A debate on the neuronal origin of focal seizures

A critical question regarding how focal seizures start is whether we can identify particular cell classes that drive the pathological process. This was the topic for debate at the recent International Conference for Technology and Analysis of Seizures (ICTALS) meeting (July 2022, Bern, CH) that we summarize here. The debate has been fueled in recent times by the introduction of powerful new ways to manipulate subpopulations of cells in relative isolation, mostly using optogenetics. The motivation for resolving the debate is to identify novel targets for therapeutic interventions through a deeper understanding of the etiology of seizures.

1 | VIEWPOINTS 1.1 | Evidence that interneurons trigger seizures A natural starting point in the debate is to ask which neurons are the earliest to be locally recruited when seizure activity begins to build up (Table 1).In many different instances, this turns out to be the interneuronal population, which can be defined based on morphology, connectivity, or molecular or electrophysiological properties; we refer here to γaminobutyric acid (GABA)ergic neurons originating from the medial and caudal ganglionic eminences. 1n rodent hippocampal slices from non-epileptic animals, seizure-like events may be elicited by increasing excitability using the potassium channel blocker 4-aminopyridine (4-AP).Interneurons increase their firing frequency during the seconds preceding the ictal event and during its initial phase characterized by fast, low-voltage activity, 2,3 whereas pyramidal cells are initially silent and discharge later with a "tonic" firing that rapidly develops into recurrent bursts during the body of the seizure.A similar order of recruitment, with interneurons active first, is seen in multiple other cases including ictal-like events triggered by electrical stimulation in hippocampus, 4 in the entorhinal cortex of the whole guinea pig brain preparation, a model in which seizures are triggered by weakening (but not fully blocking) inhibition, and also from microelectrode recordings of spontaneously occurring seizures in humans 5 (see Table for other examples).Interestingly, even in a preparation in which seizure-like activity is induced ostensibly by enhancing excitation-the 0 Mg 2+ model-as the wavefront invades new territories, interneuronal activation precedes pyramidal firing. 6f, however, all one has is the pattern of recruitment, the interpretation is not always straightforward.First, can one be sure all the relevant neurons are sampled, to be able to identify which ones were active first?Might there be, for instance, glutamatergic drive from elsewhere projecting into the local recording site?][9] The first question can be addressed using optogenetics, because the initial pattern of activation can be targeted in space and to a specific cell class (for potential caveats, see concluding remarks).It is particularly noteworthy, therefore, that interneuron-specific optogenetic activation often leads to seizures.This has been demonstrated by activating either all interneurons vesicular GABA transporter (VGAT) promoter of the superficial two thirds layers of the somatosensory cortex, 10 or the parvalbumin (PV, entorhinal cortex 11 or hippocampus 12 ) or somatostatin interneurons (SOM, in entorhinal cortex 13 ), in tissue pre-treated with either 4-AP [11][12][13] or 0 Mg 2+ 10,14 solutions (note, these two experimental treatments differ with respect to ictogenesis, 15 so it is relevant that in the 0 Mg 2+ preparation, interneuronal ictal initiation was seen only after many seizure-like events had happened, 10 and not at early stages 14 ).
Various pharmacological evidence also supports the idea that interneurons drive ictogenesis.Hyperpolarization (and, therefore, functional inhibition) of interneurons by the μopioid agonist, Tyr-D-Ala-Gly-MePhe-Gly-olenkephalin, blocks seizure-like events in epileptic temporal lobe human tissues in vitro 16 or in human periglioma cortices. 17Similarly, spontaneous epileptiform discharges in surgically resected human hippocampal brain slices were suppressed by blocking GABAergic receptors 18 ; note, however, that these events were also blocked by glutamatergic antagonists.Finally, there are clinical case reports of barbiturates triggering seizures 19,20 and they also increase seizure-like activity in 4-AP-treated human brain slices. 16On the other hand, these specific examples need to be counter-balanced by the large body of clinical evidence regarding the anti-epileptic use of both barbiturates and benzodiazepines, which work by boosting GABAergic currents.Indeed, these drugs remain the first-line treatment for status epilepticus.
Given these various data on the ictogenic properties of interneuronal output, how does this happen?0][31][32] KCC2 operates very close to its equilibrium, meaning that substantial shifts in either Cl − or K + can flip the direction of transport.Because the extracellular space is effectively shared between all cells, a rise in extracellular K + will thus have the effect of coordinating raised intracellular Cl − across the neuronal population. 33,34][37][38] Another possible mechanism is that synaptic inhibition sets sharp boundaries on when postsynaptic cells may fire, and so coordinated inhibition leads in turn to synchronization of the postsynaptic population. 9,10,39This has been demonstrated most convincingly for physiological gamma oscillatory activity, and facilitates the spread of neuronal activity through particular pathways within networks.It is interesting to note that coordinated inhibition may be disrupted when there is pathological chloride loading, since theoretical work suggests that a positive shift in E GABA creates jitter in the firing 25 (in addition to increasing neuronal activity).Notably, this has been documented in the immediate pre-ictal period, 40 and has been postulated to be a potential electrophysiological biomarker of chloride loading. 41A key point is that the synchronizing effect of the GABAergic drive depends upon the intracellular Cl − concentration of the postsynaptic population. 25f interest GABAergic function may vary dramatically across the neuropil or even within intracellular domains. 42,43In rodent hippocampal slices in which seizure-like events are triggered by N-methyl-d-aspartate (NMDA) puffs in the 4-AP and 0.5 mM Mg 2+ milieu, optogenetic activation of PV interneurons triggers ictal events in the seizure focus but decreases seizure duration and limits its propagation when interneurons are activated in the surrounding areas. 12ne final point to make at this juncture is that two recent studies have reported substantial physiological dynamics in baseline intracellular Cl − concentration in pyramidal cells, amounting to an ~19 mV shift in E GABA between the rodent's active (nightime) and rest (daytime) phase. 44,45Both studies suggest, therefore, a more nuanced view of raised Cl − concentration that does not automatically equate to a pathological state.

