induced enhanced activation of TrkB that persisted for 2–3 days as measured by Western blot analyses of membranes isolated from the hippocampus. SE was induced in both wild type and TrkB^F616A mice; following administration of diazepam to suppress SE, treatment with either vehicle or 1NMPP1 was initiated and continued for two weeks. Spontaneous recurrent seizures were detected with continuous video-EEG recordings for two weeks beginning approximately one month after SE. Whereas seizures were detected in all 19 control animals (wild type treated with vehicle or 1NMPP1; TrkB^F616A mice treated with vehicle), no seizures were detected seven of eight 1NMPP1 treated TrkB^F616A mice. A single seizure was detected in the eighth 1NMPP1 treated TrkB^F616A mouse. In addition to the seizures, SE induced anxiety-like behavior assessed by light dark test in the control groups; treatment of TrkB^F616A mice with 1NMPP1 prevented development of anxiety-like behavior. In sum, this chemical genetic approach provided unambiguous evidence that transient inhibition of TrkB kinase, initiated after SE, is sufficient to prevent SE-induced TLE and associated anxiety-like behavior in this mouse model.

While this work advanced TrkB kinase as a therapeutic target for prevention of epilepsy induced by SE, the pro-survival effects of TrkB signaling (Alcantara et al., 1997; Atwal et al., 2000) raised concern that inhibition of TrkB kinase might exacerbate death of CNS neurons induced by SE (Henshall and Meldrum, 2012). If so, such an untoward consequence would diminish enthusiasm for use of TrkB kinase inhibitors in a prevention trial. To address this issue, Fluro-Jade C (FJC) staining of hippocampal sections from mice euthanized 24 h following SE was used to quantify neuronal degeneration (Mouri et al., 2008). Chemical genetic Inhibition of TrkB kinase was initiated immediately following diazepam treatment as described above. SE induced FJC staining of hippocampal neurons, preferentially involving CA3 pyramidal cells in the three control groups (Wild type treated with vehicle or 1NMPP1; TrkB^F616A mice treated with vehicle). Inhibition of TrkB kinase (1NMPP1 treatment of TrkB^F616A mice) increased the number of FJC positive cells in the CA3 and CA1 pyramidal cell layers of hippocampus by three to tenfold. These results reveal a neuroprotective benefit of SE-induced TrkB activation and a deleterious consequence of global inhibition of TrkB signaling in this context, namely increased neuronal degeneration.

The untoward consequences of global inhibition of TrkB signaling raised the question as to which signaling pathway downstream of TrkB mediated the epileptogenic consequences of SE. That is, if the downstream pathways mediating the deleterious and beneficial effects of TrkB signaling induced by SE were distinct, then selective inhibition of the deleterious pathway could be an attractive strategy for drug development. Evidence linking PLCγ1 to long term potentiation of excitatory synapses suggested that activation of PLCγ1 signaling mediated the epileptogenic consequences of TrkB activation. Gu et al. (2015) first asked whether genetic inhibition of this signaling pathway might impair induction of SE (Gu et al., 2015). Studies of TrkB^PLC/PLC mice (in which a phenylalanine is substituted for tyrosine at residue 816 of TrkB and thereby disrupts TrkB-mediated PLCγ1 signaling [Minichiello et al., 2002]) revealed partial inhibition of KA-induced SE in comparison to wild type controls. Similar results were obtained with PLCγ1^+/− heterozygous mice (Gu et al., 2015), these antiseizure effects raising the possibility that this pathway contributes to epileptogenesis.

