Seizing the opportunity to therapeutically address neuronal hyperexcitability in Alzheimer's disease

The global prevalence of Alzheimer's disease (AD) is estimated to double by 2060. ¹ While new therapies have gained FDA approval, ² safer and more effective disease modifying interventions are needed. People with AD are at greater risk of unprovoked seizures that worsen cognitive and functional decline. ³ Indeed, some studies suggest that the risk of unprovoked seizures is up to 87-fold higher in individuals with AD onset between 50–59 years old versus that of the general population. ⁴ Even individuals who get AD later in life have a higher chance of having seizures (hazard ratio, 8.06; 95% confidence interval, 3.23–16.61). ⁵ Older adults are the fastest growing group with epilepsy diagnosis, making seizure control in AD an untapped therapeutic opportunity to uncover novel disease modifying treatments.

Focal seizures and epileptiform activity are consistently documented in AD cases^6,7 and animal models.^8–12 However, recording spontaneous epileptiform activity or overt focal seizures in preclinical settings is labor-intensive and costly, requiring a tremendous resource and technical investment to obtain sufficient statistical power to discern pharmacologic benefit. In this regard, preclinical studies are increasingly turning to higher throughput models of evoked seizures or acquired epilepsy to expedite investigational drug screening and more efficiently assess whether novel treatments may impact disease burden in AD-associated models.^13–17 The gold-standard in epilepsy research is the recording of spontaneous recurrent seizures (SRS) and epileptiform discharges; however, this is not always feasible in early drug discovery or with costly aged rodents. Kindling is a well-established phenomenon used to model acquired epileptogenesis (the development of epilepsy) in rats and mice, which has been used for decades to advance antiseizure medicine discovery and mechanistic understanding of epilepsy.^18–24 While kindling models do have some limitations, including infrequent SRS,^25,26 they do reproduce several key pathological and behavioral features of clinical epilepsy.^27–30 Indeed, kindling models were instrumental to the discovery of one of the most impactful therapies for focal onset seizures in people with epilepsy – levetiracetam ³¹ – which is now in clinical studies for mild cognitive impairment ³² and AD.^33,34 Thus, the application of kindling paradigms to rodents with AD-related genotypes expedites the process to justify or refute further in-depth, resource-intensive evaluation of therapeutic interventions against infrequent SRS and cognitive decline.

The exact mechanisms underlying seizure generation in AD are still poorly understood. ³⁵ Neuronal depolarization with high frequency firing (i.e., a seizure) induces vesicular trafficking and glutamate release into the synapse where it interacts with both NMDA- and AMPA-type glutamate receptors to increase intracellular Ca²⁺ concentrations. High frequency neuronal firing can itself lead to cleavage of amyloid-β protein precursor (AβPP) ³⁶ to generate β- and γ-C-terminal AβPP fragments (CTFs) that underlie amyloid-β (Aβ) plaque formation. ³⁷ Indeed, several amyloidogenic AD models show increased soluble Aβ expression with repeated kindled seizures.^14,38 Extracellular Aβ induces astrocytic glutamate release, as well as blocks expression of the GLT-1 transporter that normally clears synaptic glutamate. ³⁹ As a result, excess synaptic glutamate potentiates NMDA receptor activation, propagating a cycle that drives hyperexcitability and reduces seizure threshold. Glutamate-mediated excitotoxicity in epilepsy is known to promote neuroinflammation and microglial activation,^40–43 which both likely further accelerate AD pathology. ⁴⁴ As a result, restoring glutamatergic imbalance may represent a novel therapeutic strategy to minimize AD burden. This hypothesis is further supported by the limited clinical evidence that levetiracetam administration, which is believed to interact with the SV2A protein to modulate glutamatergic vesicle release in excitatory neurons,^45,46 may beneficially improve cognitive scores in AD patients with epileptiform discharges. ³³ Thus, advanced mechanistic studies to identify anticonvulsant therapies specifically for AD reflects a novel precision medicine opportunity.

The study by Dejakaisaya and colleagues ⁴⁷ expands a growing body of evidence indicating that therapeutically targeting seizures in AD may beneficially mitigate the functional decline associated with unchecked neuronal hyperexcitability in mice with an AD-associated genetic background. Importantly, this study used a kindling model to assess mechanistic changes associated with health and disease, providing a clear demonstration of the feasibility of kindling models to expedite drug screening in AD models. Using wild-type versus the Tg2576 mouse model of AD, the investigators iteratively assessed how changes in the glutamate-glutamine cycle contributed to seizure susceptibility in both early and late life. The first component of this study quantified the cortical expression of two proteins involved in the glutamate-glutamine cycle in young Tg2576 mice (<6 months old) without a history of evoked amygdala kindled seizures. Analysis of homogenized cortical tissues by western blot demonstrated significant reductions in GLT-1 and glutamine synthetase expression in Tg2576 mice versus their age-matched littermates. The team next determined how amygdala kindled seizures evoked in aged wild-type mice changed the cortical and hippocampal expression of GLT-1 and glutamine synthetase. Interestingly, there were significant increases in expression of both markers in cortex, but GLT-1 expression was only significantly increased in cortex. Finally, the investigators applied the amygdala kindling model to aged (>12 months old) Tg2576 and wild type littermates to assess whether GLT-1 and glutamate synthetase protein expression changed with the sub-chronic administration of the antibiotic, ceftriaxone. Ceftriaxone is known to upregulate GLT-1 expression in a variety of neurological disease models and has also been previously reported to improve cognitive function in APP/PS1 mice.^48,49 Therefore, the team evaluated whether ceftriaxone administration could meaningfully delay seizure onset or normalize seizure susceptibility in seizure-prone Tg2576 mice in late life. Unfortunately, there was no beneficial anticonvulsant effect of ceftriaxone administration to aged Tg2576 mice, nor did ceftriaxone markedly impact hippocampal GLT-1 or glutamine synthetase expression in kindled Tg2576 mice. Treated and untreated Tg2576 mice developed the kindled seizure phenotype to an equivalent degree. Regrettably, the investigators did not assess the effect of ceftriaxone administration in kindled wild-type mice to know whether there was any change in seizure onset versus untreated mice, but this study does reveal that repeated ceftriaxone administration does not prevent the development of a hyperexcitable neuronal network in an aged AD-related mouse model. Thus, attempts to boost GLT-1 expression with this dose and frequency of ceftriaxone administration did not changes seizure onset in this AD mouse model.

While technical limitations muddy some interpretation, this study further highlights the feasibility of integrating kindling models to expediently assess novel therapeutic interventions for seizures in AD. This work also highlights the importance of the glutamate cycle in seizure susceptibility in AD. Sometimes more is learned from failures than from successes. In this regard, Dejakaisaya and colleagues have shown that more rigorous studies of ceftriaxone are needed to definitively demonstrate whether boosting GLT-1 expression could effectively modify disease course in AD through any specific impacts on seizures and network hyperexcitability. ⁴⁷ At present, this study provokes more questions than it answers. Yet, we may better learn from this preclinical failure that used a novel outcome measure (i.e., seizure development) than from another “successful” Aβ plaque clearing study in an AD mouse model. It is time that the field consider alternative preclinical approaches to gauge disease modification in AD. Mouse models have provided invaluable mechanistic insight into the neuropathological processes in AD; leveraging these mice in new and innovative manners to more efficiently assess seizure-related biomarkers, onset of kindled seizures, and seizure-related behavioral outcomes may ultimately bring us closer to identifying disease modifying therapies to meaningfully slow the pending tsunami of AD diagnosis in the twenty-first century.