| Evidence that excessive glutamatergic activity triggers seizures
This opposing debate position during the discussion at ICTALS is less about pinpointing seizure initiation specifically to pyramidal neurons, but more about taking some of the blame off interneurons.7][48][49] While accepting the observation that interneurons are often highly active at seizure onset, 29,50 what does this really mean?0][61][62][63] Inhibitory failure, on the other hand, appears to be critical for ictal progression 49,52,53,55,[64][65][66] ; when a sustained barrage of excitatory drive causes the postsynaptic cells to become chloride loaded, this erodes the feed-forward inhibitory effect.Depolarization block in interneurons may also contribute to inhibitory failure 3,59,67,68 (note that depolarization block also occurs in pyramidal cells 69,70 ), although it remains unclear if it occurs in the absence of chemoconvulsants.The important point to recognize, however, is that the appearance of the activity patterns depends upon the exact recording site (at the mm to sub-mm scale [7][8][9]71,72 ), and instances of interneurons both initiating 73 or opposing 9,51,52,68 ictal activity may appear identical.
Earlier we mentioned examples where interneuronal optogenetic activation triggers seizures, but seizures can equally be triggered by a purely glutamatergic drive too, [74][75][76][77][78][79] which leaves open the question of a secondary activation of interneurons.In addition, we should not forget that seizures can be triggered by sensory inputs (audiogenic or photogenic epilepsies), where the thalamocortical projection system delivers glutamatergic drive into the neocortex to trigger the seizure. 80,81t the neuronal level, experimental evidence suggests that seizures can be triggered by activation of inhibitory, excitatory, or mixed neuronal populations 10,11,13,14,34,77,78,[82][83][84][85] (Table 1), but also that the prior state of the tissue and the presence of additional chemoconvulsive agents may be critical. 757][88][89][90] In the absence of inhibition, even a brief stimulus of a single CA3 pyramidal neuron can trigger short local population bursts. 76,91,92n chronic focal epilepsy, enhanced CA2-CA1 excitatory connectivity may drive seizure occurrence, 58 and excitatory dentate gyrus granule cell to CA3 synapses have even been termed "potential detonators." 93In vivo, inhibition of dentate gyrus mossy cells or CA2 pyramidal neurons reduces seizure occurrence in pilocarpine-or kainate-induced chronic epilepsy, 58,77,90 while selectively activating mossy cells is ictogenic. 77,90ollectively, the present body of literature points toward an involvement of excitatory and inhibitory cell types in ictogenesis, and a complex, compartmentdependent (e.g.5][96][97] An important recent insight into the specific driving forces toward the ictal tipping point has been to recognize various positive feedback mechanisms at work. 34For instance, neuronal activation causes both extracellular [K + ] and intracellular [Cl − ] to rise, each of which makes further neuronal activation more likely.Another factor is the bistable nature of pyramidal activation, depending on whether or not dendritic plateau potentials occur.These are mediated by either NMDA receptors or voltage-gated Ca 2+ channels, so it is pertinent that influential cellular models of both cortical 91 and thalamic 98 seizures have been presented involving these conductances.It is particularly significant, therefore, that a simple optogenetic assay of dendritic excitability, involving pyramidal activation only, reliably showed a step change in the network response occurring shortly before the onset of seizure-like activity in a variety of different acute models of ictogenesis. 34This was modeled in terms of the synergistic interplay of all the factors mentioned here: elevated [K + ] o and [Cl − ] I , defective inhibition, dendritic excitability, and bursting behavior of the pyramidal population. 33,34