To test the contribution of TrkB/PLCγ1 signaling to SE-induced epileptogenesis, Gu et al. (2015) designed a novel cell- and blood-brain-barrier permeable peptide, named pY816. Knowledge of the motif of TrkB required for binding of the SH2 domain of PLCγ1 led to synthesis of a peptide containing phosphorylated tyrosine 816 (pY816, YGRKKRRQRRR-LQNLAKASPVpYLDI) (Middlemas et al., 1994; Obermeier et al., 1993). To facilitate transport across cell membranes and the blood brain barrier, an 11 amino acid fragment of the HIV-1 Tat protein transduction domain was fused to the N-terminus. The dose- and time-dependent inhibition of PLCγ1 assessed by biochemical study of hippocampal membranes following systemic administration of pY816 was used to guide study of pY816. Using an experimental design with SE induced by intra-amygdala infusion of KA similar to that deployed in studies of the chemical-genetic approach above, pY816 or a scrambled control peptide was administered following diazepam treatment and again one and two days later. Video-EEG recordings performed approximately one month thereafter revealed a 90% reduction in the number of recurrent seizures in comparison to the scrambled peptide controls. Likewise, treatment with pY816 prevented development of anxiety-like behavior evident in scrambled control animals. In sum, these findings implicate PLCγ1 as the dominant effector downstream of TrkB responsible for the epileptogenic and anxiogenic consequences of TrkB activation following SE.

Additional studies implicated a distinct TrkB-mediated signaling pathway in the neuroprotective consequences of TrkB signaling following SE, namely TrkB/Shc/Akt signaling. Whereas phosphorylation of tyrosine 816 of TrkB is critical for its binding and activation of PLCγ1, phosphorylation of tyrosine 515 of TrkB is critical for its binding the adaptor protein, Shc, and activation of Akt. The requirement of TrkB/Shc signaling in BDNF-mediated neuronal survival for mammalian neurons in vitro (Atwal et al., 2000; Minichiello et al., 1998) and for peripheral sensory neurons in vivo (Minichiello et al., 1998) led to the idea that TrkB-activated Shc/Akt signaling mediated the neuroprotective effects of TrkB activation induced by SE. These experiments utilized TrkB^Shc/Shc mutant mice in which a point mutation (Y515F) of TrkB prevents the binding of Shc to activated TrkB kinase. The severity of SE induced by intra-amygdala infusion of KA was similar in Wild Type and TrkB^Shc/Shc mutant mice. Yet SE-induced activation of the pro-survival adaptor protein Akt was partially inhibited in the TrkB^Shc/Shc mice compared to wild type controls. This was paralleled by enhanced hippocampal neuronal death induced by SE in the TrkB^Shc/Shc mice compared to wild type controls. In sharp contrast to these findings with TrkB^Shc/Shc mice, no differences in SE-induced cell death were detected in animals treated with pY816 following SE in comparison to scrambled controls. In sum, these studies provide in vivo proof of concept evidence that it is possible to disentangle the desirable (neuroprotective) from the undesirable (epileptogenic and anxiogenic) consequences of TrkB signaling by selectively disrupting pathways downstream from the receptor tyrosine kinase.

4. Seizure-suppressant effects of BDNF

In contrast to studies of BDNF/TrkB signaling on development of TLE induced by SE, additional studies asked whether chronic treatment with BDNF suppressed seizures following their emergence weeks following SE. Chronic treatment with BDNF was achieved by implanting a cell line secreting BDNF into the hippocampus (Falcicchia et al., 2018). Given the evidence implicating enhanced BDNF/TrkB signaling in promoting epileptogenesis induced by SE, one might have expected that introduction of cells releasing BDNF into the hippocampus of “epileptic” rats would increase seizure frequency. Yet the opposite result was recently reported (Falcicchia et al., 2018). In these experiments, Falcicchia et al. (2018) injected a cell line expressing recombinant human mature BDNF into the hippocampus of adult rats three weeks following SE induced by systemic pilocarpine. Seizures were detected by behavioral observation by blinded observers. Cell lines were encapsulated in matrix that was separated from host brain tissue by a thin polymer membrane with pores sufficient to allow oxygen and nutrients to enter and BDNF to exit, yet prevented entrance of host immune cells. Following emergence of spontaneous recurrent seizures, devices containing BDNF secreting cells or devices containing cells lacking BDNF were implanted bilaterally in ventral hippocampus and the occurrence of seizures detected by behavioral observation over the next three weeks. Novel object recognition tests of memory function were performed prior to SE and at varying intervals thereafter. A diversity of histological analyses were performed three weeks following implantation.