| Concluding remarks
It goes without saying that experiments inherently simplify complex real-world phenomena, and both experimenters and enthusiastic readers need to consider carefully the advantages and shortcomings of each respective study.For example, although optogenetics have undoubtedly allowed insights into the inner workings of ictogenic neural networks with unprecedented neuroanatomical precision, differences in optogenetically targeted population size (e.g., local vs wide-field) or stimulation paradigms (e.g., fixed frequency at 10 or 20 Hz, or continuous illumination) do not exactly recapitulate the nature of spontaneously occurring seizures in chronically epileptic animals or humans, or the spontaneous neuronal activation patterns.Attention must also be paid to the pattern of opsin expression.Thus, although the Cre/loxP system, in principle, allows celltype specific interference, expression may be "leaky" and include off-target cell types, potentially affecting measured effects. 99Furthermore, seemingly homogeneous cell classes (e.g., PV) may consist of sub-types with different circuit profiles (e.g.PV-pyramidal or PV-PV). 49,100,101Finally, the majority of studies on the cellular origin of seizures have been carried out in brain slices.Such preparations have obvious limitations, lacking global brain connectivity, and may also have compromised neuromodulatory systems and network homeostasis.][111] The debate at ICTALS focused primarily on the relative involvement of interneurons and pyramidal cells, right back to the early view that seizures arise either from an excess in excitatory glutamatergic activity or a deficit of GABAergic inhibition. 112Of course, these options are not mutually exclusive and, when both occur, the effects may be synergistic.However, although the dedicated debate at the conference was specified to address the relative contributions of neuronal subtypes, we should bear in mind the role of altered extracellular ion dynamics and non-neuronal cells in the pathological processes that lead to the initiation and the development of focal seizures.Besides the known role of astrocytes in network hyperexcitability, 48,[113][114][115] a number of publications underscore that microglial population states, 46,116,117 pathologically altered oligodendrocyte distribution, and demyelination processes 47,118 shape neuronal excitability and the severity of seizures.Thus seizures are not the result of an aberrant communication of excitatory and inhibitory neurons alone, but a complex interaction of different brain cell types constituting the ictogenic network.
Finally, it is important to recognize that the precise timing of a seizure is dictated by the occurrence of a specific trigger, but also permissive factors, in the same way that a forest fire occurs only if the wood is dry before the spark is applied.[121][122] AUTHOR CONTRIBUTIONS G.H. and M.W. gave short presentations on the two viewpoints reviewed herein, at an on-stage debate on the cellular origin of focal seizures at ICTALS 2022 in Bern, Switzerland.The debate session was moderated by D.B.G.All authors who contributed to this review attended the debate session and actively contributed to the discussion.The article was written by M.W., G.H., and A.J.T; the table was compiled by M.d.C.; and all authors edited the manuscript.

Model Examined area Primary involvement at onset of focal seizures Glutamatergic neurons GABAergic interneurons Ref # doi
Primary involvement of neuronal subtypes at onset of focal seizures.An overview of publications that include recordings and/or manipulation of subtype-specific neuronal activity in the context of focal seizure onset, intended as a foundation for researchers interested in the topic of the cellular origin of focal seizures.Note that the provided overview is non-exhaustive and simplified to some extent, as some studies use multiple or combined approaches, and sub-type specific interactions during ictogenesis are complex (see main text).