Results revealed a striking reduction (80%) of seizure frequency. In contrast to the progressive increase of seizure frequency in the control groups, the BDNF treated animals exhibited a significant and persistent 80% reduction of seizure frequency. Performance in novel object recognition testing deteriorated when assessed three weeks following SE in control groups and deterioration persisted when tested an additional three weeks later. By contrast, SE-induced impairments of cognitive performance improved three weeks following implantation of BDNF secreting cells. Histological analyses revealed a 30% reduction of hippocampal volume following SE in the several control groups yet no reduction in the group with BDNF secreting cells. The number of PV immunoreactive neurons was reduced by 50% in the control groups yet only 20% in the group treated with BDNF. Striking increases of Flurojade C positive cells indicative of robust and active cell death were evident in the control groups yet limited numbers were evident in the BDNF treated animals.

Whether the BDNF secreting cell lines produced these beneficial effects by increasing or decreasing activation of TrkB is uncertain. Brief application of BDNF to cultured CNS neurons or a cell line expressing TrkB enhances TrkB activation as assessed by Western blot analyses of both TrkB and a surrogate of TrkB activation (phosphorylated TrkB) (Sommerfeld et al., 2000); however, sustained exposure to BDNF induces reduction of TrkB activation which is associated with TrkB receptor internalization. Sustained infusion of BDNF into the hippocampus by an osmotic minipump caused reductions of both TrkB and phosphorylated TrkB which was paralleled by inhibiting epileptogenesis in the kindling model (Xu et al., 2004). Thus, it seems plausible that the mechanism by which BDNF secreting cells inhibited seizures is by reducing activation of TrkB. Levels of TrkB and phosphorylated TrkB assessed by Western blotting were not reported by Falcicchia et al. (2018), leaving this an open question.

Uncertainties as to the molecular mechanism notwithstanding, these nicely designed studies clearly establish a diversity of beneficial effects of intrahippocampally implanted devices containing BDNF secreting cell lines. Because the devices were present until death of the animals, it is unclear whether the effect of the BDNF was simply to suppress seizures; if so, the seizures would be expected to recur upon removal of the devices. Also unclear is whether improvement in cognitive function is due to suppression of seizures; notably, viral-mediated delivery of BDNF to entorhinal cortex improved cognitive function in animal models of Alzheimer’s disease in the absence of overt seizures (Nagahara et al., 2009). The enhanced number of PV expressing interneurons and suggested enhancement of synaptic inhibition provides one potential mechanism underlying the beneficial effects.

5. Path to the clinic for BDNF/TrkB signaling

The studies of pathologic hyperexcitability induced by lesions that transect axons raises the question as to whether perturbing BDNF/TrkB signaling may be deployed to prevent epilepsy following traumatic brain injury. Brain trauma can be caused by acceleration/deceleration injury resulting in diffuse damage but also by focal injuries resulting in skull fracture, dural penetration, and hematoma. The extent to which hyperexcitability caused by axonal transection contributes to epilepsy arising after brain trauma in these diverse settings is uncertain. Ongoing work by multiple groups is centered on developing animal models aligned with defined subsets of brain trauma in which spontaneous recurrent seizures arise, preferably predicted by a biomarker detected shortly after injury. Availability of such models will enable study of perturbing BDNF/TrkB signaling on development of post-traumatic epilepsy. If axonal transection induced hyperexcitability proves to be an important factor in pathogenesis of post-traumatic epilepsy, the opposite effects of TrkB signaling on hyperexcitability caused by lesions in neocortex versus hippocampus provides a cautionary note in design of such studies.

For temporal lobe epilepsy, current evidence implicating BDNF/TrkB/PLCγ1 signaling as a molecular mechanism by which SE induces temporal lobe epilepsy (Gu et al., 2015; Liu et al., 2013) provides the rationale for pursuing inhibitors of this signaling pathway for possible disease prevention. The limited duration of exposure (three days) to an inhibitor enhances the attractiveness of this target. The fact that inhibition of PLCγ1 of just 50–60% exerted powerful preventive effects (Gu et al., 2015) further enhances chances of successful trial. The availability of a biomarker assessed during or shortly after SE that identified patients destined to become epileptic would be a valuable tool for design and execution of a successful clinical trial.

Similar to findings in SE models, convulsive seizures lasting one to two days induced by brief hypoxia in P10 rat pups activated TrkB signaling evidenced by increased in TrkB phosphorylation detected by Western blotting. Treatment with CEP-701, a drug that inhibits TrkB activation, immediately after hypoxic seizures and again 12 h later reduced severity of seizures induced by kainic acid at P14 (Obeid et al., 2014). Although effects of CEP-701 are not selective for TrkB, these findings nonetheless support the idea that inhibition of TrkB signaling transiently in this context may limit epileptogenesis following a distinct insult, namely hypoxia.

Finally, the use of matrix encapsulated cell lines secreting human BDNF provides an interesting option for patients with medically refractory temporal lobe epilepsy. Even if the effect solely consists of seizure suppression, these cell lines could prove valuable clinically. That is, if implanting such cell lines eliminated seizures or transformed medically refractory to medically responsive patients, such cell lines could provide an attractive alternative to surgical removal of epileptic tissue. This could be particularly appealing in subsets of patients with evidence of seizures arising from both temporal lobes who would not be candidates due to the bilateral nature of their disease. The fact that these matrix encapsulated cell lines can be removed surgically in the event of unforeseen deleterious consequences further enhances their attractiveness.

HIGHLIGHTS.

• BDNF/TrkB signaling plays a pathophysiologic role in TBI- and SE-induced epilepsies.

• Contribution of BDNF/TrkB signaling to epileptogenesis in axonal transection models is opposite in lesions of neocortex versus hippocampus.

• Peptide inhibitor is a valuable tool for inhibiting BDNF/TrkB signaling.

Abbreviations

AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

BDNF

brain derived neurotrophic factor

CNS

central nervous system

EEG

electroencephalogram

ELISA

enzyme-linked immunosorbent assay

FJC

fluoro-jade C

GABA

gamma aminobutyric acid

GAD

glutamic acid decarboxylase

GAP-43

growth associated protein 43 kDA

KA

kainic acid

mEPSCs

miniature excitatory post synaptic currents

mIPSCs

miniature inhibitory post synaptic currents

NMDA

N-methyl-D-aspartate

PLCγ

phospholipase C-γ

PTE

post-traumatic epilepsy

PV

parvalbuminEEG electroencephalogram

SE

status epilepticus

sEPSCs

spontaneous excitatory post synaptic currents

sIPSCs

spontaneous inhibitory post synaptic currents

TBI

traumatic brain injury

TLE

temporal lobe epilepsy

TrkB

tropomyosin related kinase B

VGAT

vesicular GABA transporter

Appendix

Tools to define cellular/subcellular localization of BDNF/TrkB signaling

Knowing the cellular distribution and expression of both BDNF and its cognate receptor, TrkB, provides valuable information for design and interpretation of experiments examining BDNF/TrkB signaling in epileptogenesis. Availability of antibodies that selectively bind these targets with high sensitivity under conditions of immunohistochemistry has been, and continues to be, a limitation for such experiments. Validating the antibodies by comparing signal in sections from wild type and mutant mice engineered to lack the desired target is essential. The low concentrations of BDNF in particular (Barde et al., 1982; Kolbeck et al., 1999) together with its conserved sequence (Aid et al., 2007; Pruunsild et al., 2007) likely contribute to limited availability of BDNF antibodies that selectively label BDNF with high sensitivity (Barde et al., 1982; Kolbeck et al., 1999). Among several available antibodies against BDNF, two monoclonal antibodies (Kolbeck et al., 1999) and one polyclonal antibody (Conner et al., 1997; Yan et al., 1997) have been well characterized and confirmed; unfortunately, only the monoclonal antibodies are commercially available and the sensitivity of these reagents appears limited. Similar concerns arise with respect to antibodies for TrkB as well as for phosphorylated tyrosine residues of TrkB; validation of selectivity of the antibody by comparing signals in sections from wild type and mutant mice selectively lacking the target is essential. An alternative to antibodies for protein detection is in situ hybridization using probes for detection of mRNA. The specificity of a probe is more readily established and this method can provide valuable information with respect to cellular localization as well as regulation of expression, albeit with mRNA as opposed to protein.